# ![@TusharMishra23 Avatar](https://lunarcrush.com/gi/w:26/cr:twitter::1918285228403253249.png) @TusharMishra23 ƬⲘ

ƬⲘ posts on X about chat, book, model, finish the most. They currently have [-----] followers and [---] posts still getting attention that total [-----] engagements in the last [--] hours.

### Engagements: [-----] [#](/creator/twitter::1918285228403253249/interactions)
![Engagements Line Chart](https://lunarcrush.com/gi/w:600/cr:twitter::1918285228403253249/c:line/m:interactions.svg)

- [--] Week [------] +12,519%
- [--] Month [------] +538%
- [--] Months [---------] +24,603%

### Mentions: [--] [#](/creator/twitter::1918285228403253249/posts_active)
![Mentions Line Chart](https://lunarcrush.com/gi/w:600/cr:twitter::1918285228403253249/c:line/m:posts_active.svg)

- [--] Months [---] +181%

### Followers: [-----] [#](/creator/twitter::1918285228403253249/followers)
![Followers Line Chart](https://lunarcrush.com/gi/w:600/cr:twitter::1918285228403253249/c:line/m:followers.svg)

- [--] Week [-----] no change
- [--] Month [-----] +2.50%
- [--] Months [-----] +1,469%

### CreatorRank: [-------] [#](/creator/twitter::1918285228403253249/influencer_rank)
![CreatorRank Line Chart](https://lunarcrush.com/gi/w:600/cr:twitter::1918285228403253249/c:line/m:influencer_rank.svg)

### Social Influence

**Social category influence**
[technology brands](/list/technology-brands)  [finance](/list/finance)  [social networks](/list/social-networks)  [countries](/list/countries)  [stocks](/list/stocks)  [celebrities](/list/celebrities)  [automotive brands](/list/automotive-brands) 

**Social topic influence**
[chat](/topic/chat), [book](/topic/book), [model](/topic/model), [finish](/topic/finish), [abt](/topic/abt), [paper](/topic/paper), [neural](/topic/neural), [link](/topic/link), [code](/topic/code), [books](/topic/books)

**Top accounts mentioned or mentioned by**
[@sharpeyewnl](/creator/undefined) [@veritusai](/creator/undefined) [@grok](/creator/undefined) [@neuralavb](/creator/undefined) [@memoizedk](/creator/undefined) [@xibyte](/creator/undefined) [@maidishahoon](/creator/undefined) [@pseudosid26](/creator/undefined) [@devloperhs](/creator/undefined) [@gammaapp](/creator/undefined) [@kimimoonshot](/creator/undefined) [@cneuralnetwork](/creator/undefined) [@ninzo121](/creator/undefined) [@theglobalminima](/creator/undefined) [@anonymanasensei](/creator/undefined) [@vasumanmoza](/creator/undefined) [@sujaltaneja](/creator/undefined) [@ramaniragh763](/creator/undefined) [@imbearguy](/creator/undefined) [@tanayvasishtha](/creator/undefined)

**Top assets mentioned**
[WORK (WORK)](/topic/work-work)
### Top Social Posts
Top posts by engagements in the last [--] hours

"@0xmitsurii @harshbhaiyaa Yeah i personally learned MERN React Native from them. For DSA i prefer learning myself but can bet on them"  
[X Link](https://x.com/TusharMishra23/status/1933932880633205124)  2025-06-14T17:01Z [--] followers, [--] engagements


"Implemented some great features par Vercel pe deploy krne pe issue aa raha tha toh plan is to move on to MongoDB after that this project will be completed 🤘"  
[X Link](https://x.com/TusharMishra23/status/1935007016269046118)  2025-06-17T16:09Z [--] followers, [--] engagements


"Fck it's finally done. Thank u MongoDB alot and fck rls"  
[X Link](https://x.com/TusharMishra23/status/1935478928988344578)  2025-06-18T23:25Z [--] followers, [--] engagements


