#  @initlayers initlayers initlayers posts on X about the next, llm, the project, ai the most. They currently have [-----] followers and [---] posts still getting attention that total [---] engagements in the last [--] hours. ### Engagements: [---] [#](/creator/twitter::1935724547451547649/interactions)  - [--] Week [------] +4,275% - [--] Month [------] -21% - [--] Months [-------] +16,986% ### Mentions: [--] [#](/creator/twitter::1935724547451547649/posts_active)  - [--] Months [--] +41% ### Followers: [-----] [#](/creator/twitter::1935724547451547649/followers)  - [--] Week [-----] +9% - [--] Month [-----] +20% - [--] Months [-----] +3,716% ### CreatorRank: [---------] [#](/creator/twitter::1935724547451547649/influencer_rank)  ### Social Influence **Social category influence** [technology brands](/list/technology-brands) [finance](/list/finance) [stocks](/list/stocks) [social networks](/list/social-networks) [countries](/list/countries) [celebrities](/list/celebrities) [products](/list/products) **Social topic influence** [the next](/topic/the-next), [llm](/topic/llm), [the project](/topic/the-project), [ai](/topic/ai), [oops](/topic/oops), [finish](/topic/finish), [woke](/topic/woke), [if you](/topic/if-you), [update](/topic/update), [night](/topic/night) **Top assets mentioned** [Spotify Technology (SPOT)](/topic/$spot) ### Top Social Posts Top posts by engagements in the last [--] hours "@perceptron_cnn What is the theme" [X Link](https://x.com/initlayers/status/1939377905739444459) 2025-06-29T17:38Z [--] followers, [--] engagements "Okay. I did not get shortlisted in Adobe India Hackathon. And to be very honest I am not even surprised am I lost chat 🥲" [X Link](https://x.com/initlayers/status/1945221144237265132) 2025-07-15T20:37Z [--] followers, [--] engagements "@Pseudo_Sid26 I think total access to the codebase should be given only after 60% of the payment is done 🙂 there are a lot of scamsters in freelancing. But it depends" [X Link](https://x.com/initlayers/status/1945480943772336537) 2025-07-16T13:49Z [--] followers, [--] engagements "Day 1: We'll kickstart the project. The prompt is [---] words (5345 characters). This should be enough to get the basic project done. I'll just keep refining this. I am trying perplexity let's see how there routing system works I am excited" [X Link](https://x.com/initlayers/status/1946489544729354571) 2025-07-19T08:37Z [--] followers, [--] engagements "Day [--] Progress (20/07/2025): Designed a systematic protocol for hyperparameter tuning Took lr=1e-3 and epochs=50 as baseline for all the models. Trained [--] times to produce 87.5% validation accuracy for MFCC model 1/2" [X Link](https://x.com/initlayers/status/1947044574360392027) 2025-07-20T21:22Z [--] followers, [--] engagements "Tweaked parameters [--] times to achieve 98.6% accuracy on mel Ran the model [--] times and did early stopping at epoch [---] to reach 97.2% on fusion model Perfomance ranking: Mel Fusion MFCC Finalized the results and prepared a dedicated section for the paper. 2/2" [X Link](https://x.com/initlayers/status/1947044586305864072) 2025-07-20T21:22Z [--] followers, [--] engagements "@sageadvik Used ChatGPT for [--] months. It sucks at many things. I am using this Perplexity pro which I got for free recently. I am liking the fresh responses. It doesn't care about what I feel it just does the work. ChatGPT focuses more on what I am saying rather than actually working" [X Link](https://x.com/initlayers/status/1947394637225988312) 2025-07-21T20:33Z [--] followers, [--] engagements "Day [--] Progress (21/07/2025): Freezed my hyper parameters for each model. Employed early-stopping to improve efficiency and save compute Enabled GPU and installed new packages causing epoch shifts. Tried to roll back to the previous version of the codebase but failed" [X Link](https://x.com/initlayers/status/1947404212385812896) 2025-07-21T21:11Z [--] followers, [--] engagements "Apparently [--] epoch shift is acceptable This occurs due to inherent stochasticity in DL training Finalized the results for the paper Employed 5-fold CV for best model of each MFCC Mel & Fusion CV takes too much time Still awaiting thermal image data " [X Link](https://x.com/initlayers/status/1947404215451849145) 2025-07-21T21:12Z [--] followers, [--] engagements "@adxtya_jha @perplexity_ai The interface is shit. Interpretability is non existent. Finally an AI that doesn't care about me abusing. I don't know if I should be happy for that. Atleast I am getting free deep research. I would still prefer ChatGPT if I was allowed limited deep research on both" [X Link](https://x.com/initlayers/status/1947584674240548916) 2025-07-22T09:09Z [--] followers, [---] engagements "@adxtya_jha @perplexity_ai The responses are fresh and it follows the instructions mostly. ChatGPT research is good when it understands the prompt and you ask for the first time. One more try and it changes it's faith. I started to believe GPT hates being asked twice" [X Link](https://x.com/initlayers/status/1947586133917114448) 2025-07-22T09:14Z [--] followers, [--] engagements "Day [--] Progress (22/07/2025): Performed [--] fold cross validation on the best model of mel mfcc and fusion Mel (98.6%) Fusion (97.2%) MFCC (875%) in peak single-run accuracy Fusion exhibits the highest average cross-validated accuracy (96.8% 2.8%)" [X Link](https://x.com/initlayers/status/1947706827321577942) 2025-07-22T17:14Z [--] followers, [--] engagements "@purusa0x6c None. We are busy vibe coding stuff that has been made [---] times already but with a minor tweak" [X Link](https://x.com/initlayers/status/1950967151353012622) 2025-07-31T17:09Z [--] followers, [--] engagements "I wanted to perform LOSO Cross Validation to get better insights for paper. But the problem is I have [--] subjects and I didn't want to manually run for each model so I made one single code. Now this will do [--] folds validation on [--] models 💀 I am leaving it to run overnight " [X Link](https://x.com/initlayers/status/1951745989003620751) 2025-08-02T20:44Z [--] followers, [---] engagements "@Maaztwts Spotify's music recommendation algorithm is still top notch" [X Link](https://x.com/initlayers/status/1951877725464956935) 2025-08-03T05:28Z [--] followers, [--] engagements "@Maaztwts They made their services exclusive to subscribers. It was still good for normal users some years ago. Spotify was the first subscription I ever purchased. I liked it so much but now it's different" [X Link](https://x.com/initlayers/status/1951885259504525383) 2025-08-03T05:58Z [--] followers, [--] engagements "After [--] hours of running the script overnight. I got the results for all the [--] models. And the results are shocking 💀 mel and fusion wre constantly beating eachother in different analyses" [X Link](https://x.com/initlayers/status/1951901311110459581) 2025-08-03T07:01Z [--] followers, [--] engagements "I had a nightmare last night. It stopped when my alarm rang at 5:30 AM. The story setting was IIT Guwahati and I am responsible for a civil war. I had to run away from two factions fighting eachother with grenades and all sort of stuff. I saw the majestic Brahmaputra turn red" [X Link](https://x.com/initlayers/status/1952182704394600612) 2025-08-04T01:40Z [--] followers, [--] engagements "@iamAdityaAnjana We got Pattern Recognition and Machine Learning Natural Language Processing Intelligent Data Analysis Computational Complexity Cryptography and Network Security. Labs for PRML NLP and IDA" [X Link](https://x.com/initlayers/status/1952275329718038725) 2025-08-04T07:48Z [--] followers, [--] engagements "Subjects for this semester: Natural Language Processing Pattern Recognition & Machine Learning Intelligent Data Analysis Computational Complexity Cryptography and Network Security Minor Project [--] based on Industrial Training Labs: NLP PRML IDA" [X Link](https://x.com/initlayers/status/1955267666106622302) 2025-08-12T13:58Z [--] followers, [---] engagements "@freshlimesofa You are doing good. No fixes needed" [X Link](https://x.com/initlayers/status/1955675046917427379) 2025-08-13T16:57Z [--] followers, [--] engagements "I am thinking of learning Web Design. I started with a [--] hour crash course from flux academy. However most of it is bouncing of my head. I don't know if I am learning anything. Can someone please help me with this" [X Link](https://x.com/initlayers/status/1956022324261740950) 2025-08-14T15:57Z [--] followers, [--] engagements "The mid-term is finally done. From tomorrow it's grind time kicking off my productivity arc and pushing my limits. I'll try to update here everyday consistently" [X Link](https://x.com/anyuser/status/1967272881689342435) 2025-09-14T17:02Z [----] followers, [---] engagements "@sharpeye_wnl I just don't trust them in the long term. Even a Samsung product will work but not Xiaomi" [X Link](https://x.com/initlayers/status/1969270162122498201) 2025-09-20T05:19Z [---] followers, [--] engagements "@daddy__broccoli The urge to leave everything aside and learn assembly is getting stronger with each passing day" [X Link](https://x.com/initlayers/status/1969287175909892411) 2025-09-20T06:27Z [---] followers, [--] engagements "Tomorrow is the deadline I have to build an MVP ASAP else NGMI. Our presentation will include pictures of the UI so I am the bottleneck now this night will be legendary 💀" [X Link](https://x.com/anyuser/status/1972651600088088918) 2025-09-29T13:16Z [----] followers, [---] engagements "@Arpit_2023 Is this that difficult I did the same using ChatGPT it was pretty smooth and fast" [X Link](https://x.com/initlayers/status/1974697731152285929) 2025-10-05T04:46Z [---] followers, [--] engagements "@X_Ibyte @Pseudo_Sid26 It is good but pain the *. Run time is too big but the main problem is I never get optimal parameters with it. Maybe I am using it wrong or something I always fallback to Regular Grid Search" [X Link](https://x.com/initlayers/status/1974723169803227334) 2025-10-05T06:27Z [---] followers, [--] engagements "@X_Ibyte @Pseudo_Sid26 We should still define the range right Maybe issues there how many combinations do you let it try" [X Link](https://x.com/initlayers/status/1974724523732955641) 2025-10-05T06:33Z [---] followers, [--] engagements "@Arpit_2023 What is RTC 😅" [X Link](https://x.com/initlayers/status/1974789657306214449) 2025-10-05T10:51Z [---] followers, [--] engagements "@Ndevsayzz Tf what else are you using bruh" [X Link](https://x.com/initlayers/status/1974834231735201861) 2025-10-05T13:49Z [---] followers, [--] engagements "@farhanxansari Al the best bro" [X Link](https://x.com/initlayers/status/1977416302345302128) 2025-10-12T16:49Z [---] followers, [--] engagements "15 October [----] Selected a PS built a prototype and recorded a demo video for SIH (took around [--] hours) Revised filtering layer nomenclature" [X Link](https://x.com/anyuser/status/1978550976165376447) 2025-10-15T19:58Z [----] followers, [---] engagements "ML Engineer Internship - Strong Candidate Sharing the profile of Vaibhavi Thakur (@Messing_up20023) a pre-final year ML engineer who had been building production grade ML systems with a strong focus on data reliability label quality and evaluation loops. She has: Built end-to-end ML systems Worked on RAG systems and reliability-focused ML infrastructure Co-authored a research paper on Multimodal Fatigue Detection (NIT Rourkela) Hands-on experience shipping real systems. Vaibhavi is actively looking for ML/AI Engineering internships and would be a great fit for teams working on real-world ML" [X Link](https://x.com/initlayers/status/2007074242664567041) 2026-01-02T12:59Z [----] followers, 67.6K engagements "19 October [----] Studied pre-training fine-tuning types of fine-tuning and stages of building LLMs Tried to understand Transformers Architecture tokenization self attention vector embeddings BERT vs GPT Project: confirmed the stack we will use free tools/models" [X Link](https://x.com/anyuser/status/1980000258768904534) 2025-10-19T19:57Z [----] followers, [---] engagements "Instagram and TikTok are in a league of their own when it comes to algorithms. YouTube Spotify and X follow behind but none match how fast Instagram learns your mood. It's scary how good it's become at keeping you watching. Even though it's affecting attention spans and careers it remains a masterpiece both algorithmically and psychologically" [X Link](https://x.com/anyuser/status/1981734594111303741) 2025-10-24T14:48Z [----] followers, [---] engagements "24 October [----] Revised LLM basics stages of building LLMs OOPs - mostly around classes and objects Project: Finally resumed. I built core modules to test project feasibility. I then downloaded Ollama and tested TinyLlama locally. Obviously it's not very great but with my GPU and RAM I am satisfied with anything I can possibly get. The next couple of days will be intense. I will try to code stuff from scratch" [X Link](https://x.com/anyuser/status/1981835386201727448) 2025-10-24T21:29Z [----] followers, [---] engagements "When my cousin was [--] years old we realized he wasn't like other kids. While most children played with toys he spent hours flipping through books literally any book he could find. His parents noticed this early and encouraged him giving him every resource they could get their hands on. Then something incredible happened. By the time he turned [--] he could speak Marathi fluently. No one in the family spoke Marathi which made it even stranger. And one day he sat down at a piano and started playing simple melodies perfectly. That's when his parents realized something profound. Even though each book" [X Link](https://x.com/anyuser/status/1982095329273983093) 2025-10-25T14:42Z [----] followers, [---] engagements "@farhanxansari It is notttttttt" [X Link](https://x.com/initlayers/status/1982103543525122237) 2025-10-25T15:14Z [---] followers, [--] engagements "@dv_zsh ChatGPT is also trained on data from stackoverflow. So in a way or so we will continue to rely on it" [X Link](https://x.com/initlayers/status/1982193746805407765) 2025-10-25T21:13Z [---] followers, [--] engagements "@OjasSharma276 Bruh wdym I am standing right next to you why talking to the wall" [X Link](https://x.com/initlayers/status/1982196726355165500) 2025-10-25T21:25Z [---] followers, [--] engagements "@bit2swaz I dunno it is looking monotonous right now. Keep generating you might get a better one" [X Link](https://x.com/initlayers/status/1982305251056042065) 2025-10-26T04:36Z [---] followers, [--] engagements "Today I'll be doing two lectures of LLMs. Will try to code as much as I can. I'll complete the OOPs revision and study DVC. I'll also complete the UI for my project. It will be a big day. Let's go" [X Link](https://x.com/anyuser/status/1982308978391687512) 2025-10-26T04:51Z [----] followers, [---] engagements "26 October [----] Studied the evolution of GPT from transformers to GPT-5. A brief understanding of what's changed Revisited the working of GPT-3 Studied different types of tokenizations schemes Byte Pair Encoding Finally completed revising OOPs: Static method encapsulation setter and getter methods Project: Pushed everything to git. Created a streamlit UI using AI. Everything worked well. I'll take [--] more days and launch it here. Not satisfied with the grind 😕" [X Link](https://x.com/anyuser/status/1982568154607067143) 2025-10-26T22:00Z [----] followers, [---] engagements "@LowKeyDevs Learn Python no matter what your main language is. It's like learning English it opens you up to a world of knowledge beyond your circle" [X Link](https://x.com/initlayers/status/1982681222716506287) 2025-10-27T05:30Z [---] followers, [--] engagements "@krishdotdev Selective consumerism and extremely high product consciousness. India has all of it. Accounting for each rupee spent" [X Link](https://x.com/initlayers/status/1983113880084922534) 2025-10-28T10:09Z [---] followers, [---] engagements "Ever wondered what people mean when they say "context window" in LLMs Think of an LLM like a human with limited short-term memory. The context window is that memory span - it decides how much text the model can "remember" while generating a response. GPT-3 had a context windows of 2K-4K tokens. GPT-4 Turbo 128K. Claude [---] Even more. Larger context = more room to think recall and reason. You can feed entire books conversations or codebases and still get coherent answers. But here's the catch - a bigger window doesn't mean perfect memory. LLMs don't store information like humans they re-read" [X Link](https://x.com/anyuser/status/1983390838764851374) 2025-10-29T04:30Z [----] followers, [---] engagements "@OjasSharma276 I hope you are okay. What did you apply for though Like role" [X Link](https://x.com/initlayers/status/1983444577823629336) 2025-10-29T08:03Z [---] followers, [--] engagements "One week of using a premium account. It has been good not the best. I'll keep analyzing what I can improve over the next [--] weeks. Thanks for all the engagement 😊" [X Link](https://x.com/anyuser/status/1983539577324552410) 2025-10-29T14:21Z [----] followers, [---] engagements "Your text doesn't go straight into the model. It first passes through a tokenizer - a system that breaks words into small chunks called tokens. Example: understanding - under + standing LLMs don't read words. They read numbers" [X Link](https://x.com/initlayers/status/1983775019294667249) 2025-10-30T05:56Z [---] followers, [--] engagements "Once tokenized your request turns into a long list of numbers. This is sent to OpenAI's servers (or whichever model host you're using). Every message carries context - everything you typed so far. That's how ChatGPT "remembers" your conversation" [X Link](https://x.com/initlayers/status/1983775036847861808) 2025-10-30T05:56Z [---] followers, [--] engagements "Now the real magic begins. Inside giant clusters of GPUs the model looks at your tokens and predicts what comes next. one token at a time. The entire essay you got back Generated word by word. Prediction by prediction" [X Link](https://x.com/initlayers/status/1983775043743313952) 2025-10-30T05:56Z [---] followers, [--] engagements "Each token depends on everything before it. That's why longer prompts slow it down - it's carrying your whole chat history each time. This running memory is called the context window. GPT-4 Turbo can handle up to [------] tokens (around a 300-page book)" [X Link](https://x.com/initlayers/status/1983775053100839169) 2025-10-30T05:56Z [---] followers, [---] engagements "As each token is predicted it's streamed back to you in real time. That's why you see words appear one by one. You're literally watching the model "think."" [X Link](https://x.com/initlayers/status/1983775072973410695) 2025-10-30T05:56Z [---] followers, [--] engagements "When the reply ends your entire chat (your message + its answer) is stored temporarily for continuity. Once the context window fills up older parts are forgotten. ChatGPT doesn't have memory. It has context" [X Link](https://x.com/initlayers/status/1983775082515460495) 2025-10-30T05:56Z [---] followers, [--] engagements "@dv_zsh Bro's aura is unmatchable" [X Link](https://x.com/initlayers/status/1984294272120803568) 2025-10-31T16:19Z [---] followers, [--] engagements "@kmeanskaran That's great I have spent a lot of time building strong fundamentals. If you are looking for ML or LLM Engineer roles I'd be a good fit" [X Link](https://x.com/initlayers/status/1984699933270950049) 2025-11-01T19:11Z [---] followers, [---] engagements "November [----] Goals: Complete 2nd MLOps project Update resume & start applying Land an ML Engineer internship Finish MLOps lectures (3/25 done) Grow to 1K followers on X (470 now) & [---] verified (160 now) Complete half of the LLM lectures (currently at 9/43 done) Win a hackathon Score 75%+ in mid-terms This month will be intense. Mid-terms in week [--] lectures piling up projects and hackathons waiting. Somehow I've got to balance it all and land that internship and connect with more people here. Also. it's my birthday month 😊" [X Link](https://x.com/anyuser/status/1984851364250599783) 2025-11-02T05:13Z [----] followers, [----] engagements "@aaravshukla_10 Give ChatGPT all your details and ask for a good resume in LaTeX format. Put it on Overleaf. Keep iterating your code using GPT. It will probably take 1-2 days. But you'll end up with a very high quality resume" [X Link](https://x.com/initlayers/status/1984880884827505097) 2025-11-02T07:10Z [---] followers, [--] engagements "@aaravshukla_10 Yes. You can post it here for further feedback. We have an incredible community" [X Link](https://x.com/initlayers/status/1984882436665459170) 2025-11-02T07:17Z [---] followers, [--] engagements "I am working on an open-source RAG system focused on efficient hybrid retrieval and evaluation. It's been [--] days since I started the project. With mid-terms coming up next week I've decided to finish it tonight no matter what. I will be working for the next [--] hours taking little breaks. Let's see how far we can go" [X Link](https://x.com/anyuser/status/1984908480986116541) 2025-11-02T09:00Z [----] followers, 20K engagements "@AkwaIbomBreed @harshalvk_ Learn Git/GitHub OOPs Data Versioning using DVC CI using mlflow Dockerization CD on AWS Kubernetes Prometheus and Grafana. Just learn these concepts individually from any tutor. That should be a great start" [X Link](https://x.com/initlayers/status/1985336005654905016) 2025-11-03T13:19Z [---] followers, [--] engagements "The most dangerous thing about AI isn't that it lies - it's that it lies confidently. You ask a model something simple. It replies instantly flawlessly persuasively and yet it's wrong. That's called a hallucination. Large Language Models don't know things. They don't store facts or check truth. They just predict the next word based on patterns they've seen before. So when the model doesn't know something it doesn't say "I'm not sure." It guesses. But it guesses beautifully. It's like that one kid in class who didn't study but still wrote a full essay with perfect grammar and total confidence." [X Link](https://x.com/anyuser/status/1986092585337974927) 2025-11-05T15:25Z [----] followers, [----] engagements "@autom8nerd Haha. I remember that. I don't know how these things work but I feel using RL for such tasks is better than training under supervision" [X Link](https://x.com/initlayers/status/1986116762077896832) 2025-11-05T17:01Z [---] followers, [--] engagements "It's my 🎈 day - just turned [--]. My first birthday on X. I've met many awesome people here since I joined in June. Hopefully by my next birthday you'll look at me as someone who adds real value here" [X Link](https://x.com/anyuser/status/1986511458457518286) 2025-11-06T19:10Z [----] followers, [----] engagements "@OjasSharma276 @RockstarGames 🤣🤣" [X Link](https://x.com/initlayers/status/1986642865641169120) 2025-11-07T03:52Z [---] followers, [--] engagements "8 Nov [----] IDA Mid Prep: Clustering Analysis K-means K-mediods Agglomerative Divisive Distance Metrics for Clustering Algorithms DB Scan Evaluation of Clustering PRML Mid Prep: Perceptron Algorithm Support Vector Machine Hard and Soft Margin SVM" [X Link](https://x.com/anyuser/status/1987257439625122249) 2025-11-08T20:34Z [----] followers, [----] engagements "9 November [----] PRML Mid Prep: Kernel Function Linear Fisher Polynomial Gaussian Radial Basis Function Cosine Two-class problem Hard constraint SVM Soft Margin SVM. Linear models for classification Discriminant functions fisher's discriminant Probabilistic discriminative and generative models" [X Link](https://x.com/anyuser/status/1987601725470245343) 2025-11-09T19:22Z [----] followers, [---] engagements "12 November [----] CNS Mid Prep: RSA Encryption Diffie-Hellman Key Agreement SHA-512 Algorithm Kerberos Password Management SSL Architecture SignCryption NLP Mid Prep: Discourse Analysis Dialogue Analysis QA System Rule Based and Probabilistic Parsing NER" [X Link](https://x.com/anyuser/status/1988703649426821400) 2025-11-12T20:21Z [----] followers, [---] engagements "@Sarthak4Alpha @sachinyadav699 San Francisco" [X Link](https://x.com/initlayers/status/1989794342815633533) 2025-11-15T20:35Z [---] followers, [--] engagements "@devcelio How come I never heard of this. This is awesome" [X Link](https://x.com/initlayers/status/1990109086999052772) 2025-11-16T17:25Z [---] followers, [----] engagements "I've delayed the project launch again this time it's because of the front-end. The entire repository is ready but I used Bolt for the front-end and the output looks way more polished than my original version. Now I'm stuck in a dilemma: should I launch the boring current version or take the pain update the front-end properly and then release it" [X Link](https://x.com/initlayers/status/1990515207891030346) 2025-11-17T20:19Z [---] followers, [---] engagements "19 November [----] Token embeddings how are they created for LLMs coding embedding layers intuition behind embedding layers and linear neural networks" [X Link](https://x.com/initlayers/status/1991231517587079603) 2025-11-19T19:46Z [---] followers, [---] engagements "If you want to learn the Theory of computation this is the best book I've come across: "Theory of Computer Science (Automata Languages and Computation)" by K.L.P. Mishra and N. Chandrasekharan. @philearning I used it myself. The explanations of automata grammars turing machines decidability and complexity are very clear. You don't need to be a hardcore theory person; you can open any chapter and understand the concepts without getting lost. And if you're into AI/ML TOC actually matters more than people think: Automata helps you understand how parsers and tokenizers work. Grammars relate" [X Link](https://x.com/initlayers/status/1991460217415168381) 2025-11-20T10:54Z [---] followers, [---] engagements "I just promoted Nano Banana 🍌 Pro: "Visualise how LLMs work from the very foundations as an academic image that can be used for illustration purposes. I need neatly labelled descriptions." And here's the output. Honestly I'm impressed with what it generated. Seeing an AI generated image with zero typos actually made my day" [X Link](https://x.com/initlayers/status/1991748476913557929) 2025-11-21T06:00Z [---] followers, [---] engagements "@autom8nerd Yes. ChatGPT sucks at image generation. It produces inaccurate images with a lot of typos" [X Link](https://x.com/initlayers/status/1991790479944630501) 2025-11-21T08:47Z [---] followers, [--] engagements "@bit2swaz Yes. I am finding this model good. I will test it more" [X Link](https://x.com/initlayers/status/1991790750678536681) 2025-11-21T08:48Z [---] followers, [--] engagements "@karanjagtiani04 I didn't have to version the embeddings. It's plug and play just swap components and get the results" [X Link](https://x.com/initlayers/status/1991792692720963620) 2025-11-21T08:55Z [---] followers, [--] engagements "@autom8nerd Yes agreed. I presume there must be some way to deal with that as well" [X Link](https://x.com/initlayers/status/1991857959232819694) 2025-11-21T13:15Z [---] followers, [--] engagements "@shivamxtech I got many more. Stay tuned 😉" [X Link](https://x.com/initlayers/status/1992166122595217635) 2025-11-22T09:39Z [---] followers, [--] engagements "@Rushu_Tushu They're still general LLMs. They just get way more code training and coding-specific optimization. That's why they write code well and chat normally" [X Link](https://x.com/initlayers/status/1992260503377506548) 2025-11-22T15:54Z [---] followers, [--] engagements "@devsterxyz That's some good intuition. It's actually interesting if you go a little deeper" [X Link](https://x.com/initlayers/status/1992331505037938741) 2025-11-22T20:37Z [---] followers, [--] engagements "I really pushed my limits for this project. It took a lot of effort. Since RAG is trending right now I picked up that direction and built a system that combines keyword search and semantic search. My first step toward grounding LLMs to a real knowledge base and reducing hallucinations. Please check it out and comment on what you feel about it. I spent the last few weeks building VectorFlow-RAG a full semantic search + QA system because I wanted to understand what it actually takes to ship an ML system end-to-end. Why I Built It Keyword search breaks when you need semantic understanding. Pure" [X Link](https://x.com/initlayers/status/1992640000476021167) 2025-11-23T17:02Z [---] followers, [---] engagements "We've spent decades trying to catch liars by reading their faces. Turns out the most reliable signal isn't a micro expression or a twitch - it's heat. This paper shows that thermal patterns around the eyes can reveal deception far more consistently than human intuition. The authors tracked periorbital temperature variations with a mid-infrared camera and analysed [---] lie-truth responses. The authors captured [---] responses (249 lies [---] truths) and tracked [----] pixel periorbital patches across [--] frames (1.7 seconds) around each answer. The result is a compact 102-dimensional thermal signature" [X Link](https://x.com/initlayers/status/1992838423321928171) 2025-11-24T06:11Z [---] followers, [---] engagements "India is fighting over a piece of invisible real estate - the [--] GHz spectrum. And it might decide whether your next home Wi-Fi is lightning fast or painfully mid. Most people don't even know this fight is happening but here's why it actually matters. [--]. First: Why is [--] GHz such a big deal Think of internet signals as traffic on roads. Your current Wi-Fi runs on [---] GHz (slow crowded) and [--] GHz (faster but still crowded) [--] GHz is a brand new empty highway. No traffic no interference huge speeds. This is the band that unlocks Wi-Fi 6E and Wi-Fi [--] which can genually hit multi-gigabit speeds at" [X Link](https://x.com/initlayers/status/1993004090704773497) 2025-11-24T17:09Z [---] followers, [---] engagements "@karthikponna19 Jupyter notebook support as well" [X Link](https://x.com/initlayers/status/1993286313035153819) 2025-11-25T11:51Z [---] followers, [--] engagements "@OjasSharma276 @karthikponna19 Unit [--] - How to fine-tune an LLM from Scratch" [X Link](https://x.com/initlayers/status/1993604986530619747) 2025-11-26T08:57Z [---] followers, [--] engagements "29 November [----] Worked on project [--] today - easily my most productive day so far. Vibe-coded the entire backend and it boosted my speed like crazy. Almost finished the whole project which I didn't expect. Didn't study but I did code an LLM tokenizer from scratch in Python" [X Link](https://x.com/initlayers/status/1994864710924341721) 2025-11-29T20:23Z [---] followers, [---] engagements "Project 3: Ran into issues with docker-compose last night. It's supposed to start [--] services but only one comes online. Not sure if it's an endpoint issue or something inside Docker but it's taking a lot of time to debug" [X Link](https://x.com/initlayers/status/1995054908090020130) 2025-11-30T08:58Z [---] followers, [---] engagements "@Rushu_Tushu Yeah. It's done. Now I am working on hosting the repo on render" [X Link](https://x.com/initlayers/status/1995097626832974075) 2025-11-30T11:48Z [---] followers, [--] engagements "1 December [----] LLMs: Attention Types of Attention Mechanisms Problems with Longer Sequences Working of RNNs. Project 3: Still working on deploying the back-end on render. There is some issue with the environment variables I presume. Trying to debug" [X Link](https://x.com/initlayers/status/1995588320479641636) 2025-12-01T20:18Z [---] followers, [---] engagements "Tried the new X Android update. The animations are great and the UI feels fresh but it was so different that I switched back within [--] minutes. It completely changed how X feels and I'm not sure I can get used to it" [X Link](https://x.com/initlayers/status/1997909901625233873) 2025-12-08T06:03Z [---] followers, [----] engagements "@original_ngv @Starlink @elonmusk Did you do a speed test" [X Link](https://x.com/initlayers/status/1997949305605550287) 2025-12-08T08:40Z [---] followers, [--] engagements "Most people think attention was introduced by the [----] "Attention Is All You Need" - but that's not true. The real origin is a [----] paper by Bahdanau Cho and Bengio: Neural Machine Translation by Jointly Learning to Align and Translate. That paper introduced what we now call "attention": instead of compressing an input sentence into a single vector the decoder dynamically attends to different encoder hidden states per output token. That solved the fixed length bottleneck for long sentences. In [----] the Transformer architecture replaced recurrence entirely - stacking self attention (and cross" [X Link](https://x.com/initlayers/status/1998779029718184317) 2025-12-10T15:37Z [---] followers, [----] engagements "Practical 1/3 - Pattern Recognition & Machine Learning I reached late and wasn't allowed to sit for the exam at first. After requesting the professor multiple times she allowed me to take the test with other late comers. I couldn't cheat properly. During the implementation session I just copy pasted the code from ChatGPT. VIVA was simple af. The external asked: "Where do we use Python" "Difference between supervised and unsupervised learning." "Train-test split ratios." "What is reinforcement learning" "Real world applications of pattern recognition." It was an okayish day. VIVA cheered me up" [X Link](https://x.com/initlayers/status/1999564324361453715) 2025-12-12T19:37Z [---] followers, [---] engagements "Practical 2/3 - Intelligent Data Analysis Reached on time today. The test started late. I couldn't cheat because I'm not used to it. I didn't write the code. I implemented the same KMeans algorithm using ChatGPT but the evaluator didn't give me any marks because my answer sheet is empty. I mean what's the point of implementing then VIVA kept postponing because my professor is flirting with the external. The evaluator changed thrice and the final one asked me about: "What is dimensionality reduction" "Support vector machines are used for" "What is regression" I fumbled a bit at the SVM" [X Link](https://x.com/initlayers/status/1999847527915647446) 2025-12-13T14:23Z [---] followers, [---] engagements "Presented my group's project today (even though I barely worked on it). The project itself was entirely AI-generated. My teammates asked me to generate the report which I did. We didn't even manage to make a proper PPT due to time constraints. A good chunk of our time went into me explaining the project to my own teammates before explaining it to the evaluator. The external didn't help either - most of the questions were irrelevant and confusing. Easily the worst VIVA experience I've had. One thing became clear today: I don't work very well in teams yet. That's something I need to improve" [X Link](https://x.com/initlayers/status/2001288716015595866) 2025-12-17T13:49Z [---] followers, [---] engagements "AI/ML can feel overwhelming and there's a good reason for that. New architectures papers tools and techniques show up almost every day. If you pause for too long the gap becomes noticeable very quickly. Not because you're bad at it but because the field itself is moving fast. This isn't hype. AI/ML is one of the few domains where research industry and tooling evolve simultaneously. Papers turn into production systems in months. What was state-of-the-art last year can become baseline knowledge today. The mistake is thinking you need to learn everything. You don't. The people who survive" [X Link](https://x.com/initlayers/status/2003094444749533517) 2025-12-22T13:25Z [---] followers, [--] engagements "@yugb_twt LinkedIn and Cold Mailing ☠" [X Link](https://x.com/initlayers/status/2003357274727952680) 2025-12-23T06:49Z [---] followers, [--] engagements "AI/ML can feel overwhelming and there's a good reason for that. New architectures papers tools and techniques show up almost every day. If you pause for too long the gap becomes noticeable very quickly. Not because you're bad at it but because the field itself is moving fast. This isn't hype. AI/ML is one of the few domains where research industry and tooling evolve simultaneously. Papers turn into production systems in months. What was state-of-the-art last year can become baseline knowledge today. The mistake is thinking you need to learn everything. You don't. The people who survive" [X Link](https://x.com/initlayers/status/2003443950225342494) 2025-12-23T12:33Z [---] followers, [---] engagements "Took [--] AI interviews on Internshala today. Honestly it felt a bit soulless. Maybe I'm just more used to human evaluators. The first two interviews were mostly communication based. The third was Data Science (didn't go great) but the fourth was Machine Learning and that one went surprisingly well. The questions were basically overall. I'm still unsure about the internshala though. It feels slow and exhausting and I'm not convinced people actually land internships through it. Has anyone here got an internship/job from Internshala I'd like to know your experience" [X Link](https://x.com/initlayers/status/2004205077515636970) 