"agar Support vector machines face data pe itna sahi kaam karti h toh digits mai toh crazzy bhai (try krne ja raha)"  
[X Link](https://x.com/TusharMishra23/status/1937013543943180581)  2025-06-23T05:03Z [--] followers, [--] engagements


"@KartikeyStack that's the only thing typescript hates bro😄"  
[X Link](https://x.com/TusharMishra23/status/1937588842590273724)  2025-06-24T19:09Z [--] followers, [--] engagements


"React native tu toh gya bete😤"  
[X Link](https://x.com/TusharMishra23/status/1937717860014706826)  2025-06-25T03:41Z [--] followers, [--] engagements


"530. Minimum Absolute Difference in BST Hint: isme kya hint lega bhai😅"  
[X Link](https://x.com/TusharMishra23/status/1938410436137042031)  2025-06-27T01:33Z [--] followers, [--] engagements


"Looks like i've cleared the basics of BST [---]. Find Largest Value in Each Tree Row Hint: Do BFS then just append the max val in ans array"  
[X Link](https://x.com/TusharMishra23/status/1939147970689277989)  2025-06-29T02:24Z [--] followers, [--] engagements


"For anyone out there who wants to start Machine Learning this book is the pure gold contains numpy pandas matplotlib scikit-learn in good depth Link: https://archive.org/details/python-data-science-handbook.pdf/mode/1upview=theater https://archive.org/details/python-data-science-handbook.pdf/mode/1upview=theater"  
[X Link](https://x.com/TusharMishra23/status/1939187694615298408)  2025-06-29T05:02Z [--] followers, [--] engagements


"how to get into ml research like is there a bar like when you know Neural net LLM or some math derivation. Anyone and also how to do one 🙂"  
[X Link](https://x.com/TusharMishra23/status/1940356734121656444)  2025-07-02T10:27Z [--] followers, [--] engagements


"@cneuralnetwork how to get into ml research like is there a bar like when you know Neural net LLM or some math derivation. and also how to get or do one 🙂"  
[X Link](https://x.com/TusharMishra23/status/1940386408721735747)  2025-07-02T12:25Z [--] followers, [--] engagements


"When u realize Multi Layer Perceptron in reality is just Mathematics 🧠 At its core an MLP is a feedforward artificial neural network that uses a series of mathematical operations to transform input data into desired outputs"  
[X Link](https://x.com/tm23twt/status/1941055301085811186)  2025-07-04T08:43Z [--] followers, [--] engagements


"This is Multi layer perceptron when there is deep stack of hidden layer it is called deep neural network"  
[X Link](https://x.com/tm23twt/status/1941077800821444769)  2025-07-04T10:13Z [--] followers, [--] engagements


"In a neural network: = all weights and biases (can be millions of parameters). Backpropagation computes L efficiently for all of them. You use gradient descent (or a variant) to update them all at once"  
[X Link](https://x.com/tm23twt/status/1941096724812169231)  2025-07-04T11:28Z [--] followers, [--] engagements


"@amritwt INSANITY: Popular mf Amrit changed his pfp after gaining 20k + follower from that rainbow glasses man Is this going TOO far"  
[X Link](https://x.com/tm23twt/status/1941477903746859354)  2025-07-05T12:42Z [--] followers, [---] engagements


"@wordsofteekay will check it out currently i've just finished basics of neural nets like backpropagation😀"  
[X Link](https://x.com/tm23twt/status/1941766098971668521)  2025-07-06T07:48Z [--] followers, [--] engagements


"Ever heard of Universal Approximation Theorem "A large enough Deep Neural Net with nonlinear activations can theoretically approximate any continuous function." ✔ The nonlinearity is essential. ✔ Thats the power of deep learning"  
[X Link](https://x.com/tm23twt/status/1941768164544741647)  2025-07-06T07:56Z [--] followers, [--] engagements


"Wrote my first Multi Layer Perceptron Regressor in Scikit learn (next time in Keras) with understanding of how neural nets works. This Shit is really cool"  
[X Link](https://x.com/tm23twt/status/1941810831022542901)  2025-07-06T10:45Z [--] followers, [--] engagements