2025-12-25T14:58Z [---] followers, [----] engagements "@autom8nerd Reddit is 🐐" [X Link](https://x.com/initlayers/status/2004234184760684974) 2025-12-25T16:54Z [---] followers, [--] engagements "@Rushu_Tushu Thanks. I will pivot to other platforms" [X Link](https://x.com/initlayers/status/2004438815830859921) 2025-12-26T06:27Z [---] followers, [--] engagements "@tm23twt WTF is this. Why am I seeing Machine Learning and Thermodynamics on the same paper" [X Link](https://x.com/initlayers/status/2004518515760668826) 2025-12-26T11:43Z [---] followers, [--] engagements "Looking for an ML Engineering Internship. I've built end-to-end ML pipelines worked on RAG systems and co-authored a research paper. I also have research internship experience at NIT Rourkela. I'm actively applying but I'd really appreciate referrals leads or even advice. If you know someone hiring ML Engineers or if your company is open to interns please let me know. Resume attached. Any help means a lot" [X Link](https://x.com/initlayers/status/2004539968388248045) 2025-12-26T13:09Z [----] followers, 81K engagements "Reading Karpathy says he feels behind is strangely comforting. If he feels FOMO one of the best minds we have it shows how insanely fast this field is moving. AI is rewriting the role of a programmer in real-time. New abstractions new layers new ways of thinking. You aren't supposed to have it all figured out none of us do. The only real skill now is staying in motion experimenting and learning faster than yesterday. If the legends are adapting we should too. I've never felt this much behind as a programmer. The profession is being dramatically refactored as the bits contributed by the" [X Link](https://x.com/initlayers/status/2004610143254970624) 2025-12-26T17:47Z [----] followers, [----] engagements "@munkeeDLuffy Bruh. They'll kill me when you use such terms I am hardly surviving 🥲" [X Link](https://x.com/initlayers/status/2004908639249989661) 2025-12-27T13:34Z [---] followers, [--] engagements "@OjasSharma276 Broski please get a Mac. You'll never regret listening to me" [X Link](https://x.com/initlayers/status/2005263193715909095) 2025-12-28T13:02Z [---] followers, [----] engagements "We're entering a strange phase in human-machine cognition. Large language models make high-level reasoning instantly accessible reducing the need to hold information in working memory or recall facts manually. Research on cognitive offloading already shows that when tools perform thinking steps for us our brain naturally delegate tasks to them. The pattern is visible with calculators maps and now reasoning itself. But something interesting is happening in parallel. When used deliberately LLMs also amplify human reasoning by generating alternate perspectives exposing blind spots and enabling" [X Link](https://x.com/initlayers/status/2005264095323558318) 2025-12-28T13:06Z [----] followers, [---] engagements "@OjasSharma276 I use M1 the basic model. It's smooth like butter. It feels better than [--] GB RAM Victus Asus Acer laptops" [X Link](https://x.com/initlayers/status/2005313068675539380) 2025-12-28T16:21Z [----] followers, [---] engagements "@RanjanRavi00 @vaemaski @OjasSharma276 I never happened to use a MacBook that heats more than Windows" [X Link](https://x.com/initlayers/status/2005553900108251210) 2025-12-29T08:18Z [---] followers, [--] engagements "@autom8nerd @StillTypeWritin Big brother is not responding I already contacted 😔" [X Link](https://x.com/initlayers/status/2005616151506849987) 2025-12-29T12:25Z [---] followers, [--] engagements "@devsterxyz Cracked dev is still an uncertainty become a good communicator instead" [X Link](https://x.com/initlayers/status/2005677244132573231) 2025-12-29T16:28Z [---] followers, [--] engagements "We finally hit [----] 🎊 I created this account on June [--] [----]. What started as random posts slowly became a habit then a discipline and now a part of my routine. There were days when I wanted to quit no reach no growth no validation. I'd watch others go viral while I posted quietly into the void. Many times I wondered if it was even worth it. But I kept showing up. Even when it felt slow. Even when it felt pointless. Even when consistency felt like a burden. I used to tell @OjasSharma276 that I'd hit 1K by the year-end. At the time I didn't fully believe myself but I kept going anyway. And" [X Link](https://x.com/initlayers/status/2006389323411963968) 2025-12-31T15:37Z [----] followers, [---] engagements "Goals for [----] Become financially independent Focus on physical fitness Never worry about unemployment Improve communication skills Reach 10K followers on X Start monetization on X Win atleast [--] hackathons Deep dive into LLMs and MLOps Improve programming skills Act on good habits immediately Stop aimless scrolling Reduce mobile usage https://twitter.com/i/web/status/2006714420278272055 https://twitter.com/i/web/status/2006714420278272055" [X Link](https://x.com/initlayers/status/2006714420278272055) 2026-01-01T13:09Z [----] followers, [---] engagements "Goals for [----] Become financially independent Focus on physical fitness Never worry about unemployment Improve communication skills Reach 10K followers on X Start monetization on X Win atleast [--] hackathons Deep dive into LLMs and MLOps Improve programming skills Act on good habits immediately Stop aimless scrolling Reduce mobile usage https://twitter.com/i/web/status/2006730256376533338 https://twitter.com/i/web/status/2006730256376533338" [X Link](https://x.com/initlayers/status/2006730256376533338) 2026-01-01T14:12Z [----] followers, [---] engagements "@Messing_up20023 Prompt engineering is pretty essential these days. Happy that you are covering it 🫂" [X Link](https://x.com/initlayers/status/2007311803458765292) 2026-01-03T04:43Z [----] followers, [--] engagements "I get a lot of DMs asking for good AI/ML/DSA resources. But the honest truth I don't know if the ones I followed are the best. So let's settle this once and for all let's crowdsource the best resources. If you're learning or working in any AI/ML domain drop what actually helped you. You can format it like this: ML/DL - Hands-on Machine Learning with Scikit-Learn and PyTorch by Aurlien Gron (Book) LLMs - Building LLMs from scratch by Vizuara (YT Playlist) MLOps - . CV/NLP/RL - . Books are preferred but high-quality playlists or courses are welcome too. Feel free to share full roadmaps - ML" [X Link](https://x.com/initlayers/status/2007734681140396288) 2026-01-04T08:43Z [----] followers, [----] engagements "@Pushkar11420566 Precise and to the point 🤝🏻" [X Link](https://x.com/initlayers/status/2007741115441750462) 2026-01-04T09:09Z [----] followers, [--] engagements "PATTERN RECOGNITION and MACHINE LEARNING - B. Tech. (Hon's) (Fifth Semester) Examination UNIT 1: [--]. (a) Explain the role of covariance in multivariate Gaussian distributions and justify how it influences the shape and orientation of the PDF. [--]. (b) Derive the expectation and variance of a linear combination of random variables. Analyse how these results support the Law of Large Numbers [--]. (c) Suppose a random vector x = (x1 x2) follows a multivariate Gaussian. Write down its joint PDF and interpret its parameters. [--]. (d) Formulate and justify a decision rule that minimizes expected loss in" [X Link](https://x.com/initlayers/status/2008476185207206226) 2026-01-06T09:50Z [----] followers, [----] engagements "@CanadaPRaspiran This is the reference book" [X Link](https://x.com/initlayers/status/2008540639773028581) 2026-01-06T14:06Z [----] followers, [--] engagements "INTELLIGENT DATA ANALYSIS - B. Tech. (Hon's) (Fifth Semester) Examination UNIT 1: [--]. (a) Explain the curse of dimensionality. [--]. (b) Differentiate between mean median and mode with one example. [--]. (c) What are the various methods of measuring data dispersion Illustrate with an example dataset. [--]. (d) Explaine the role of data transformation and data discretization in the pre-processing phase. Provide suitable examples. UNIT 2: [--]. (a) What is the principle of cross-validation Why is it used [--]. (b) Compare decision tree rule-based classifier and nearest-neighbor classifier based on accuracy" [X Link](https://x.com/initlayers/status/2009234121458094582) 2026-01-08T12:01Z [----] followers, [----] engagements "@Pseudo_Sid26 Bro is a formula [--] racer 🚗" [X Link](https://x.com/initlayers/status/2009956483518374107) 2026-01-10T11:52Z [----] followers, [--] engagements "@Pseudo_Sid26 Amazing article Sid. You are doing great 🔥" [X Link](https://x.com/initlayers/status/2010292634695467165) 2026-01-11T10:08Z [----] followers, [---] engagements "AI feels less like a tool and more like a craft. Easy to use but hard to master. You can pick up AI products quickly and pivot into adjacent roles but building a stable career here is different. It demands patience depth and years of compounding effort. This isn't a field where shortcuts work. If you're not ready to put in at least 4-5 years expect very little in return. I've never seen a domain evolve this fast. What you learn today can feel outdated tomorrow. The only way to survive is to keep learning and keep building. https://twitter.com/i/web/status/2010388722051924340" [X Link](https://x.com/initlayers/status/2010388722051924340) 2026-01-11T16:29Z [----] followers, [---] engagements "NATURAL LANGUAGE PROCESSING - B. Tech. (Hon's) (Fifth Semester) Examination UNIT 1: [--]. (a) Define Natural Language Processing. What are its main goal [--]. (b) Describe the key challenges of Natural Language Processing. [--]. (c) Explain the importance of NLP in sentiment analysis. [--]. (d) Write the history of Natural Language Processing. UNIT 2: [--]. (a) What is the Markov assumption [--]. (b) Explain in detail how regular expressions and automata are used in NLP. [--]. (c) Explain the three branches of phonetics: articulatory acoustic and auditory. [--]. (d) Describe different word-formation processes in" [X Link](https://x.com/initlayers/status/2010644809636295147) 2026-01-12T09:27Z [----] followers, [---] engagements "@Dev__Vishwajeet This semester will be intense though a lot of work and uncertainty" [X Link](https://x.com/initlayers/status/2011856881661464709) 2026-01-15T17:43Z [----] followers, [--] engagements "1200 now. [----] - [----] took [--] days 🙂 We are [----] now. [----] - [----] took [--] hours 🙃 https://t.co/os5CRDmcLU We are [----] now. [----] - [----] took [--] hours 🙃 https://t.co/os5CRDmcLU" [X Link](https://x.com/initlayers/status/2012711462595444830) 2026-01-18T02:19Z [----] followers, [----] engagements "I'm currently piled up with multiple research opportunities from multiple IITs. I'll be picking one or two soon and share an update. The last couple of days were a bit rough so I went quiet. Back now" [X Link](https://x.com/initlayers/status/2014022640512438673) 2026-01-21T17:09Z [----] followers, [----] engagements "Starting February [--] I'm committing the next [--] days to sharpening my fundamentals the right way. Python GitHub Linux Shell Scripting Consistent daily work until these tools feel natural and no more intimidating. If you're also working on your fundamentals this is your sign to go all in. https://twitter.com/i/web/status/2015127013171032190 https://twitter.com/i/web/status/2015127013171032190" [X Link](https://x.com/initlayers/status/2015127013171032190) 2026-01-24T18:18Z [----] followers, [----] engagements "AI makes coding faster and prototyping easier but it doesn't replace learning to code. Models can generate decent code sure but when something breaks in production you can't rely on someone who doesn't understand what's happening. Real skill is knowing how to debug fix and take responsibility when things go wrong" [X Link](https://x.com/initlayers/status/2016942968209600763) 2026-01-29T18:34Z [----] followers, [---] engagements "@_Creation22 Some OCR stuff AI Validation and Blockchain" [X Link](https://x.com/initlayers/status/2019822140066050422) 2026-02-06T17:14Z [----] followers, [--] engagements "If you're hiring ML/AI interns look at this 👇🏻 @Yash_xml has experience with LLMs agents deep learning and internships at IBM & EY. Resume attached. Please RT to help him reach the right people" [X Link](https://x.com/initlayers/status/2014694123815764080) 2026-01-23T13:38Z [----] followers, 19.4K engagements "The lab gave me a PC with 32GB RAM and an i7-10700 basically a beast for research work. It came with a pirated Windows [--] Education copy so I waited patiently thinking they'd eventually provide a proper key but I'm done waiting. Installing Ubuntu [-----] now and setting it up the way I want. This is going to be my workstation for the next 5-6 months so it better be clean stable and ready for some serious work" [X Link](https://x.com/initlayers/status/2020802377058812179) 2026-02-09T10:10Z [----] followers, 43.9K engagements "This striked my mind today. Skip connections and residual blocks may appear to be identical but that's not the case. A skip connection is a technique while a residual block is a structural component that utilizes that technique. Skip connections are direct pathways that skips one or more layers allowing information to bypass non-linearities. Residual block is a specific architectural unit where a skip connection is used to add the input x to the output of set of layers F(x) resulting in F(x) + x. This is commonly seen in ResNet architectures. To put it simply all residual block use skip" [X Link](https://x.com/initlayers/status/2021211607758668258) 2026-02-10T13:16Z [----] followers, [----] engagements "Unpaid internships are almost never worth it matter how desperate you feel. If a company can't pay you they usually won't mentor you either. You'll end up doing random work burning time and calling it experience. Upskill instead. Be patient. Build projects. Network. Apply better. The only exception I'd consider is a genuine research internship where you're actually learning and working closely with a good mentor. But unpaid industry internships Don't normalize that" [X Link](https://x.com/initlayers/status/2021585897259716813) 2026-02-11T14:03Z [----] followers, [----] engagements "RAG (Retrieval Augmented Generation) - what's the actual point People talk about RAG a lot but the idea is simple: LLMs are great at language not memory. They don't store facts. They don't know your PDFs. They just predict the next token. RAG fixes that. Here's how it actually works: [--]. Retrieval You take your documents chunk them convert each chunk into embeddings and store them in a vector database. When a query comes in you embed the query and pull the top-k most relevant chunks using vector similarity. [--]. Augmentation You attach those retrieved chunks to the user query. This gives the" [X Link](https://x.com/initlayers/status/1991859301842100625) 2025-11-21T13:20Z [----] followers, [---] engagements "Reached the lab. Today's reading list: DnCNN DRUNet Restormer and SCUNet. All of them are strong baselines for image restoration and I want to understand what makes each one work. Let's see how much I can cover today" [X Link](https://x.com/initlayers/status/2021510135596925011) 2026-02-11T09:02Z [----] followers, [----] engagements "@tusharuii Depends on your interest. If you are okay with CNNs start implementing U-Net" [X Link](https://x.com/initlayers/status/2022152402519068735) 2026-02-13T03:34Z [----] followers, [---] engagements "@VarunrRao I started in my 3rd year. Feeling FOMO 😮" [X Link](https://x.com/initlayers/status/2022345349793518010) 2026-02-13T16:21Z [----] followers, [--] engagements "Working on Sundays so I can make weekdays easier" [X Link](https://x.com/initlayers/status/2022887821716185422) 2026-02-15T04:16Z [----] followers, [---] engagements "If you already have a basic understanding of deep learning and you're in your 2nd or 3rd year start reading papers and implementing them. Don't wait until you feel ready. That moment never comes. Use AI tools to break down the paper understand the intuition and replicate the code step by step. You'll learn more from one implementation than from weeks of passive courses. And if PyTorch feels like a barrier just watch a good one shot PyTorch video and start building immediately. You'll thank youself later. https://twitter.com/i/web/status/2022146198132732325" [X Link](https://x.com/initlayers/status/2022146198132732325) 2026-02-13T03:09Z [----] followers, [----] engagements "Valentine's Day started with a [--] AM meeting with my supervisor. [--] minutes later I got humbled at least a hundred times. I couldn't clearly explain convolutions there was miscommunication and I realized I don't understand some of the architectures deeply enough for the literature review I'm doing. The intuition wasn't there and it showed. Now I'm back in my room with a lot to revisit and half my valentine's Day is already gone. Research really doesn't care what day it is. https://twitter.com/i/web/status/2022571053273493549 https://twitter.com/i/web/status/2022571053273493549" [X Link](https://x.com/initlayers/status/2022571053273493549) 2026-02-14T07:18Z [----] followers, [---] engagements "10 hours into this project today. I've been using ChatGPT-5.2 and Perplexity Deep Research so far but at some point the pressure hit hard and genuinely felt like I was stuck and would be trapped in this project forever. Then I installed GPT-5.3 Codex just to see what the hype was about and honestly this is scary. I've never designed and built things this fast in my life I was cynical about context window and rate limits but for the kind of changes and iterations I needed it was more than sufficient. I still have around [--] more hours of work left but I'm pretty sure the day is going to end" [X Link](https://x.com/initlayers/status/2023037057224904741) 2026-02-15T14:09Z [----] followers, [---] engagements "I am working on an open-source RAG system focused on efficient hybrid retrieval and evaluation. It's been [--] days since I started the project. With mid-terms coming up next week I've decided to finish it tonight no matter what. I will be working for the next [--] hours taking little breaks. Let's see how far we can go" [X Link](https://x.com/anyuser/status/1984908480986116541) 2025-11-02T09:00Z [----] followers, 20K engagements "It's my 🎈 day - just turned [--]. My first birthday on X. I've met many awesome people here since I joined in June. Hopefully by my next birthday you'll look at me as someone who adds real value here" [X Link](https://x.com/anyuser/status/1986511458457518286) 