"@_rajtwt This is it 50LPA on the way bhai"  
[X Link](https://x.com/tm23twt/status/1942113592045941027)  2025-07-07T06:48Z [--] followers, [--] engagements


"Do ml or web-dev (any fcking one) Do leetcode since codeforces will overwhelm u Do internship & hackathons Do Build in public (everything) Do AWS Azure or some shit Can't really say about system design Grab anything u get fck numbers This is my 3-year college roadmap to secure a [--] LPA package 🗿😤 https://t.co/dsUwxFBMRh This is my 3-year college roadmap to secure a [--] LPA package 🗿😤 https://t.co/dsUwxFBMRh"  
[X Link](https://x.com/tm23twt/status/1942632586834260108)  2025-07-08T17:11Z [--] followers, [---] engagements


"ADVICE❗ if using TypeScript never use any type fix it on the spot. or u will get fcked🤡"  
[X Link](https://x.com/tm23twt/status/1942904562266018056)  2025-07-09T11:11Z [--] followers, [--] engagements


"If your place doesn't look like this you're ngmi"  
[X Link](https://x.com/tm23twt/status/1943405364730552764)  2025-07-10T20:22Z [--] followers, [--] engagements


"Typescript WTF IS THIS WARNING"  
[X Link](https://x.com/tm23twt/status/1943744168322576695)  2025-07-11T18:48Z [--] followers, [--] engagements


"is svelte really good planning to switch from next js to svelte any advice would be helpful"  
[X Link](https://x.com/tm23twt/status/1945672551134548390)  2025-07-17T02:30Z [--] followers, [--] engagements


"started learning react native just for exploring & freelancing our main shit is ml & that shit is cool af"  
[X Link](https://x.com/tm23twt/status/1946200490519511525)  2025-07-18T13:28Z [--] followers, [--] engagements


"f-ML💢 Day [--] Around 3-4 hrs (clg exams) ADAGrad RMSProp Adam optimizer got better intuition on diff btw various optimizers next is variants of Adam then optimizers will be done building a react native app"  
[X Link](https://x.com/tm23twt/status/1946272842104131735)  2025-07-18T18:16Z [--] followers, [--] engagements


"f-ML💢 Day [--] Around 4-5 hrs Adam optimizer variants - NAdam WAdam Adamax Learning rate scheduling - power exponential piecewise performance 1cycle next is avoiding overfitting through regularization learning react native"  
[X Link](https://x.com/tm23twt/status/1946655106197491768)  2025-07-19T19:35Z [----] followers, [--] engagements


"f-ML💢 Day [--] Around 4-5 hrs Adam optimizer variants - NAdam WAdam Adamax Learning rate scheduling - power exponential piecewise performance 1cycle next is avoiding overfitting through regularization learning react native"  
[X Link](https://x.com/tm23twt/status/1946655309331861666)  2025-07-19T19:36Z [----] followers, [---] engagements


"Term "state-of-the-art" (SOTA) in the context of neural networks refers to a model or technique that achieves the best performance on a specific standardized task at a particular point in time. Think of it as the current "world record holder" for a given problem"  
[X Link](https://x.com/tm23twt/status/1946886661893410904)  2025-07-20T10:55Z [--] followers, [--] engagements


"this is how much i learnt in [--] chapter - Glorot & He initialization ReLU Sigmoid LeakyReLU Swish ELU SELU GELU Mish Batch Normalization Gradient clipping - Momentum NAG ADAM RMSProp AdaGrad - l1 l2 regularization Dropout MC dropout max norm somemore shit"  
[X Link](https://x.com/tm23twt/status/1946962851102666781)  2025-07-20T15:58Z [--] followers, [--] engagements


"this is how much i learnt in [--] chapter - Glorot & He initialization ReLU Sigmoid LeakyReLU Swish ELU SELU GELU Mish Batch Normalization Gradient clipping - Momentum NAG ADAM RMSProp AdaGrad - l1 l2 regularization Dropout MC dropout max norm some more shit"  
[X Link](https://x.com/tm23twt/status/1946963113640902742)  2025-07-20T15:59Z [--] followers, [--] engagements


"f-ML💢 Day [--] Around 8-9 hrs Overfitting through regularization - MCDropout Dropout Max norm l1 & l2 regularization Chapter [--] : Custom model & training with Tensorflow next is tensors & operation some tensor shit learning react native updating portfolio website"  
[X Link](https://x.com/tm23twt/status/1947008585155916071)  2025-07-20T18:59Z [--] followers, [--] engagements


"f-ML💢 Day [--] Around 8-9 hrs Overfitting through regularization - MCDropout Dropout Max norm l1 & l2 regularization Chapter [--] : Custom model & training with Tensorflow next is tensors & operation some tensor shit learning react native updating portfolio website"  
[X Link](https://x.com/tm23twt/status/1947008664784797839)  2025-07-20T19:00Z [--] followers, [--] engagements


"f-ML💢 Day [--] Around 8-9 hrs Chapter [--] : Custom model & training with Tensorflow custom loss func metrics regularizers activation (basically customizing shit) learning react native updating portfolio website next is custom layers model gradients using autodiff"  
[X Link](https://x.com/tm23twt/status/1947380227812032819)  2025-07-21T19:36Z [--] followers, [--] engagements


"f-ML💢 Day [--] Around 8-9 hrs Chapter [--] : Custom model & training with Tensorflow custom loss func metrics regularizers activation (basically customizing shit) learning react native updating portfolio website next is custom layers model gradients using autodiff"  
[X Link](https://x.com/tm23twt/status/1947380316991328534)  2025-07-21T19:37Z [--] followers, [--] engagements


"f-ML💢 Day [--] Around 8-9 hrs Chapter [--] custom loss func metrics regularizers activation (basically customizing shit) custom metrics layers models gradients using autodiff learning react native -fck updating portfolio website next is custom training loops"  
[X Link](https://x.com/tm23twt/status/1947718296742306122)  2025-07-22T18:00Z [---] followers, [--] engagements


"f-ML💢 Day [--] Around 8-9 hrs Chapter [--] custom loss func metrics regularizers activation (basically customizing shit) custom metrics layers models gradients using autodiff learning react native -fck updating portfolio website next is custom training loops"  
[X Link](https://x.com/tm23twt/status/1947718357798859049)  2025-07-22T18:00Z [--] followers, [--] engagements


"@samireey @Divyanshii_1707 @samireey bro par abhi projects toh wahi cancer detection house price detection basic hi banenge shouldn't we go deep into theory like CNN RNN NLP and then some projects. which approach u would suggest as a senior 🫡"  
[X Link](https://x.com/tm23twt/status/1948015591673237972)  2025-07-23T13:41Z [--] followers, [--] engagements


"f-ML💢 Day [--] Around 8-9 hrs Chapter [--] custom training loops AutoGraph & Tracing tensorflow data API (cool shit) exploring CERN HSF for open source next is TFRecord format preprocessing layers pending task : resume portfolio update website + react native"  
[X Link](https://x.com/tm23twt/status/1948081974096322633)  2025-07-23T18:05Z [--] followers, [--] engagements


"Insane U can test ur routes without postman or some shit Just go to your browser console and fetch the request something like this 👇"  
[X Link](https://x.com/tm23twt/status/1948251419242369523)  2025-07-24T05:18Z [--] followers, [--] engagements


"I actually started liking PostgreSQL Neon Prisma & started hating MongoDB reason : man it's fast like 500ms response time it's good ngl"  
[X Link](https://x.com/tm23twt/status/1948407397833388441)  2025-07-24T15:38Z [--] followers, [--] engagements


"f-ML💢 Day [--] Around 8-9 hrs Chapter [--] AutoGraph & Tracing TFRecord format preprocessing layers tf protocols tensorflow data API (cool shit) working on web project next is Keras preprocessing layers pending task : resume portfolio update react native"  
[X Link](https://x.com/tm23twt/status/1948470601481683257)  2025-07-24T19:49Z [---] followers, [----] engagements