2025-11-06T19:10Z [----] followers, [----] engagements "November [----] Goals: Complete 2nd MLOps project Update resume & start applying Land an ML Engineer internship Finish MLOps lectures (3/25 done) Grow to 1K followers on X (470 now) & [---] verified (160 now) Complete half of the LLM lectures (currently at 9/43 done) Win a hackathon Score 75%+ in mid-terms This month will be intense. Mid-terms in week [--] lectures piling up projects and hackathons waiting. Somehow I've got to balance it all and land that internship and connect with more people here. Also. it's my birthday month 😊" [X Link](https://x.com/anyuser/status/1984851364250599783) 2025-11-02T05:13Z [----] followers, [----] engagements "I finally completed my first research paper 💀 I'll submit somewhere early september. It is [--] pages long" [X Link](https://x.com/anyuser/status/1953873068457439430) 2025-08-08T17:36Z [----] followers, [----] engagements "What the actual fak. @cneuralnetwork -- this guy inspired me to start posting on X. Not many people know that my account name itself is inspired by his. So it took him [---] days for him to notice me. Not bad. Hope it's not a mistake 😭" [X Link](https://x.com/anyuser/status/1981427895114616938) 2025-10-23T18:29Z [----] followers, [----] engagements "October [----] Rewind: Studied CNNs Did two submissions for SIH (many sleepless nights) Participated in two more hackathons Started LLMs and MLOps Finished a RAG project Got gifted a premium account X Grew from [---] to [---] followers Honestly this month wasn't anything special. But I learnt a lot in the process. I am more serious now about building on X and sharpening my skills. Also met some amazing people here this month" [X Link](https://x.com/anyuser/status/1984605184753090645) 2025-11-01T12:55Z [----] followers, [----] engagements "600 🙂 This growth needs to be studied. It took me [--] hours from [---] - [---] [---] followers it is. Took [--] hours to go from [---]. Grateful for the amazing people on this platform. https://t.co/fU5Frc4rVw [---] followers it is. Took [--] hours to go from [---]. Grateful for the amazing people on this platform. https://t.co/fU5Frc4rVw" [X Link](https://x.com/anyuser/status/1985332354999849147) 2025-11-03T13:04Z [----] followers, [----] engagements "500 followers it is. Took [--] hours to go from [---]. Grateful for the amazing people on this platform. We are at [---] now. It took [--] days and [--] hours from [---] 🦧 https://t.co/QYZbHRsDR2 We are at [---] now. It took [--] days and [--] hours from [---] 🦧 https://t.co/QYZbHRsDR2" [X Link](https://x.com/anyuser/status/1984985389191635121) 2025-11-02T14:06Z [----] followers, [----] engagements "Thanks to @sharpeye_wnl and @TanayVasishtha for sponsoring my Premium. All the best to you guys as well 🫂 Got my blue tick within [--] minutes of payment which is super fast Feels like I'm on the right path after all. As part of refactoring my account I removed around [--] followers who were basically bots. Not sure if that's the best move algorithmically but I hope to build a genuine community that engages and responds. Since I have a month of these benefits I'll try to engage more post consistently and learn from you all. Reach is pretty rough right now but I'll focus on adding value. Usually" [X Link](https://x.com/anyuser/status/1981384959316389949) 2025-10-23T15:39Z [----] followers, [----] engagements "@yikoshi Holding hands and crying in her lap" [X Link](https://x.com/anyuser/status/1984662099956727939) 2025-11-01T16:41Z [----] followers, [----] engagements "8 Nov [----] IDA Mid Prep: Clustering Analysis K-means K-mediods Agglomerative Divisive Distance Metrics for Clustering Algorithms DB Scan Evaluation of Clustering PRML Mid Prep: Perceptron Algorithm Support Vector Machine Hard and Soft Margin SVM" [X Link](https://x.com/anyuser/status/1987257439625122249) 2025-11-08T20:34Z [----] followers, [----] engagements "Guess who didn't sleep all night just for making a PPT. There you go it's me 🤥 I have my internal hackathon tomorrow (we'll technically today) wish me luck" [X Link](https://x.com/anyuser/status/1970995373658472860) 2025-09-24T23:34Z [----] followers, [----] engagements "Time to switch gears. I just wrapped up my RAG project - easily one of the toughest and most rewarding things I've built so far. Mid-terms are now around the corner so I probably won't be diving deep into LLMs or MLOps until they're done. I'll share the project repo after exams - once the documentation is solid. No compromises here. Exams will be over by the 15th. Until then I'll post small things I'm learning - tiny concepts insights and reflections. I know why most of you are here - it's for LLMs and MLOps. I'll do my best to stay consistent and keep adding value even during this break." [X Link](https://x.com/anyuser/status/1985720202299121967) 2025-11-04T14:46Z [----] followers, [---] engagements "9 November [----] PRML Mid Prep: Kernel Function Linear Fisher Polynomial Gaussian Radial Basis Function Cosine Two-class problem Hard constraint SVM Soft Margin SVM. Linear models for classification Discriminant functions fisher's discriminant Probabilistic discriminative and generative models" [X Link](https://x.com/anyuser/status/1987601725470245343) 2025-11-09T19:22Z [----] followers, [---] engagements "26 October [----] Studied the evolution of GPT from transformers to GPT-5. A brief understanding of what's changed Revisited the working of GPT-3 Studied different types of tokenizations schemes Byte Pair Encoding Finally completed revising OOPs: Static method encapsulation setter and getter methods Project: Pushed everything to git. Created a streamlit UI using AI. Everything worked well. I'll take [--] more days and launch it here. Not satisfied with the grind 😕" [X Link](https://x.com/anyuser/status/1982568154607067143) 2025-10-26T22:00Z [----] followers, [---] engagements "The most dangerous thing about AI isn't that it lies - it's that it lies confidently. You ask a model something simple. It replies instantly flawlessly persuasively and yet it's wrong. That's called a hallucination. Large Language Models don't know things. They don't store facts or check truth. They just predict the next word based on patterns they've seen before. So when the model doesn't know something it doesn't say "I'm not sure." It guesses. But it guesses beautifully. It's like that one kid in class who didn't study but still wrote a full essay with perfect grammar and total confidence." [X Link](https://x.com/anyuser/status/1986092585337974927) 2025-11-05T15:25Z [----] followers, [----] engagements "25 October [----] Revised Pre-training Fine-tuning Fine-tuning techniques building schematic simplified transformers architecture some more concepts Coded an LLM Tokenizer from scratch using regular expressions and python Project: Completed the skeleton of the project. All modules are working well. I debugged LLM client errors. Created a script for an integrated pipeline combining all modules it took a lot of time. The UI is pending (not very keen on). Further refinement and add-ons are to be made" [X Link](https://x.com/anyuser/status/1982196827639406611) 2025-10-25T21:25Z [----] followers, [----] engagements "It took roughly a month this time to gain the next [--] followers but it's worth it. The last [--] have been of high quality 😃 I think I can finally say this now that I have a small community here: [---] it is https://t.co/XGhbLFsoXE I think I can finally say this now that I have a small community here: [---] it is https://t.co/XGhbLFsoXE" [X Link](https://x.com/anyuser/status/1981458673147994389) 2025-10-23T20:32Z [----] followers, [----] engagements "I think I can finally say this now that I have a small community here: [---] it is Okay so this happened as well 😮 https://t.co/dyyx5IHjFv Okay so this happened as well 😮 https://t.co/dyyx5IHjFv" [X Link](https://x.com/anyuser/status/1972349353315733711) 2025-09-28T17:15Z [----] followers, [----] engagements "28 October [----] Studied the need for data-target pair creation. Looked at custom dataset and data loader generation The lecture was dense and I couldn't complete it. Didn't do any project work or MLOps as well for the same reason" [X Link](https://x.com/anyuser/status/1983288998069186621) 2025-10-28T21:45Z [----] followers, [---] engagements "One week of using a premium account. It has been good not the best. I'll keep analyzing what I can improve over the next [--] weeks. Thanks for all the engagement 😊" [X Link](https://x.com/anyuser/status/1983539577324552410) 2025-10-29T14:21Z [----] followers, [---] engagements "@abd_mukadam It's a career update for the company" [X Link](https://x.com/anyuser/status/1983062938325004691) 2025-10-28T06:47Z [----] followers, [---] engagements "We are at [---] now. It took [--] days and [--] hours from [---] 🦧 We're at [---] now. It took [--] hours this time - not for [--] but for [---]. It blew up so fast I couldn't even post anything at [---] 🙂 https://t.co/tkofzjzvv1 We're at [---] now. It took [--] hours this time - not for [--] but for [---]. It blew up so fast I couldn't even post anything at [---] 🙂 https://t.co/tkofzjzvv1" [X Link](https://x.com/anyuser/status/1983941700348473783) 2025-10-30T16:58Z [----] followers, [----] engagements "We're at [---] now. It took [--] hours this time - not for [--] but for [---]. It blew up so fast I couldn't even post anything at [---] 🙂 It took roughly a month this time to gain the next [--] followers but it's worth it. The last [--] have been of high quality 😃 https://t.co/mk1XyPZVD3 It took roughly a month this time to gain the next [--] followers but it's worth it. The last [--] have been of high quality 😃 https://t.co/mk1XyPZVD3" [X Link](https://x.com/anyuser/status/1982379375954907399) 2025-10-26T09:30Z [----] followers, [----] engagements "It took me [--] days to reach my first [--] followers and just [--] days to reach the next [--]. That's [--] days on this app so far. Greatful for the journey and all the learnings thank you 🙏🏻" [X Link](https://x.com/anyuser/status/1968193691853656145) 2025-09-17T06:01Z [----] followers, [---] engagements "1 November [----] The entire CI/CD pipeline was setup nicely. I have created a bunch of unit integration and performance tests. I had to spend a good amount of time writing and debugging these. This project is slowly moving towards MLOps. I want to demonstrate all my skills. Couldn't study LLMs today" [X Link](https://x.com/anyuser/status/1984715100062957688) 2025-11-01T20:12Z [----] followers, [---] engagements "29 October [----] Revised LLM Intro classes & objects constructor initializer method vs functions special methods encapsulation getter and setter methods static function. Revised tokenization algorithm types BPE algorithm input-target pair generation custom dataset and data loader creation from Pytorch components. Studied batch_size and num_workers in great detail. These will help me with model training" [X Link](https://x.com/anyuser/status/1983630925788540966) 2025-10-29T20:24Z [----] followers, [---] engagements "27 October [----] Revised stages of building LLMs Revised OOPs - second lap Project: Did not do much. I understood that there are several issues with multiple modules of my project. I tried to understand my code better and understand limitations. Today was a L. Least productive day ever" [X Link](https://x.com/anyuser/status/1982913597710971071) 2025-10-27T20:53Z [----] followers, [---] engagements "I talked to @Divyansh91565 last night. He is a humble guy and helped me getting my basics right in DSA and CP. Hoping to connect with many great people here who are doing amazing in their fields" [X Link](https://x.com/anyuser/status/1977952502311162007) 2025-10-14T04:19Z [----] followers, [---] engagements "Okay so this happened as well 😮 It now took me only [--] hours for the next [--] 😐 what is this growth Feels like I am speed running X https://t.co/1o3hHxUgEX It now took me only [--] hours for the next [--] 😐 what is this growth Feels like I am speed running X https://t.co/1o3hHxUgEX" [X Link](https://x.com/anyuser/status/1969242100458783072) 2025-09-20T03:27Z [----] followers, [---] engagements "1 October [----] I learnt Git and GitHub from scratch. We had a [--] hour meeting today to discuss about our SIH progress and upcoming hackathons. Today was a bit productive" [X Link](https://x.com/anyuser/status/1973442592126935420) 2025-10-01T17:39Z [----] followers, [---] engagements "23 September [----] DL: Filters and Convolutions in 1D image data MLOps: Introduction to MLOps need for MLOps Statistics: Describing datasets different plotting techniques MLOps: ML project life cycle Software development life cycle I will be consistent with my work 🤞🏻" [X Link](https://x.com/anyuser/status/1970576749298905440) 2025-09-23T19:51Z [----] followers, [---] engagements "31 October [----] I couldn't study today. Worked on my project again. [--] hours almost. I finally setup CI/CD for my project using GitHub Actions. Initial tests worked. So I prepared more comprehensive unit tests. I am waiting for all of this to be finished so that I can go to the last stages and refine parts of the project. https://x.com/initlayers/status/1984355416156094474/photo/1 https://x.com/initlayers/status/1984355416156094474/photo/1" [X Link](https://x.com/anyuser/status/1984355416156094474) 2025-10-31T20:22Z [----] followers, [---] engagements "24 September [----] I studied 2D filters in CNNs today. How the filters are applied to the depth of the images instead of individual matrices. I tried to find an isolated pixel using custom filters. It was a fun ride" [X Link](https://x.com/anyuser/status/1970977145485918625) 2025-09-24T22:22Z [----] followers, [---] engagements "14 October [----] Revised CNNs: Why CNNs over ANNs Filters Hands-on 1D & 2D data convolutions multiple filters Studied Filtering Layer Nomenclature Revised Git commands ML lifecycle I don't think it's a great day or something. But everyday counts I will be patient" [X Link](https://x.com/anyuser/status/1978177124860113130) 2025-10-14T19:12Z [----] followers, [---] engagements "30 October [----] Couldn't study LLMs or MLOps I worked on the project for [--] hours. Performed ablation studies trying to document everything. I am trying to set-up CI/CD with GitHub Actions. I dunno if it'll be completed tonight. Today was rusty though I couldn't study anything" [X Link](https://x.com/anyuser/status/1984008988011417957) 2025-10-30T21:26Z [----] followers, [---] engagements "I am stuck with these tests. Chances of finishing the project tonight are slim 🥺" [X Link](https://x.com/anyuser/status/1984321187720413415) 2025-10-31T18:06Z [----] followers, [---] engagements "In the Chinese Room thought experiment the person inside knows only English and has an instruction manual. People outside pass him Chinese question through a slot. The person follows the manual and produces perfect Chinese responses. From the outside it seems like he understands Chinese but in reality he doesn't. The intelligence isn't in the person alone but perhaps in the system (the person + the instruction manual) Now here's the question: if we were to simulate every neural network and pathway of a human brain would that count as true intelligence" [X Link](https://x.com/anyuser/status/1981548701773648313) 2025-10-24T02:30Z [----] followers, [---] engagements "24 October [----] Revised LLM basics stages of building LLMs OOPs - mostly around classes and objects Project: Finally resumed. I built core modules to test project feasibility. I then downloaded Ollama and tested TinyLlama locally. Obviously it's not very great but with my GPU and RAM I am satisfied with anything I can possibly get. The next couple of days will be intense. I will try to code stuff from scratch" [X Link](https://x.com/anyuser/status/1981835386201727448) 2025-10-24T21:29Z [----] followers, [---] engagements "21 October [----] Studied stages of building LLMs again but in more detail Finally started the project-1 implementation just set-up the environment and dependencies after careful consideration I have decided to document my choices and learnings throughout the project" [X Link](https://x.com/anyuser/status/1980744386238640324) 2025-10-21T21:13Z [----] followers, [---] engagements "15 October [----] Selected a PS built a prototype and recorded a demo video for SIH (took around [--] hours) Revised filtering layer nomenclature" [X Link](https://x.com/anyuser/status/1978550976165376447) 2025-10-15T19:58Z [----] followers, [---] engagements "19 October [----] Studied pre-training fine-tuning types of fine-tuning and stages of building LLMs Tried to understand Transformers Architecture tokenization self attention vector embeddings BERT vs GPT Project: confirmed the stack we will use free tools/models" [X Link](https://x.com/anyuser/status/1980000258768904534) 2025-10-19T19:57Z [----] followers, [---] engagements "It now took me only [--] hours for the next [--] 😐 what is this growth Feels like I am speed running X It took me [--] days to reach my first [--] followers and just [--] days to reach the next [--]. That's [--] days on this app so far. Greatful for the journey and all the learnings thank you 🙏🏻 https://t.co/5eSDF9w7VA It took me [--] days to reach my first [--] followers and just [--] days to reach the next [--]. That's [--] days on this app so far. Greatful for the journey and all the learnings thank you 🙏🏻 https://t.co/5eSDF9w7VA" [X Link](https://x.com/anyuser/status/1968515509735014417) 2025-09-18T03:20Z [----] followers, [---] engagements "Small account happiness 🙂" [X Link](https://x.com/anyuser/status/1968510076592816418) 2025-09-18T02:59Z [----] followers, [---] engagements "I'm still working on this project. It's been around [--] hours already and things aren't looking great. The code has finally passed all [--] unit integration & performance tests but the SQLite backend in Chroma is set to read-only which is causing issues during CI. I'll fix this it should be an easy one. I'll do everything I can to finish the project tonight" [X Link](https://x.com/anyuser/status/1985050124243108100) 2025-11-02T18:23Z [----] followers, [---] engagements "It took me [--] hours but I am finally done with the back-end it's working great. I'll start the front-end as well I have to complete this tonight at any cost" [X Link](https://x.com/anyuser/status/1972754815605510307) 2025-09-29T20:06Z [----] followers, [---] engagements "Tomorrow is the deadline I have to build an MVP ASAP else NGMI. Our presentation will include pictures of the UI so I am the bottleneck now this night will be legendary 💀" [X Link](https://x.com/anyuser/status/1972651600088088918) 2025-09-29T13:16Z [----] followers, [---] engagements "I think I'll break the record for spending the most midnights on SIH. Hope it's all worth it 😐" [X Link](https://x.com/anyuser/status/1975337344640426222) 2025-10-06T23:08Z [----] followers, [---] engagements "20 Sept [----] I designed a basic neural network for classification using TensorFlow it felt a lot different to Pytorch. I studied basics of vector algebra and span of two vectors spent most of the time shortlisting problem statements for SIH" [X Link](https://x.com/anyuser/status/1969463220155847082) 2025-09-20T18:06Z [----] followers, [---] engagements "Guys I found something cool: The automatic recognition of machine or hand-printed characters is what we refer to by Optical Character Recognition (OCR) The recognition of human printed or cursive handwriting as the user is writing it's called "On-line Handwriting Recognition"" [X Link](https://x.com/anyuser/status/1963204623499714792) 2025-09-03T11:37Z [----] followers, [---] engagements "16 October [----] Studied max pooling layer and CNN architecture. Revised filtering layer nomenclature and working" [X Link](https://x.com/anyuser/status/1978897351507869820) 2025-10-16T18:54Z [----] followers, [---] engagements "Good morning 🌞 The sun's aura is too high that I couldn't see him eye to eye even from [---] million kilometers" [X Link](https://x.com/anyuser/status/1969212326852039120) 2025-09-20T01:29Z [----] followers, [---] engagements "Our second SIH submission (made on 15th October) wasn't successful. We worked hard for it but couldn't quite pull it off. Now we're waiting for the batch [--] results. Let's see if our first submission can work some magic" [X Link](https://x.com/anyuser/status/1987381885136478705) 2025-11-09T04:48Z [----] followers, [---] engagements "Project Update: I finally made it. I completed the CI/CD pipeline and managed to not break anything. The project is now officially finished. I'll wrap up the remaining documentation and front-end work after my mid-terms. The last 2-3 days have been a hell of a ride. I learnt a lot - built a comprehensive test suite integrated LLM inference around 2K+ lines of code and spent countless hours debugging. It's been an amazing journey" [X Link](https://x.com/anyuser/status/1985486432082018545) 2025-11-03T23:17Z [----] followers, [---] engagements "Instagram and TikTok are in a league of their own when it comes to algorithms. YouTube Spotify and X follow behind but none match how fast Instagram learns your mood. It's scary how good it's become at keeping you watching. Even though it's affecting attention spans and careers it remains a masterpiece both algorithmically and psychologically" [X Link](https://x.com/anyuser/status/1981734594111303741) 2025-10-24T14:48Z [----] followers, [---] engagements "Brainstorming hackathon ideas for two challenges: one from IITB and the other from IITR. Both are kind of focussing on environmental issues. Let's see what would be finalized" [X Link](https://x.com/anyuser/status/1972301476564537578) 2025-09-28T14:04Z [----] followers, [---] engagements Limited data mode. Full metrics available with subscription: lunarcrush.com/pricing
@initlayers initlayers
initlayers posts on X about the next, llm, the project, ai the most. They currently have [-----] followers and [---] posts still getting attention that total [---] engagements in the last [--] hours.
Engagements: [---] #
- [--] Week [------] +4,275%
- [--] Month [------] -21%
- [--] Months [-------] +16,986%
Mentions: [--] #
- [--] Months [--] +41%
Followers: [-----] #
- [--] Week [-----] +9%
- [--] Month [-----] +20%
- [--] Months [-----] +3,716%
CreatorRank: [---------] #
Social Influence
Social category influence technology brands finance stocks social networks countries celebrities products
Social topic influence the next, llm, the project, ai, oops, finish, woke, if you, update, night
Top assets mentioned Spotify Technology (SPOT)
Top Social Posts
Top posts by engagements in the last [--] hours
"@perceptron_cnn What is the theme"
X Link 2025-06-29T17:38Z [--] followers, [--] engagements
"Okay. I did not get shortlisted in Adobe India Hackathon. And to be very honest I am not even surprised am I lost chat 🥲"
X Link 2025-07-15T20:37Z [--] followers, [--] engagements
"@Pseudo_Sid26 I think total access to the codebase should be given only after 60% of the payment is done 🙂 there are a lot of scamsters in freelancing. But it depends"
X Link 2025-07-16T13:49Z [--] followers, [--] engagements
"Day 1: We'll kickstart the project. The prompt is [---] words (5345 characters). This should be enough to get the basic project done. I'll just keep refining this. I am trying perplexity let's see how there routing system works I am excited"
X Link 2025-07-19T08:37Z [--] followers, [--] engagements
"Day [--] Progress (20/07/2025): Designed a systematic protocol for hyperparameter tuning Took lr=1e-3 and epochs=50 as baseline for all the models. Trained [--] times to produce 87.5% validation accuracy for MFCC model 1/2"
X Link 2025-07-20T21:22Z [--] followers, [--] engagements
"Tweaked parameters [--] times to achieve 98.6% accuracy on mel Ran the model [--] times and did early stopping at epoch [---] to reach 97.2% on fusion model Perfomance ranking: Mel Fusion MFCC Finalized the results and prepared a dedicated section for the paper. 2/2"
X Link 2025-07-20T21:22Z [--] followers, [--] engagements
"@sageadvik Used ChatGPT for [--] months. It sucks at many things. I am using this Perplexity pro which I got for free recently. I am liking the fresh responses. It doesn't care about what I feel it just does the work. ChatGPT focuses more on what I am saying rather than actually working"
X Link 2025-07-21T20:33Z [--] followers, [--] engagements
"Day [--] Progress (21/07/2025): Freezed my hyper parameters for each model. Employed early-stopping to improve efficiency and save compute Enabled GPU and installed new packages causing epoch shifts. Tried to roll back to the previous version of the codebase but failed"
X Link 2025-07-21T21:11Z [--] followers, [--] engagements
"Apparently [--] epoch shift is acceptable This occurs due to inherent stochasticity in DL training Finalized the results for the paper Employed 5-fold CV for best model of each MFCC Mel & Fusion CV takes too much time Still awaiting thermal image data "
X Link 2025-07-21T21:12Z [--] followers, [--] engagements
"@adxtya_jha @perplexity_ai The interface is shit. Interpretability is non existent. Finally an AI that doesn't care about me abusing. I don't know if I should be happy for that. Atleast I am getting free deep research. I would still prefer ChatGPT if I was allowed limited deep research on both"
X Link 2025-07-22T09:09Z [--] followers, [---] engagements
"@adxtya_jha @perplexity_ai The responses are fresh and it follows the instructions mostly. ChatGPT research is good when it understands the prompt and you ask for the first time. One more try and it changes it's faith. I started to believe GPT hates being asked twice"
X Link 2025-07-22T09:14Z [--] followers, [--] engagements
"Day [--] Progress (22/07/2025): Performed [--] fold cross validation on the best model of mel mfcc and fusion Mel (98.6%) Fusion (97.2%) MFCC (875%) in peak single-run accuracy Fusion exhibits the highest average cross-validated accuracy (96.8% 2.8%)"
X Link 2025-07-22T17:14Z [--] followers, [--] engagements
"@purusa0x6c None. We are busy vibe coding stuff that has been made [---] times already but with a minor tweak"
X Link 2025-07-31T17:09Z [--] followers, [--] engagements
"I wanted to perform LOSO Cross Validation to get better insights for paper. But the problem is I have [--] subjects and I didn't want to manually run for each model so I made one single code. Now this will do [--] folds validation on [--] models 💀 I am leaving it to run overnight "
X Link 2025-08-02T20:44Z [--] followers, [---] engagements
"@Maaztwts Spotify's music recommendation algorithm is still top notch"
X Link 2025-08-03T05:28Z [--] followers, [--] engagements
"@Maaztwts They made their services exclusive to subscribers. It was still good for normal users some years ago. Spotify was the first subscription I ever purchased. I liked it so much but now it's different"
X Link 2025-08-03T05:58Z [--] followers, [--] engagements
"After [--] hours of running the script overnight. I got the results for all the [--] models. And the results are shocking 💀 mel and fusion wre constantly beating eachother in different analyses"
X Link 2025-08-03T07:01Z [--] followers, [--] engagements
"I had a nightmare last night. It stopped when my alarm rang at 5:30 AM. The story setting was IIT Guwahati and I am responsible for a civil war. I had to run away from two factions fighting eachother with grenades and all sort of stuff. I saw the majestic Brahmaputra turn red"
X Link 2025-08-04T01:40Z [--] followers, [--] engagements
"@iamAdityaAnjana We got Pattern Recognition and Machine Learning Natural Language Processing Intelligent Data Analysis Computational Complexity Cryptography and Network Security. Labs for PRML NLP and IDA"
X Link 2025-08-04T07:48Z [--] followers, [--] engagements
"Subjects for this semester: Natural Language Processing Pattern Recognition & Machine Learning Intelligent Data Analysis Computational Complexity Cryptography and Network Security Minor Project [--] based on Industrial Training Labs: NLP PRML IDA"
X Link 2025-08-12T13:58Z [--] followers, [---] engagements
"@freshlimesofa You are doing good. No fixes needed"
X Link 2025-08-13T16:57Z [--] followers, [--] engagements
"I am thinking of learning Web Design. I started with a [--] hour crash course from flux academy. However most of it is bouncing of my head. I don't know if I am learning anything. Can someone please help me with this"
X Link 2025-08-14T15:57Z [--] followers, [--] engagements
"The mid-term is finally done. From tomorrow it's grind time kicking off my productivity arc and pushing my limits. I'll try to update here everyday consistently"
X Link 2025-09-14T17:02Z [----] followers, [---] engagements
"@sharpeye_wnl I just don't trust them in the long term. Even a Samsung product will work but not Xiaomi"
X Link 2025-09-20T05:19Z [---] followers, [--] engagements
"@daddy__broccoli The urge to leave everything aside and learn assembly is getting stronger with each passing day"
X Link 2025-09-20T06:27Z [---] followers, [--] engagements
"Tomorrow is the deadline I have to build an MVP ASAP else NGMI. Our presentation will include pictures of the UI so I am the bottleneck now this night will be legendary 💀"
X Link 2025-09-29T13:16Z [----] followers, [---] engagements
"@Arpit_2023 Is this that difficult I did the same using ChatGPT it was pretty smooth and fast"
X Link 2025-10-05T04:46Z [---] followers, [--] engagements
"@X_Ibyte @Pseudo_Sid26 It is good but pain the *. Run time is too big but the main problem is I never get optimal parameters with it. Maybe I am using it wrong or something I always fallback to Regular Grid Search"
X Link 2025-10-05T06:27Z [---] followers, [--] engagements
"@X_Ibyte @Pseudo_Sid26 We should still define the range right Maybe issues there how many combinations do you let it try"
X Link 2025-10-05T06:33Z [---] followers, [--] engagements
"@Arpit_2023 What is RTC 😅"
X Link 2025-10-05T10:51Z [---] followers, [--] engagements
"@Ndevsayzz Tf what else are you using bruh"
X Link 2025-10-05T13:49Z [---] followers, [--] engagements
"@farhanxansari Al the best bro"
X Link 2025-10-12T16:49Z [---] followers, [--] engagements
"15 October [----] Selected a PS built a prototype and recorded a demo video for SIH (took around [--] hours) Revised filtering layer nomenclature"
X Link 2025-10-15T19:58Z [----] followers, [---] engagements
"ML Engineer Internship - Strong Candidate Sharing the profile of Vaibhavi Thakur (@Messing_up20023) a pre-final year ML engineer who had been building production grade ML systems with a strong focus on data reliability label quality and evaluation loops. She has: Built end-to-end ML systems Worked on RAG systems and reliability-focused ML infrastructure Co-authored a research paper on Multimodal Fatigue Detection (NIT Rourkela) Hands-on experience shipping real systems. Vaibhavi is actively looking for ML/AI Engineering internships and would be a great fit for teams working on real-world ML"
X Link 2026-01-02T12:59Z [----] followers, 67.6K engagements
"19 October [----] Studied pre-training fine-tuning types of fine-tuning and stages of building LLMs Tried to understand Transformers Architecture tokenization self attention vector embeddings BERT vs GPT Project: confirmed the stack we will use free tools/models"
X Link 2025-10-19T19:57Z [----] followers, [---] engagements
"Instagram and TikTok are in a league of their own when it comes to algorithms. YouTube Spotify and X follow behind but none match how fast Instagram learns your mood. It's scary how good it's become at keeping you watching. Even though it's affecting attention spans and careers it remains a masterpiece both algorithmically and psychologically"
X Link 2025-10-24T14:48Z [----] followers, [---] engagements
"24 October [----] Revised LLM basics stages of building LLMs OOPs - mostly around classes and objects Project: Finally resumed. I built core modules to test project feasibility. I then downloaded Ollama and tested TinyLlama locally. Obviously it's not very great but with my GPU and RAM I am satisfied with anything I can possibly get. The next couple of days will be intense. I will try to code stuff from scratch"
X Link 2025-10-24T21:29Z [----] followers, [---] engagements
"When my cousin was [--] years old we realized he wasn't like other kids. While most children played with toys he spent hours flipping through books literally any book he could find. His parents noticed this early and encouraged him giving him every resource they could get their hands on. Then something incredible happened. By the time he turned [--] he could speak Marathi fluently. No one in the family spoke Marathi which made it even stranger. And one day he sat down at a piano and started playing simple melodies perfectly. That's when his parents realized something profound. Even though each book"
X Link 2025-10-25T14:42Z [----] followers, [---] engagements
"@farhanxansari It is notttttttt"
X Link 2025-10-25T15:14Z [---] followers, [--] engagements
"@dv_zsh ChatGPT is also trained on data from stackoverflow. So in a way or so we will continue to rely on it"
X Link 2025-10-25T21:13Z [---] followers, [--] engagements
"@OjasSharma276 Bruh wdym I am standing right next to you why talking to the wall"
X Link 2025-10-25T21:25Z [---] followers, [--] engagements
"@bit2swaz I dunno it is looking monotonous right now. Keep generating you might get a better one"
X Link 2025-10-26T04:36Z [---] followers, [--] engagements
"Today I'll be doing two lectures of LLMs. Will try to code as much as I can. I'll complete the OOPs revision and study DVC. I'll also complete the UI for my project. It will be a big day. Let's go"
X Link 2025-10-26T04:51Z [----] followers, [---] engagements
"26 October [----] Studied the evolution of GPT from transformers to GPT-5. A brief understanding of what's changed Revisited the working of GPT-3 Studied different types of tokenizations schemes Byte Pair Encoding Finally completed revising OOPs: Static method encapsulation setter and getter methods Project: Pushed everything to git. Created a streamlit UI using AI. Everything worked well. I'll take [--] more days and launch it here. Not satisfied with the grind 😕"
X Link 2025-10-26T22:00Z [----] followers, [---] engagements
"@LowKeyDevs Learn Python no matter what your main language is. It's like learning English it opens you up to a world of knowledge beyond your circle"
X Link 2025-10-27T05:30Z [---] followers, [--] engagements
"@krishdotdev Selective consumerism and extremely high product consciousness. India has all of it. Accounting for each rupee spent"
X Link 2025-10-28T10:09Z [---] followers, [---] engagements
"Ever wondered what people mean when they say "context window" in LLMs Think of an LLM like a human with limited short-term memory. The context window is that memory span - it decides how much text the model can "remember" while generating a response. GPT-3 had a context windows of 2K-4K tokens. GPT-4 Turbo 128K. Claude [---] Even more. Larger context = more room to think recall and reason. You can feed entire books conversations or codebases and still get coherent answers. But here's the catch - a bigger window doesn't mean perfect memory. LLMs don't store information like humans they re-read"
X Link 2025-10-29T04:30Z [----] followers, [---] engagements
"@OjasSharma276 I hope you are okay. What did you apply for though Like role"
X Link 2025-10-29T08:03Z [---] followers, [--] engagements
"One week of using a premium account. It has been good not the best. I'll keep analyzing what I can improve over the next [--] weeks. Thanks for all the engagement 😊"
X Link 2025-10-29T14:21Z [----] followers, [---] engagements
"Your text doesn't go straight into the model. It first passes through a tokenizer - a system that breaks words into small chunks called tokens. Example: understanding - under + standing LLMs don't read words. They read numbers"
X Link 2025-10-30T05:56Z [---] followers, [--] engagements
"Once tokenized your request turns into a long list of numbers. This is sent to OpenAI's servers (or whichever model host you're using). Every message carries context - everything you typed so far. That's how ChatGPT "remembers" your conversation"
X Link 2025-10-30T05:56Z [---] followers, [--] engagements
"Now the real magic begins. Inside giant clusters of GPUs the model looks at your tokens and predicts what comes next. one token at a time. The entire essay you got back Generated word by word. Prediction by prediction"
X Link 2025-10-30T05:56Z [---] followers, [--] engagements
"Each token depends on everything before it. That's why longer prompts slow it down - it's carrying your whole chat history each time. This running memory is called the context window. GPT-4 Turbo can handle up to [------] tokens (around a 300-page book)"
X Link 2025-10-30T05:56Z [---] followers, [---] engagements
"As each token is predicted it's streamed back to you in real time. That's why you see words appear one by one. You're literally watching the model "think.""