"f-ML💢 Day [--] Around 8-9 hrs Chapter [--] AutoGraph & Tracing TFRecord format preprocessing layers tf protocols tensorflow data API (cool shit) working on web project next is Keras preprocessing layers pending task : resume portfolio update react native"  
[X Link](https://x.com/tm23twt/status/1948470708633301019)  2025-07-24T19:49Z [--] followers, [--] engagements


"mfs : You can study ml without mathematics 🤟 Day [--] : 🤮"  
[X Link](https://x.com/tm23twt/status/1949739040447566298)  2025-07-28T07:49Z [--] followers, [---] engagements


"For anyone who knows next js & typescript planning to go for app dev this is really good advice also follow this man 👇 @tm23twt diving in react native is one of the best decisions someone can make imo. the best wall you can face is react and typescript but since you said you already know react and nextjs then theres no wall at all. i use expo to build react native apps think of it as the framework. it @tm23twt diving in react native is one of the best decisions someone can make imo. the best wall you can face is react and typescript but since you said you already know react and nextjs then"  
[X Link](https://x.com/tm23twt/status/1949812246697173467)  2025-07-28T12:40Z [---] followers, [---] engagements


"Building a Pizza Store App 🥳 To learn React native + Expo (integration with Supabase) Covered around [--] hrs of content Link : Not that smooth journey 🤢 Now going for ml ✌ https://www.youtube.com/watchv=rIYzLhkG9TA https://www.youtube.com/watchv=rIYzLhkG9TA"  
[X Link](https://x.com/tm23twt/status/1950537409269043700)  2025-07-30T12:42Z [---] followers, [---] engagements


"f-ML💢 Day [--] Around 8-9 hrs Chapter [--] : CNN (ended 🥳) Object Detection Fully Conv Network YOLO Object Tracking Semantic segmentation Super resolution Transpose Conv layer lil progress in react native 🤢 next up : RNN (Chapter 15) 🫦"  
[X Link](https://x.com/tm23twt/status/1951230981752443018)  2025-08-01T10:38Z [---] followers, [---] engagements


"August [----] Focus of the Month : - finish HOML book (500 pages btw) - land a internship (ml) - build a project (react-native) - sleep schedule (idgaf about this)"  
[X Link](https://x.com/tm23twt/status/1951359944521031723)  2025-08-01T19:10Z [---] followers, [----] engagements


"confession time 🤡 planning to go full time to ml recently i thought maybe react native & next js with ml will help (but it's boring man) now i'll aim to be too good in this ml fcking shit. heard some man saying : if what ur doing is not ur goal then fck it man"  
[X Link](https://x.com/tm23twt/status/1951380063959621980)  2025-08-01T20:30Z [---] followers, [---] engagements


"f-ML💢 Day [--] 🥳 Ch [--] NLP with RNN & attention Sentiment Analysis of movie reviews Masking Pretrained Embedding & Language model Encoder - Decoder network for Neural machine Translation Project : English to spanish translation next : Attention mechanism 🤢"  
[X Link](https://x.com/tm23twt/status/1953771969238765694)  2025-08-08T10:55Z [---] followers, [--] engagements


"f-ML💢 Day [--] 🥳 Ch [--] NLP with RNN & attention Sentiment Analysis of movie reviews Masking Pretrained Embedding & Language model Encoder - Decoder network for Neural machine Translation Project : English to spanish translation next : Attention mechanism 🤢"  
[X Link](https://x.com/tm23twt/status/1953772045033959921)  2025-08-08T10:55Z [---] followers, [---] engagements


"life is good when you're reading this"  
[X Link](https://x.com/tm23twt/status/1956676657147711986)  2025-08-16T11:17Z [---] followers, [---] engagements


"f-ML💢 Day [--] 🥳 Ch [--] (finished ch 16) Avalanche of Transformer models BERT GPT DistilBERT Vision Transformers Hugging face 🤬 Fine tuning pre training self distillation need to read some papers & implement them next : Autoencoders GANs and Diffusion Models"  
[X Link](https://x.com/tm23twt/status/1957138202004701304)  2025-08-17T17:51Z [---] followers, [---] engagements