X Link 2025-10-30T05:56Z [---] followers, [--] engagements
"When the reply ends your entire chat (your message + its answer) is stored temporarily for continuity. Once the context window fills up older parts are forgotten. ChatGPT doesn't have memory. It has context"
X Link 2025-10-30T05:56Z [---] followers, [--] engagements
"@dv_zsh Bro's aura is unmatchable"
X Link 2025-10-31T16:19Z [---] followers, [--] engagements
"@kmeanskaran That's great I have spent a lot of time building strong fundamentals. If you are looking for ML or LLM Engineer roles I'd be a good fit"
X Link 2025-11-01T19:11Z [---] followers, [---] engagements
"November [----] Goals: Complete 2nd MLOps project Update resume & start applying Land an ML Engineer internship Finish MLOps lectures (3/25 done) Grow to 1K followers on X (470 now) & [---] verified (160 now) Complete half of the LLM lectures (currently at 9/43 done) Win a hackathon Score 75%+ in mid-terms This month will be intense. Mid-terms in week [--] lectures piling up projects and hackathons waiting. Somehow I've got to balance it all and land that internship and connect with more people here. Also. it's my birthday month 😊"
X Link 2025-11-02T05:13Z [----] followers, [----] engagements
"@aaravshukla_10 Give ChatGPT all your details and ask for a good resume in LaTeX format. Put it on Overleaf. Keep iterating your code using GPT. It will probably take 1-2 days. But you'll end up with a very high quality resume"
X Link 2025-11-02T07:10Z [---] followers, [--] engagements
"@aaravshukla_10 Yes. You can post it here for further feedback. We have an incredible community"
X Link 2025-11-02T07:17Z [---] followers, [--] engagements
"I am working on an open-source RAG system focused on efficient hybrid retrieval and evaluation. It's been [--] days since I started the project. With mid-terms coming up next week I've decided to finish it tonight no matter what. I will be working for the next [--] hours taking little breaks. Let's see how far we can go"
X Link 2025-11-02T09:00Z [----] followers, 20K engagements
"@AkwaIbomBreed @harshalvk_ Learn Git/GitHub OOPs Data Versioning using DVC CI using mlflow Dockerization CD on AWS Kubernetes Prometheus and Grafana. Just learn these concepts individually from any tutor. That should be a great start"
X Link 2025-11-03T13:19Z [---] followers, [--] engagements
"The most dangerous thing about AI isn't that it lies - it's that it lies confidently. You ask a model something simple. It replies instantly flawlessly persuasively and yet it's wrong. That's called a hallucination. Large Language Models don't know things. They don't store facts or check truth. They just predict the next word based on patterns they've seen before. So when the model doesn't know something it doesn't say "I'm not sure." It guesses. But it guesses beautifully. It's like that one kid in class who didn't study but still wrote a full essay with perfect grammar and total confidence."
X Link 2025-11-05T15:25Z [----] followers, [----] engagements
"@autom8nerd Haha. I remember that. I don't know how these things work but I feel using RL for such tasks is better than training under supervision"
X Link 2025-11-05T17:01Z [---] followers, [--] engagements
"It's my 🎈 day - just turned [--]. My first birthday on X. I've met many awesome people here since I joined in June. Hopefully by my next birthday you'll look at me as someone who adds real value here"
X Link 2025-11-06T19:10Z [----] followers, [----] engagements
"@OjasSharma276 @RockstarGames 🤣🤣"
X Link 2025-11-07T03:52Z [---] followers, [--] engagements
"8 Nov [----] IDA Mid Prep: Clustering Analysis K-means K-mediods Agglomerative Divisive Distance Metrics for Clustering Algorithms DB Scan Evaluation of Clustering PRML Mid Prep: Perceptron Algorithm Support Vector Machine Hard and Soft Margin SVM"
X Link 2025-11-08T20:34Z [----] followers, [----] engagements
"9 November [----] PRML Mid Prep: Kernel Function Linear Fisher Polynomial Gaussian Radial Basis Function Cosine Two-class problem Hard constraint SVM Soft Margin SVM. Linear models for classification Discriminant functions fisher's discriminant Probabilistic discriminative and generative models"
X Link 2025-11-09T19:22Z [----] followers, [---] engagements
"12 November [----] CNS Mid Prep: RSA Encryption Diffie-Hellman Key Agreement SHA-512 Algorithm Kerberos Password Management SSL Architecture SignCryption NLP Mid Prep: Discourse Analysis Dialogue Analysis QA System Rule Based and Probabilistic Parsing NER"
X Link 2025-11-12T20:21Z [----] followers, [---] engagements
"@Sarthak4Alpha @sachinyadav699 San Francisco"
X Link 2025-11-15T20:35Z [---] followers, [--] engagements
"@devcelio How come I never heard of this. This is awesome"
X Link 2025-11-16T17:25Z [---] followers, [----] engagements
"I've delayed the project launch again this time it's because of the front-end. The entire repository is ready but I used Bolt for the front-end and the output looks way more polished than my original version. Now I'm stuck in a dilemma: should I launch the boring current version or take the pain update the front-end properly and then release it"
X Link 2025-11-17T20:19Z [---] followers, [---] engagements
"19 November [----] Token embeddings how are they created for LLMs coding embedding layers intuition behind embedding layers and linear neural networks"
X Link 2025-11-19T19:46Z [---] followers, [---] engagements
"If you want to learn the Theory of computation this is the best book I've come across: "Theory of Computer Science (Automata Languages and Computation)" by K.L.P. Mishra and N. Chandrasekharan. @philearning I used it myself. The explanations of automata grammars turing machines decidability and complexity are very clear. You don't need to be a hardcore theory person; you can open any chapter and understand the concepts without getting lost. And if you're into AI/ML TOC actually matters more than people think: Automata helps you understand how parsers and tokenizers work. Grammars relate"
X Link 2025-11-20T10:54Z [---] followers, [---] engagements
"I just promoted Nano Banana 🍌 Pro: "Visualise how LLMs work from the very foundations as an academic image that can be used for illustration purposes. I need neatly labelled descriptions." And here's the output. Honestly I'm impressed with what it generated. Seeing an AI generated image with zero typos actually made my day"
X Link 2025-11-21T06:00Z [---] followers, [---] engagements
"@autom8nerd Yes. ChatGPT sucks at image generation. It produces inaccurate images with a lot of typos"
X Link 2025-11-21T08:47Z [---] followers, [--] engagements
"@bit2swaz Yes. I am finding this model good. I will test it more"
X Link 2025-11-21T08:48Z [---] followers, [--] engagements
"@karanjagtiani04 I didn't have to version the embeddings. It's plug and play just swap components and get the results"
X Link 2025-11-21T08:55Z [---] followers, [--] engagements
"@autom8nerd Yes agreed. I presume there must be some way to deal with that as well"
X Link 2025-11-21T13:15Z [---] followers, [--] engagements
"@shivamxtech I got many more. Stay tuned 😉"
X Link 2025-11-22T09:39Z [---] followers, [--] engagements
"@Rushu_Tushu They're still general LLMs. They just get way more code training and coding-specific optimization. That's why they write code well and chat normally"
X Link 2025-11-22T15:54Z [---] followers, [--] engagements
"@devsterxyz That's some good intuition. It's actually interesting if you go a little deeper"
X Link 2025-11-22T20:37Z [---] followers, [--] engagements
"I really pushed my limits for this project. It took a lot of effort. Since RAG is trending right now I picked up that direction and built a system that combines keyword search and semantic search. My first step toward grounding LLMs to a real knowledge base and reducing hallucinations. Please check it out and comment on what you feel about it. I spent the last few weeks building VectorFlow-RAG a full semantic search + QA system because I wanted to understand what it actually takes to ship an ML system end-to-end. Why I Built It Keyword search breaks when you need semantic understanding. Pure"
X Link 2025-11-23T17:02Z [---] followers, [---] engagements
"We've spent decades trying to catch liars by reading their faces. Turns out the most reliable signal isn't a micro expression or a twitch - it's heat. This paper shows that thermal patterns around the eyes can reveal deception far more consistently than human intuition. The authors tracked periorbital temperature variations with a mid-infrared camera and analysed [---] lie-truth responses. The authors captured [---] responses (249 lies [---] truths) and tracked [----] pixel periorbital patches across [--] frames (1.7 seconds) around each answer. The result is a compact 102-dimensional thermal signature"
X Link 2025-11-24T06:11Z [---] followers, [---] engagements
"India is fighting over a piece of invisible real estate - the [--] GHz spectrum. And it might decide whether your next home Wi-Fi is lightning fast or painfully mid. Most people don't even know this fight is happening but here's why it actually matters. [--]. First: Why is [--] GHz such a big deal Think of internet signals as traffic on roads. Your current Wi-Fi runs on [---] GHz (slow crowded) and [--] GHz (faster but still crowded) [--] GHz is a brand new empty highway. No traffic no interference huge speeds. This is the band that unlocks Wi-Fi 6E and Wi-Fi [--] which can genually hit multi-gigabit speeds at"
X Link 2025-11-24T17:09Z [---] followers, [---] engagements
"@karthikponna19 Jupyter notebook support as well"
X Link 2025-11-25T11:51Z [---] followers, [--] engagements
"@OjasSharma276 @karthikponna19 Unit [--] - How to fine-tune an LLM from Scratch"
X Link 2025-11-26T08:57Z [---] followers, [--] engagements
"29 November [----] Worked on project [--] today - easily my most productive day so far. Vibe-coded the entire backend and it boosted my speed like crazy. Almost finished the whole project which I didn't expect. Didn't study but I did code an LLM tokenizer from scratch in Python"
X Link 2025-11-29T20:23Z [---] followers, [---] engagements
"Project 3: Ran into issues with docker-compose last night. It's supposed to start [--] services but only one comes online. Not sure if it's an endpoint issue or something inside Docker but it's taking a lot of time to debug"
X Link 2025-11-30T08:58Z [---] followers, [---] engagements
"@Rushu_Tushu Yeah. It's done. Now I am working on hosting the repo on render"
X Link 2025-11-30T11:48Z [---] followers, [--] engagements
"1 December [----] LLMs: Attention Types of Attention Mechanisms Problems with Longer Sequences Working of RNNs. Project 3: Still working on deploying the back-end on render. There is some issue with the environment variables I presume. Trying to debug"
X Link 2025-12-01T20:18Z [---] followers, [---] engagements
"Tried the new X Android update. The animations are great and the UI feels fresh but it was so different that I switched back within [--] minutes. It completely changed how X feels and I'm not sure I can get used to it"
X Link 2025-12-08T06:03Z [---] followers, [----] engagements
"@original_ngv @Starlink @elonmusk Did you do a speed test"
X Link 2025-12-08T08:40Z [---] followers, [--] engagements
"Most people think attention was introduced by the [----] "Attention Is All You Need" - but that's not true. The real origin is a [----] paper by Bahdanau Cho and Bengio: Neural Machine Translation by Jointly Learning to Align and Translate. That paper introduced what we now call "attention": instead of compressing an input sentence into a single vector the decoder dynamically attends to different encoder hidden states per output token. That solved the fixed length bottleneck for long sentences. In [----] the Transformer architecture replaced recurrence entirely - stacking self attention (and cross"
X Link 2025-12-10T15:37Z [---] followers, [----] engagements
"Practical 1/3 - Pattern Recognition & Machine Learning I reached late and wasn't allowed to sit for the exam at first. After requesting the professor multiple times she allowed me to take the test with other late comers. I couldn't cheat properly. During the implementation session I just copy pasted the code from ChatGPT. VIVA was simple af. The external asked: "Where do we use Python" "Difference between supervised and unsupervised learning." "Train-test split ratios." "What is reinforcement learning" "Real world applications of pattern recognition." It was an okayish day. VIVA cheered me up"
X Link 2025-12-12T19:37Z [---] followers, [---] engagements
"Practical 2/3 - Intelligent Data Analysis Reached on time today. The test started late. I couldn't cheat because I'm not used to it. I didn't write the code. I implemented the same KMeans algorithm using ChatGPT but the evaluator didn't give me any marks because my answer sheet is empty. I mean what's the point of implementing then VIVA kept postponing because my professor is flirting with the external. The evaluator changed thrice and the final one asked me about: "What is dimensionality reduction" "Support vector machines are used for" "What is regression" I fumbled a bit at the SVM"
X Link 2025-12-13T14:23Z [---] followers, [---] engagements
"Presented my group's project today (even though I barely worked on it). The project itself was entirely AI-generated. My teammates asked me to generate the report which I did. We didn't even manage to make a proper PPT due to time constraints. A good chunk of our time went into me explaining the project to my own teammates before explaining it to the evaluator. The external didn't help either - most of the questions were irrelevant and confusing. Easily the worst VIVA experience I've had. One thing became clear today: I don't work very well in teams yet. That's something I need to improve"
X Link 2025-12-17T13:49Z [---] followers, [---] engagements
"AI/ML can feel overwhelming and there's a good reason for that. New architectures papers tools and techniques show up almost every day. If you pause for too long the gap becomes noticeable very quickly. Not because you're bad at it but because the field itself is moving fast. This isn't hype. AI/ML is one of the few domains where research industry and tooling evolve simultaneously. Papers turn into production systems in months. What was state-of-the-art last year can become baseline knowledge today. The mistake is thinking you need to learn everything. You don't. The people who survive"
X Link 2025-12-22T13:25Z [---] followers, [--] engagements
"@yugb_twt LinkedIn and Cold Mailing ☠"
X Link 2025-12-23T06:49Z [---] followers, [--] engagements
"AI/ML can feel overwhelming and there's a good reason for that. New architectures papers tools and techniques show up almost every day. If you pause for too long the gap becomes noticeable very quickly. Not because you're bad at it but because the field itself is moving fast. This isn't hype. AI/ML is one of the few domains where research industry and tooling evolve simultaneously. Papers turn into production systems in months. What was state-of-the-art last year can become baseline knowledge today. The mistake is thinking you need to learn everything. You don't. The people who survive"
X Link 2025-12-23T12:33Z [---] followers, [---] engagements
"Took [--] AI interviews on Internshala today. Honestly it felt a bit soulless. Maybe I'm just more used to human evaluators. The first two interviews were mostly communication based. The third was Data Science (didn't go great) but the fourth was Machine Learning and that one went surprisingly well. The questions were basically overall. I'm still unsure about the internshala though. It feels slow and exhausting and I'm not convinced people actually land internships through it. Has anyone here got an internship/job from Internshala I'd like to know your experience"
X Link 2025-12-25T14:58Z [---] followers, [----] engagements
"@autom8nerd Reddit is 🐐"
X Link 2025-12-25T16:54Z [---] followers, [--] engagements
"@Rushu_Tushu Thanks. I will pivot to other platforms"
X Link 2025-12-26T06:27Z [---] followers, [--] engagements
"@tm23twt WTF is this. Why am I seeing Machine Learning and Thermodynamics on the same paper"
X Link 2025-12-26T11:43Z [---] followers, [--] engagements
"Looking for an ML Engineering Internship. I've built end-to-end ML pipelines worked on RAG systems and co-authored a research paper. I also have research internship experience at NIT Rourkela. I'm actively applying but I'd really appreciate referrals leads or even advice. If you know someone hiring ML Engineers or if your company is open to interns please let me know. Resume attached. Any help means a lot"
X Link 2025-12-26T13:09Z [----] followers, 81K engagements
"Reading Karpathy says he feels behind is strangely comforting. If he feels FOMO one of the best minds we have it shows how insanely fast this field is moving. AI is rewriting the role of a programmer in real-time. New abstractions new layers new ways of thinking. You aren't supposed to have it all figured out none of us do. The only real skill now is staying in motion experimenting and learning faster than yesterday. If the legends are adapting we should too. I've never felt this much behind as a programmer. The profession is being dramatically refactored as the bits contributed by the"
X Link 2025-12-26T17:47Z [----] followers, [----] engagements
"@munkeeDLuffy Bruh. They'll kill me when you use such terms I am hardly surviving 🥲"
X Link 2025-12-27T13:34Z [---] followers, [--] engagements
"@OjasSharma276 Broski please get a Mac. You'll never regret listening to me"
X Link 2025-12-28T13:02Z [---] followers, [----] engagements
"We're entering a strange phase in human-machine cognition. Large language models make high-level reasoning instantly accessible reducing the need to hold information in working memory or recall facts manually. Research on cognitive offloading already shows that when tools perform thinking steps for us our brain naturally delegate tasks to them. The pattern is visible with calculators maps and now reasoning itself. But something interesting is happening in parallel. When used deliberately LLMs also amplify human reasoning by generating alternate perspectives exposing blind spots and enabling"
X Link 2025-12-28T13:06Z [----] followers, [---] engagements
"@OjasSharma276 I use M1 the basic model. It's smooth like butter. It feels better than [--] GB RAM Victus Asus Acer laptops"
X Link 2025-12-28T16:21Z [----] followers, [---] engagements
"@RanjanRavi00 @vaemaski @OjasSharma276 I never happened to use a MacBook that heats more than Windows"
X Link 2025-12-29T08:18Z [---] followers, [--] engagements
"@autom8nerd @StillTypeWritin Big brother is not responding I already contacted 😔"
X Link 2025-12-29T12:25Z [---] followers, [--] engagements
"@devsterxyz Cracked dev is still an uncertainty become a good communicator instead"
X Link 2025-12-29T16:28Z [---] followers, [--] engagements
"We finally hit [----] 🎊 I created this account on June [--] [----]. What started as random posts slowly became a habit then a discipline and now a part of my routine. There were days when I wanted to quit no reach no growth no validation. I'd watch others go viral while I posted quietly into the void. Many times I wondered if it was even worth it. But I kept showing up. Even when it felt slow. Even when it felt pointless. Even when consistency felt like a burden. I used to tell @OjasSharma276 that I'd hit 1K by the year-end. At the time I didn't fully believe myself but I kept going anyway. And"
X Link 2025-12-31T15:37Z [----] followers, [---] engagements
"Goals for [----] Become financially independent Focus on physical fitness Never worry about unemployment Improve communication skills Reach 10K followers on X Start monetization on X Win atleast [--] hackathons Deep dive into LLMs and MLOps Improve programming skills Act on good habits immediately Stop aimless scrolling Reduce mobile usage https://twitter.com/i/web/status/2006714420278272055 https://twitter.com/i/web/status/2006714420278272055"
X Link 2026-01-01T13:09Z [----] followers, [---] engagements
"Goals for [----] Become financially independent Focus on physical fitness Never worry about unemployment Improve communication skills Reach 10K followers on X Start monetization on X Win atleast [--] hackathons Deep dive into LLMs and MLOps Improve programming skills Act on good habits immediately Stop aimless scrolling Reduce mobile usage https://twitter.com/i/web/status/2006730256376533338 https://twitter.com/i/web/status/2006730256376533338"
X Link 2026-01-01T14:12Z [----] followers, [---] engagements
"@Messing_up20023 Prompt engineering is pretty essential these days. Happy that you are covering it 🫂"
X Link 2026-01-03T04:43Z [----] followers, [--] engagements
"I get a lot of DMs asking for good AI/ML/DSA resources. But the honest truth I don't know if the ones I followed are the best. So let's settle this once and for all let's crowdsource the best resources. If you're learning or working in any AI/ML domain drop what actually helped you. You can format it like this: ML/DL - Hands-on Machine Learning with Scikit-Learn and PyTorch by Aurlien Gron (Book) LLMs - Building LLMs from scratch by Vizuara (YT Playlist) MLOps - . CV/NLP/RL - . Books are preferred but high-quality playlists or courses are welcome too. Feel free to share full roadmaps - ML"
X Link 2026-01-04T08:43Z [----] followers, [----] engagements
"@Pushkar11420566 Precise and to the point 🤝🏻"
X Link 2026-01-04T09:09Z [----] followers, [--] engagements
"PATTERN RECOGNITION and MACHINE LEARNING - B. Tech. (Hon's) (Fifth Semester) Examination UNIT 1: [--]. (a) Explain the role of covariance in multivariate Gaussian distributions and justify how it influences the shape and orientation of the PDF. [--]. (b) Derive the expectation and variance of a linear combination of random variables. Analyse how these results support the Law of Large Numbers [--]. (c) Suppose a random vector x = (x1 x2) follows a multivariate Gaussian. Write down its joint PDF and interpret its parameters. [--]. (d) Formulate and justify a decision rule that minimizes expected loss in"
X Link 2026-01-06T09:50Z [----] followers, [----] engagements
"@CanadaPRaspiran This is the reference book"
X Link 2026-01-06T14:06Z [----] followers, [--] engagements
"INTELLIGENT DATA ANALYSIS - B. Tech. (Hon's) (Fifth Semester) Examination UNIT 1: [--]. (a) Explain the curse of dimensionality. [--]. (b) Differentiate between mean median and mode with one example. [--]. (c) What are the various methods of measuring data dispersion Illustrate with an example dataset. [--]. (d) Explaine the role of data transformation and data discretization in the pre-processing phase. Provide suitable examples. UNIT 2: [--]. (a) What is the principle of cross-validation Why is it used [--]. (b) Compare decision tree rule-based classifier and nearest-neighbor classifier based on accuracy"
X Link 2026-01-08T12:01Z [----] followers, [----] engagements
"@Pseudo_Sid26 Bro is a formula [--] racer 🚗"
X Link 2026-01-10T11:52Z [----] followers, [--] engagements
"@Pseudo_Sid26 Amazing article Sid. You are doing great 🔥"
X Link 2026-01-11T10:08Z [----] followers, [---] engagements
"AI feels less like a tool and more like a craft. Easy to use but hard to master. You can pick up AI products quickly and pivot into adjacent roles but building a stable career here is different. It demands patience depth and years of compounding effort. This isn't a field where shortcuts work. If you're not ready to put in at least 4-5 years expect very little in return. I've never seen a domain evolve this fast. What you learn today can feel outdated tomorrow. The only way to survive is to keep learning and keep building. https://twitter.com/i/web/status/2010388722051924340"
X Link 2026-01-11T16:29Z [----] followers, [---] engagements
"NATURAL LANGUAGE PROCESSING - B. Tech. (Hon's) (Fifth Semester) Examination UNIT 1: [--]. (a) Define Natural Language Processing. What are its main goal [--]. (b) Describe the key challenges of Natural Language Processing. [--]. (c) Explain the importance of NLP in sentiment analysis. [--]. (d) Write the history of Natural Language Processing. UNIT 2: [--]. (a) What is the Markov assumption [--]. (b) Explain in detail how regular expressions and automata are used in NLP. [--]. (c) Explain the three branches of phonetics: articulatory acoustic and auditory. [--]. (d) Describe different word-formation processes in"
X Link 2026-01-12T09:27Z [----] followers, [---] engagements
"@Dev__Vishwajeet This semester will be intense though a lot of work and uncertainty"
X Link 2026-01-15T17:43Z [----] followers, [--] engagements
"1200 now. [----] - [----] took [--] days 🙂 We are [----] now. [----] - [----] took [--] hours 🙃 https://t.co/os5CRDmcLU We are [----] now. [----] - [----] took [--] hours 🙃 https://t.co/os5CRDmcLU"
X Link 2026-01-18T02:19Z [----] followers, [----] engagements
"I'm currently piled up with multiple research opportunities from multiple IITs. I'll be picking one or two soon and share an update. The last couple of days were a bit rough so I went quiet. Back now"
X Link 2026-01-21T17:09Z [----] followers, [----] engagements
"Starting February [--] I'm committing the next [--] days to sharpening my fundamentals the right way. Python GitHub Linux Shell Scripting Consistent daily work until these tools feel natural and no more intimidating. If you're also working on your fundamentals this is your sign to go all in. https://twitter.com/i/web/status/2015127013171032190 https://twitter.com/i/web/status/2015127013171032190"
X Link 2026-01-24T18:18Z [----] followers, [----] engagements
"AI makes coding faster and prototyping easier but it doesn't replace learning to code. Models can generate decent code sure but when something breaks in production you can't rely on someone who doesn't understand what's happening. Real skill is knowing how to debug fix and take responsibility when things go wrong"
X Link 2026-01-29T18:34Z [----] followers, [---] engagements
"@_Creation22 Some OCR stuff AI Validation and Blockchain"
X Link 2026-02-06T17:14Z [----] followers, [--] engagements
"If you're hiring ML/AI interns look at this 👇🏻 @Yash_xml has experience with LLMs agents deep learning and internships at IBM & EY. Resume attached. Please RT to help him reach the right people"
X Link 2026-01-23T13:38Z [----] followers, 19.4K engagements
"The lab gave me a PC with 32GB RAM and an i7-10700 basically a beast for research work. It came with a pirated Windows [--] Education copy so I waited patiently thinking they'd eventually provide a proper key but I'm done waiting. Installing Ubuntu [-----] now and setting it up the way I want. This is going to be my workstation for the next 5-6 months so it better be clean stable and ready for some serious work"
X Link 2026-02-09T10:10Z [----] followers, 43.9K engagements
"This striked my mind today. Skip connections and residual blocks may appear to be identical but that's not the case. A skip connection is a technique while a residual block is a structural component that utilizes that technique. Skip connections are direct pathways that skips one or more layers allowing information to bypass non-linearities. Residual block is a specific architectural unit where a skip connection is used to add the input x to the output of set of layers F(x) resulting in F(x) + x. This is commonly seen in ResNet architectures. To put it simply all residual block use skip"
X Link 2026-02-10T13:16Z [----] followers, [----] engagements
"Unpaid internships are almost never worth it matter how desperate you feel. If a company can't pay you they usually won't mentor you either. You'll end up doing random work burning time and calling it experience. Upskill instead. Be patient. Build projects. Network. Apply better. The only exception I'd consider is a genuine research internship where you're actually learning and working closely with a good mentor. But unpaid industry internships Don't normalize that"
X Link 2026-02-11T14:03Z [----] followers, [----] engagements
"RAG (Retrieval Augmented Generation) - what's the actual point People talk about RAG a lot but the idea is simple: LLMs are great at language not memory. They don't store facts. They don't know your PDFs. They just predict the next token. RAG fixes that. Here's how it actually works: [--]. Retrieval You take your documents chunk them convert each chunk into embeddings and store them in a vector database. When a query comes in you embed the query and pull the top-k most relevant chunks using vector similarity. [--]. Augmentation You attach those retrieved chunks to the user query. This gives the"
X Link 2025-11-21T13:20Z [----] followers, [---] engagements
"Reached the lab. Today's reading list: DnCNN DRUNet Restormer and SCUNet. All of them are strong baselines for image restoration and I want to understand what makes each one work. Let's see how much I can cover today"
X Link 2026-02-11T09:02Z [----] followers, [----] engagements
"@tusharuii Depends on your interest. If you are okay with CNNs start implementing U-Net"
X Link 2026-02-13T03:34Z [----] followers, [---] engagements
"@VarunrRao I started in my 3rd year. Feeling FOMO 😮"
X Link 2026-02-13T16:21Z [----] followers, [--] engagements
"Working on Sundays so I can make weekdays easier"
X Link 2026-02-15T04:16Z [----] followers, [---] engagements
"If you already have a basic understanding of deep learning and you're in your 2nd or 3rd year start reading papers and implementing them. Don't wait until you feel ready. That moment never comes. Use AI tools to break down the paper understand the intuition and replicate the code step by step. You'll learn more from one implementation than from weeks of passive courses. And if PyTorch feels like a barrier just watch a good one shot PyTorch video and start building immediately. You'll thank youself later. https://twitter.com/i/web/status/2022146198132732325"
X Link 2026-02-13T03:09Z [----] followers, [----] engagements
"Valentine's Day started with a [--] AM meeting with my supervisor. [--] minutes later I got humbled at least a hundred times. I couldn't clearly explain convolutions there was miscommunication and I realized I don't understand some of the architectures deeply enough for the literature review I'm doing. The intuition wasn't there and it showed. Now I'm back in my room with a lot to revisit and half my valentine's Day is already gone. Research really doesn't care what day it is. https://twitter.com/i/web/status/2022571053273493549 https://twitter.com/i/web/status/2022571053273493549"
X Link 2026-02-14T07:18Z [----] followers, [---] engagements
"10 hours into this project today. I've been using ChatGPT-5.2 and Perplexity Deep Research so far but at some point the pressure hit hard and genuinely felt like I was stuck and would be trapped in this project forever. Then I installed GPT-5.3 Codex just to see what the hype was about and honestly this is scary. I've never designed and built things this fast in my life I was cynical about context window and rate limits but for the kind of changes and iterations I needed it was more than sufficient. I still have around [--] more hours of work left but I'm pretty sure the day is going to end"
X Link 2026-02-15T14:09Z [----] followers, [---] engagements
"I am working on an open-source RAG system focused on efficient hybrid retrieval and evaluation. It's been [--] days since I started the project. With mid-terms coming up next week I've decided to finish it tonight no matter what. I will be working for the next [--] hours taking little breaks. Let's see how far we can go"
X Link 2025-11-02T09:00Z [----] followers, 20K engagements
"It's my 🎈 day - just turned [--]. My first birthday on X. I've met many awesome people here since I joined in June. Hopefully by my next birthday you'll look at me as someone who adds real value here"
X Link 2025-11-06T19:10Z [----] followers, [----] engagements
"November [----] Goals: Complete 2nd MLOps project Update resume & start applying Land an ML Engineer internship Finish MLOps lectures (3/25 done) Grow to 1K followers on X (470 now) & [---] verified (160 now) Complete half of the LLM lectures (currently at 9/43 done) Win a hackathon Score 75%+ in mid-terms This month will be intense. Mid-terms in week [--] lectures piling up projects and hackathons waiting. Somehow I've got to balance it all and land that internship and connect with more people here. Also. it's my birthday month 😊"
X Link 2025-11-02T05:13Z [----] followers, [----] engagements
"I finally completed my first research paper 💀 I'll submit somewhere early september. It is [--] pages long"
X Link 2025-08-08T17:36Z [----] followers, [----] engagements
"What the actual fak. @cneuralnetwork -- this guy inspired me to start posting on X. Not many people know that my account name itself is inspired by his. So it took him [---] days for him to notice me. Not bad. Hope it's not a mistake 😭"
X Link 2025-10-23T18:29Z [----] followers, [----] engagements
"October [----] Rewind: Studied CNNs Did two submissions for SIH (many sleepless nights) Participated in two more hackathons Started LLMs and MLOps Finished a RAG project Got gifted a premium account X Grew from [---] to [---] followers Honestly this month wasn't anything special. But I learnt a lot in the process. I am more serious now about building on X and sharpening my skills. Also met some amazing people here this month"
X Link 2025-11-01T12:55Z [----] followers, [----] engagements
"600 🙂 This growth needs to be studied. It took me [--] hours from [---] - [---] [---] followers it is. Took [--] hours to go from [---]. Grateful for the amazing people on this platform. https://t.co/fU5Frc4rVw [---] followers it is. Took [--] hours to go from [---]. Grateful for the amazing people on this platform. https://t.co/fU5Frc4rVw"
X Link 2025-11-03T13:04Z [----] followers, [----] engagements
"500 followers it is. Took [--] hours to go from [---]. Grateful for the amazing people on this platform. We are at [---] now. It took [--] days and [--] hours from [---] 🦧 https://t.co/QYZbHRsDR2 We are at [---] now. It took [--] days and [--] hours from [---] 🦧 https://t.co/QYZbHRsDR2"
X Link 2025-11-02T14:06Z [----] followers, [----] engagements
"Thanks to @sharpeye_wnl and @TanayVasishtha for sponsoring my Premium. All the best to you guys as well 🫂 Got my blue tick within [--] minutes of payment which is super fast Feels like I'm on the right path after all. As part of refactoring my account I removed around [--] followers who were basically bots. Not sure if that's the best move algorithmically but I hope to build a genuine community that engages and responds. Since I have a month of these benefits I'll try to engage more post consistently and learn from you all. Reach is pretty rough right now but I'll focus on adding value. Usually"
X Link 2025-10-23T15:39Z [----] followers, [----] engagements
"@yikoshi Holding hands and crying in her lap"
X Link 2025-11-01T16:41Z [----] followers, [----] engagements
"8 Nov [----] IDA Mid Prep: Clustering Analysis K-means K-mediods Agglomerative Divisive Distance Metrics for Clustering Algorithms DB Scan Evaluation of Clustering PRML Mid Prep: Perceptron Algorithm Support Vector Machine Hard and Soft Margin SVM"
X Link 2025-11-08T20:34Z [----] followers, [----] engagements
"Guess who didn't sleep all night just for making a PPT. There you go it's me 🤥 I have my internal hackathon tomorrow (we'll technically today) wish me luck"
X Link 2025-09-24T23:34Z [----] followers, [----] engagements
"Time to switch gears. I just wrapped up my RAG project - easily one of the toughest and most rewarding things I've built so far. Mid-terms are now around the corner so I probably won't be diving deep into LLMs or MLOps until they're done. I'll share the project repo after exams - once the documentation is solid. No compromises here. Exams will be over by the 15th. Until then I'll post small things I'm learning - tiny concepts insights and reflections. I know why most of you are here - it's for LLMs and MLOps. I'll do my best to stay consistent and keep adding value even during this break."