"Honest review : This paper is shit wrapped with grammar all it talks about is stakeholder risk assessment testing. This gave me feel of UPSC answers where u have to just expand that shit. All it talks about is nothing no practical way to apply the solutions it talks about"  
[X Link](https://x.com/tm23twt/status/1957484331414274219)  2025-08-18T16:46Z [---] followers, [---] engagements


"@neural_avb level sabke niklenge 😅"  
[X Link](https://x.com/tm23twt/status/1957860127379202298)  2025-08-19T17:40Z [---] followers, [---] engagements


"f-ML💢 Day [--] 🥳 Convolutional autoencoder denoising & sparse autoencoders Kullback - leibler divergence (not gone into derivation) Variational autoencoders (this is quite confusing) 🤮 Generative adversarial network finished research paper 😘 next : Diffusion models"  
[X Link](https://x.com/tm23twt/status/1958123832071340281)  2025-08-20T11:07Z [---] followers, [---] engagements


"f-ML💢 Day [--] 🥳 Convolutional autoencoder denoising & sparse autoencoders Kullback - leibler divergence (not gone into derivation) Variational autoencoders (this is quite confusing) 🤮 Generative adversarial network finished research paper 😘 next : Diffusion models"  
[X Link](https://x.com/tm23twt/status/1958123887868211289)  2025-08-20T11:08Z [---] followers, [---] engagements


"f-ML💢 Day [--] 🥳 Generative adversarial network Nash equilibrium mode collapse experience relay mini batch discrimination Deep Convolutional GANs Progressive Growing of GANs Mini-batch standard deviation layer Pixelwise normalization layer next : Style GANs 🥲"  
[X Link](https://x.com/tm23twt/status/1958518597170450742)  2025-08-21T13:16Z [---] followers, [---] engagements


"@memoizedk bro im in 3rd year & i feel either go deep in ml with basics of next js and some web projects or don't explore ml at all cause it is too vast knowing things from top won't help & it will consume ur time as well :)"  
[X Link](https://x.com/tm23twt/status/1959600129872482680)  2025-08-24T12:54Z [---] followers, [---] engagements


"@memoizedk Exploring what u like is not wrong in fact that is what we should do. U will go through equations architectures and models. Like the HOML book introduces concepts of genai in Ch [--] as it needs some base. My point is to find ur domain asap but keep exploring :)"  
[X Link](https://x.com/tm23twt/status/1959602480863715486)  2025-08-24T13:03Z [---] followers, [--] engagements


"Neural nets are overrated this shit will humble you :)"  
[X Link](https://x.com/anyuser/status/1959739457994993950)  2025-08-24T22:07Z [----] followers, 17.9K engagements


"chat do you also post content in instagram idk what to do there is it just about reels or you can post shit like in X also share your ig profile so i can get some idea :)"  
[X Link](https://x.com/tm23twt/status/1960227375586926614)  2025-08-26T06:26Z [---] followers, [---] engagements


"@cryptopatrick I don't think ig is worth my time X is best 😁"  
[X Link](https://x.com/tm23twt/status/1960266590680957257)  2025-08-26T09:02Z [---] followers, [--] engagements


"writing a blog to answer most of ur question regarding this book will try to cover them all if u have any comment below will address them. be more specific & pinpointed & avoid question like how u stay consistent and all that personal shit :)"  
[X Link](https://x.com/anyuser/status/1960286184518680846)  2025-08-26T10:20Z [----] followers, [----] engagements


"ayo chat can y'all briefly introduce yourself like which domain ur into your learnings experience & whatever you want to share about yourself. it's hard to check portfolios profile of each one of you :)"  
[X Link](https://x.com/tm23twt/status/1960722523223462204)  2025-08-27T15:14Z [---] followers, [---] engagements


"@memoizedk college is eating hours 🥲"  
[X Link](https://x.com/tm23twt/status/1960793476997472269)  2025-08-27T19:56Z [---] followers, [---] engagements