X Link 2025-11-04T14:46Z [----] followers, [---] engagements
"9 November [----] PRML Mid Prep: Kernel Function Linear Fisher Polynomial Gaussian Radial Basis Function Cosine Two-class problem Hard constraint SVM Soft Margin SVM. Linear models for classification Discriminant functions fisher's discriminant Probabilistic discriminative and generative models"
X Link 2025-11-09T19:22Z [----] followers, [---] engagements
"26 October [----] Studied the evolution of GPT from transformers to GPT-5. A brief understanding of what's changed Revisited the working of GPT-3 Studied different types of tokenizations schemes Byte Pair Encoding Finally completed revising OOPs: Static method encapsulation setter and getter methods Project: Pushed everything to git. Created a streamlit UI using AI. Everything worked well. I'll take [--] more days and launch it here. Not satisfied with the grind 😕"
X Link 2025-10-26T22:00Z [----] followers, [---] engagements
"The most dangerous thing about AI isn't that it lies - it's that it lies confidently. You ask a model something simple. It replies instantly flawlessly persuasively and yet it's wrong. That's called a hallucination. Large Language Models don't know things. They don't store facts or check truth. They just predict the next word based on patterns they've seen before. So when the model doesn't know something it doesn't say "I'm not sure." It guesses. But it guesses beautifully. It's like that one kid in class who didn't study but still wrote a full essay with perfect grammar and total confidence."
X Link 2025-11-05T15:25Z [----] followers, [----] engagements
"25 October [----] Revised Pre-training Fine-tuning Fine-tuning techniques building schematic simplified transformers architecture some more concepts Coded an LLM Tokenizer from scratch using regular expressions and python Project: Completed the skeleton of the project. All modules are working well. I debugged LLM client errors. Created a script for an integrated pipeline combining all modules it took a lot of time. The UI is pending (not very keen on). Further refinement and add-ons are to be made"
X Link 2025-10-25T21:25Z [----] followers, [----] engagements
"It took roughly a month this time to gain the next [--] followers but it's worth it. The last [--] have been of high quality 😃 I think I can finally say this now that I have a small community here: [---] it is https://t.co/XGhbLFsoXE I think I can finally say this now that I have a small community here: [---] it is https://t.co/XGhbLFsoXE"
X Link 2025-10-23T20:32Z [----] followers, [----] engagements
"I think I can finally say this now that I have a small community here: [---] it is Okay so this happened as well 😮 https://t.co/dyyx5IHjFv Okay so this happened as well 😮 https://t.co/dyyx5IHjFv"
X Link 2025-09-28T17:15Z [----] followers, [----] engagements
"28 October [----] Studied the need for data-target pair creation. Looked at custom dataset and data loader generation The lecture was dense and I couldn't complete it. Didn't do any project work or MLOps as well for the same reason"
X Link 2025-10-28T21:45Z [----] followers, [---] engagements
"One week of using a premium account. It has been good not the best. I'll keep analyzing what I can improve over the next [--] weeks. Thanks for all the engagement 😊"
X Link 2025-10-29T14:21Z [----] followers, [---] engagements
"@abd_mukadam It's a career update for the company"
X Link 2025-10-28T06:47Z [----] followers, [---] engagements
"We are at [---] now. It took [--] days and [--] hours from [---] 🦧 We're at [---] now. It took [--] hours this time - not for [--] but for [---]. It blew up so fast I couldn't even post anything at [---] 🙂 https://t.co/tkofzjzvv1 We're at [---] now. It took [--] hours this time - not for [--] but for [---]. It blew up so fast I couldn't even post anything at [---] 🙂 https://t.co/tkofzjzvv1"
X Link 2025-10-30T16:58Z [----] followers, [----] engagements
"We're at [---] now. It took [--] hours this time - not for [--] but for [---]. It blew up so fast I couldn't even post anything at [---] 🙂 It took roughly a month this time to gain the next [--] followers but it's worth it. The last [--] have been of high quality 😃 https://t.co/mk1XyPZVD3 It took roughly a month this time to gain the next [--] followers but it's worth it. The last [--] have been of high quality 😃 https://t.co/mk1XyPZVD3"
X Link 2025-10-26T09:30Z [----] followers, [----] engagements
"It took me [--] days to reach my first [--] followers and just [--] days to reach the next [--]. That's [--] days on this app so far. Greatful for the journey and all the learnings thank you 🙏🏻"
X Link 2025-09-17T06:01Z [----] followers, [---] engagements
"1 November [----] The entire CI/CD pipeline was setup nicely. I have created a bunch of unit integration and performance tests. I had to spend a good amount of time writing and debugging these. This project is slowly moving towards MLOps. I want to demonstrate all my skills. Couldn't study LLMs today"
X Link 2025-11-01T20:12Z [----] followers, [---] engagements
"29 October [----] Revised LLM Intro classes & objects constructor initializer method vs functions special methods encapsulation getter and setter methods static function. Revised tokenization algorithm types BPE algorithm input-target pair generation custom dataset and data loader creation from Pytorch components. Studied batch_size and num_workers in great detail. These will help me with model training"
X Link 2025-10-29T20:24Z [----] followers, [---] engagements
"27 October [----] Revised stages of building LLMs Revised OOPs - second lap Project: Did not do much. I understood that there are several issues with multiple modules of my project. I tried to understand my code better and understand limitations. Today was a L. Least productive day ever"
X Link 2025-10-27T20:53Z [----] followers, [---] engagements
"I talked to @Divyansh91565 last night. He is a humble guy and helped me getting my basics right in DSA and CP. Hoping to connect with many great people here who are doing amazing in their fields"
X Link 2025-10-14T04:19Z [----] followers, [---] engagements
"Okay so this happened as well 😮 It now took me only [--] hours for the next [--] 😐 what is this growth Feels like I am speed running X https://t.co/1o3hHxUgEX It now took me only [--] hours for the next [--] 😐 what is this growth Feels like I am speed running X https://t.co/1o3hHxUgEX"
X Link 2025-09-20T03:27Z [----] followers, [---] engagements
"1 October [----] I learnt Git and GitHub from scratch. We had a [--] hour meeting today to discuss about our SIH progress and upcoming hackathons. Today was a bit productive"
X Link 2025-10-01T17:39Z [----] followers, [---] engagements
"23 September [----] DL: Filters and Convolutions in 1D image data MLOps: Introduction to MLOps need for MLOps Statistics: Describing datasets different plotting techniques MLOps: ML project life cycle Software development life cycle I will be consistent with my work 🤞🏻"
X Link 2025-09-23T19:51Z [----] followers, [---] engagements
"31 October [----] I couldn't study today. Worked on my project again. [--] hours almost. I finally setup CI/CD for my project using GitHub Actions. Initial tests worked. So I prepared more comprehensive unit tests. I am waiting for all of this to be finished so that I can go to the last stages and refine parts of the project. https://x.com/initlayers/status/1984355416156094474/photo/1 https://x.com/initlayers/status/1984355416156094474/photo/1"
X Link 2025-10-31T20:22Z [----] followers, [---] engagements
"24 September [----] I studied 2D filters in CNNs today. How the filters are applied to the depth of the images instead of individual matrices. I tried to find an isolated pixel using custom filters. It was a fun ride"
X Link 2025-09-24T22:22Z [----] followers, [---] engagements
"14 October [----] Revised CNNs: Why CNNs over ANNs Filters Hands-on 1D & 2D data convolutions multiple filters Studied Filtering Layer Nomenclature Revised Git commands ML lifecycle I don't think it's a great day or something. But everyday counts I will be patient"
X Link 2025-10-14T19:12Z [----] followers, [---] engagements
"30 October [----] Couldn't study LLMs or MLOps I worked on the project for [--] hours. Performed ablation studies trying to document everything. I am trying to set-up CI/CD with GitHub Actions. I dunno if it'll be completed tonight. Today was rusty though I couldn't study anything"
X Link 2025-10-30T21:26Z [----] followers, [---] engagements
"I am stuck with these tests. Chances of finishing the project tonight are slim 🥺"
X Link 2025-10-31T18:06Z [----] followers, [---] engagements
"In the Chinese Room thought experiment the person inside knows only English and has an instruction manual. People outside pass him Chinese question through a slot. The person follows the manual and produces perfect Chinese responses. From the outside it seems like he understands Chinese but in reality he doesn't. The intelligence isn't in the person alone but perhaps in the system (the person + the instruction manual) Now here's the question: if we were to simulate every neural network and pathway of a human brain would that count as true intelligence"
X Link 2025-10-24T02:30Z [----] followers, [---] engagements
"24 October [----] Revised LLM basics stages of building LLMs OOPs - mostly around classes and objects Project: Finally resumed. I built core modules to test project feasibility. I then downloaded Ollama and tested TinyLlama locally. Obviously it's not very great but with my GPU and RAM I am satisfied with anything I can possibly get. The next couple of days will be intense. I will try to code stuff from scratch"
X Link 2025-10-24T21:29Z [----] followers, [---] engagements
"21 October [----] Studied stages of building LLMs again but in more detail Finally started the project-1 implementation just set-up the environment and dependencies after careful consideration I have decided to document my choices and learnings throughout the project"
X Link 2025-10-21T21:13Z [----] followers, [---] engagements
"15 October [----] Selected a PS built a prototype and recorded a demo video for SIH (took around [--] hours) Revised filtering layer nomenclature"
X Link 2025-10-15T19:58Z [----] followers, [---] engagements
"19 October [----] Studied pre-training fine-tuning types of fine-tuning and stages of building LLMs Tried to understand Transformers Architecture tokenization self attention vector embeddings BERT vs GPT Project: confirmed the stack we will use free tools/models"
X Link 2025-10-19T19:57Z [----] followers, [---] engagements
"It now took me only [--] hours for the next [--] 😐 what is this growth Feels like I am speed running X It took me [--] days to reach my first [--] followers and just [--] days to reach the next [--]. That's [--] days on this app so far. Greatful for the journey and all the learnings thank you 🙏🏻 https://t.co/5eSDF9w7VA It took me [--] days to reach my first [--] followers and just [--] days to reach the next [--]. That's [--] days on this app so far. Greatful for the journey and all the learnings thank you 🙏🏻 https://t.co/5eSDF9w7VA"
X Link 2025-09-18T03:20Z [----] followers, [---] engagements
"Small account happiness 🙂"
X Link 2025-09-18T02:59Z [----] followers, [---] engagements
"I'm still working on this project. It's been around [--] hours already and things aren't looking great. The code has finally passed all [--] unit integration & performance tests but the SQLite backend in Chroma is set to read-only which is causing issues during CI. I'll fix this it should be an easy one. I'll do everything I can to finish the project tonight"
X Link 2025-11-02T18:23Z [----] followers, [---] engagements
"It took me [--] hours but I am finally done with the back-end it's working great. I'll start the front-end as well I have to complete this tonight at any cost"
X Link 2025-09-29T20:06Z [----] followers, [---] engagements
"Tomorrow is the deadline I have to build an MVP ASAP else NGMI. Our presentation will include pictures of the UI so I am the bottleneck now this night will be legendary 💀"
X Link 2025-09-29T13:16Z [----] followers, [---] engagements
"I think I'll break the record for spending the most midnights on SIH. Hope it's all worth it 😐"
X Link 2025-10-06T23:08Z [----] followers, [---] engagements
"20 Sept [----] I designed a basic neural network for classification using TensorFlow it felt a lot different to Pytorch. I studied basics of vector algebra and span of two vectors spent most of the time shortlisting problem statements for SIH"
X Link 2025-09-20T18:06Z [----] followers, [---] engagements
"Guys I found something cool: The automatic recognition of machine or hand-printed characters is what we refer to by Optical Character Recognition (OCR) The recognition of human printed or cursive handwriting as the user is writing it's called "On-line Handwriting Recognition""
X Link 2025-09-03T11:37Z [----] followers, [---] engagements
"16 October [----] Studied max pooling layer and CNN architecture. Revised filtering layer nomenclature and working"
X Link 2025-10-16T18:54Z [----] followers, [---] engagements
"Good morning 🌞 The sun's aura is too high that I couldn't see him eye to eye even from [---] million kilometers"
X Link 2025-09-20T01:29Z [----] followers, [---] engagements
"Our second SIH submission (made on 15th October) wasn't successful. We worked hard for it but couldn't quite pull it off. Now we're waiting for the batch [--] results. Let's see if our first submission can work some magic"
X Link 2025-11-09T04:48Z [----] followers, [---] engagements
"Project Update: I finally made it. I completed the CI/CD pipeline and managed to not break anything. The project is now officially finished. I'll wrap up the remaining documentation and front-end work after my mid-terms. The last 2-3 days have been a hell of a ride. I learnt a lot - built a comprehensive test suite integrated LLM inference around 2K+ lines of code and spent countless hours debugging. It's been an amazing journey"
X Link 2025-11-03T23:17Z [----] followers, [---] engagements
"Instagram and TikTok are in a league of their own when it comes to algorithms. YouTube Spotify and X follow behind but none match how fast Instagram learns your mood. It's scary how good it's become at keeping you watching. Even though it's affecting attention spans and careers it remains a masterpiece both algorithmically and psychologically"
X Link 2025-10-24T14:48Z [----] followers, [---] engagements
"Brainstorming hackathon ideas for two challenges: one from IITB and the other from IITR. Both are kind of focussing on environmental issues. Let's see what would be finalized"
X Link 2025-09-28T14:04Z [----] followers, [---] engagements
Limited data mode. Full metrics available with subscription: lunarcrush.com/pricing