"@Divyansh91565 fck phir bolte h kyu nhi ho rhi padhai 😭"  
[X Link](https://x.com/tm23twt/status/1960864781734142329)  2025-08-28T00:39Z [---] followers, [--] engagements


"chat LLM blog is published currently im reading HOLLM book & along that im making my notes so this will be in depth also i'll keep on updating this & there will be many next parts to this. you can read this along the book this will help you a lot. link below 👇"  
[X Link](https://x.com/tm23twt/status/1961063896422269363)  2025-08-28T13:50Z [---] followers, [----] engagements


"@sharpeye_wnl bro how the whole shit works like is it depends on impressions only or what"  
[X Link](https://x.com/tm23twt/status/1961649539800256631)  2025-08-30T04:37Z [---] followers, [--] engagements


"machine learning beginners blog around 20+ queries from resources roadmaps jobs internship math non cs beginners guide almost touched everything. took around 3hrs to cover each query thoroughly any doubt you've ever had about ml will be cleared in this one peace✌ link👇"  
[X Link](https://x.com/anyuser/status/1962056651630064059)  2025-08-31T07:35Z [----] followers, 19.2K engagements


"@stefnox Between AI debates about gradient descent vs reinforcement learning --- this looks familiar 🤔"  
[X Link](https://x.com/tm23twt/status/1962753887296987618)  2025-09-02T05:46Z [----] followers, [--] engagements


"Implemented Transformer Architecture in pytorch :) finally chat took around [--] days to get this shit done around 8-9 classes are there now i got a good understanding of this beautiful shit. Will finish the remaining small portion of paper today only then we move ahead :) link👇"  
[X Link](https://x.com/anyuser/status/1963149944405033420)  2025-09-03T07:59Z [----] followers, 10.1K engagements


"ml grind - 09/03 - finished research paper - Attention is all u need - added content in llm blog - implemented Transformer architecture novel components - understood kv caching will finish tomorrow - now we'll maybe code GPT or back to book we locked in chat :)"  
[X Link](https://x.com/anyuser/status/1963262498456367231)  2025-09-03T15:27Z [----] followers, 16.8K engagements


"onto third chapter baby i think this llm phase will go pretty smooth new topics will be introduced from next chapter ig also built a song recommendation system (quite easy). after this one will read paper or more theory of Byte pair encoding & other methods what abt u chat :)"  
[X Link](https://x.com/tm23twt/status/1963897040515301545)  2025-09-05T09:28Z [----] followers, [----] engagements


"fck it read around [--] pages today will deal with this sparse attention shit tomorrow also maybe we'll code byte pair encoding from scratch to get better understanding. will try to finish this book asap then we'll read some papers & code some more shit from scratch :)"  
[X Link](https://x.com/anyuser/status/1963973145221587430)  2025-09-05T14:31Z [----] followers, [----] engagements


"ml grind - 09/05 - read around 100+ pages in last [--] days - tokenization scheme & methods - introduced to various hugging face pre trained models - more clarity on contrastive training relevance scoring & kv caching - next up sparse attention bpe getting some hate nowadays :)"  
[X Link](https://x.com/tm23twt/status/1963989567339380893)  2025-09-05T15:36Z [----] followers, [----] engagements


"image source - pinterest :)"  
[X Link](https://x.com/tm23twt/status/1963989806523760776)  2025-09-05T15:37Z [----] followers, [---] engagements


"plans for the day chat - finish chapter [--] (around [--] pages left) - read paper of Byte Pair Encoding (popular one) - implement BPE from scratch - maybe RoPE & Flash attention as well what abt u chat :)"  
[X Link](https://x.com/tm23twt/status/1964158657467470094)  2025-09-06T02:48Z [----] followers, [----] engagements


"a man is born after going through Neural networks :)"  
[X Link](https://x.com/anyuser/status/1965007063760712161)  2025-09-08T10:59Z [----] followers, [----] engagements


"@neural_avb so cool sir excited for this one :)"  
[X Link](https://x.com/tm23twt/status/1965036067163455772)  2025-09-08T12:54Z [----] followers, [---] engagements


"ml grind - 09/09 - Implemented Rotary Absolute & Relative positional encodings - read around [--] pages - read paper of Byte Pair Encoding - got introduced to RLHF prompt eng instruction training & preference tuning why it looks so less anyways that's all :)"  
[X Link](https://x.com/anyuser/status/1965429136429908459)  2025-09-09T14:56Z [----] followers, [----] engagements


"yo chat onto chapter [--] only one fcking thing is coming to my mind after seeing prompt engineering yes ik you got it :)"  
[X Link](https://x.com/tm23twt/status/1965744349897273633)  2025-09-10T11:49Z [----] followers, [----] engagements


"the satisfaction after you have completed ur tasks is unreal. nowadays there is a loop of motivation for me i study then post you guys appreciate it so i try to keep up the efforts. but beside ml there are things i want to explore ig in 20s we'll detach a lil from ml :)"  
[X Link](https://x.com/tm23twt/status/1966441695244857472)  2025-09-12T10:00Z [----] followers, [----] engagements


"maths + llm 💀"  
[X Link](https://x.com/anyuser/status/1966486111791841763)  2025-09-12T12:56Z [----] followers, 48.1K engagements


"@sharpeye_wnl @X_Ibyte @lazy_Neuron Jaha dekhta hu bas bhidte rehte ho tum dono 🤣"  
[X Link](https://x.com/tm23twt/status/1966543439358763033)  2025-09-12T16:44Z [----] followers, [--] engagements


"ml goals for the day chat - now in end game final section of book that deals with Training & Fine tuning language models - will target to read around 70-80 pages - topics are SBERT creating & fine tuning embedding models currently at pg [---] chapter [--] what abt u chat :)"  
[X Link](https://x.com/anyuser/status/1966714109111747000)  2025-09-13T04:02Z [----] followers, [----] engagements


"@neuralpat @maidishahoon @grok mujhe toh gussa agya tha bhai 😡"  
[X Link](https://x.com/tm23twt/status/1966885109904404732)  2025-09-13T15:22Z [----] followers, [--] engagements


"@sharpeye_wnl bhai lekin impressions kite the aap toh 100+ wale ho 😈"  
[X Link](https://x.com/tm23twt/status/1966901434596946167)  2025-09-13T16:26Z [----] followers, [--] engagements


"yo chat onto chapter [--] two more chapters to go im lil confused as well cause there are bunch of things that i need to attend together like there are two books then papers alongside some orgs i wanna contribute. so things will not be linear from now on ig :("  
[X Link](https://x.com/tm23twt/status/1967138906148905091)  2025-09-14T08:10Z [----] followers, [----] engagements


"@ErenYea9er happy happy happy 😊"  
[X Link](https://x.com/tm23twt/status/1967457860633452690)  2025-09-15T05:18Z [----] followers, [--] engagements


"ml goals for the day chat - yesterday blog was done but only read around 5-6 pages - now we'll finish chapter [--] today - after this book maybe i should build [--] projects lil confused btw books & projects blog did rly good chat thanks to y'all what are you guys doing today :)"  
[X Link](https://x.com/anyuser/status/1967473096828555646)  2025-09-15T06:18Z [----] followers, [----] engagements


"@Aaron_Warner__ for dl go for homl this book uses pretrained llm & build things like rag on top of it will write a blog on this book soon :)"  
[X Link](https://x.com/tm23twt/status/1967815029920764025)  2025-09-16T04:57Z [----] followers, [--] engagements


"yo chat we done with chapter [--] though the named entity recognition needs more clarity my approach is to move fast then revisit things rather than going too slow. after this book will share the thoughts on it :)"  
[X Link](https://x.com/tm23twt/status/1967856042559275266)  2025-09-16T07:40Z [----] followers, [---] engagements


"@raydotsh ded 🥲"  
[X Link](https://x.com/tm23twt/status/1967922119691604424)  2025-09-16T12:02Z [----] followers, [--] engagements


"my picks : [--] [--] [--] [--] [--] ur :"  
[X Link](https://x.com/anyuser/status/1968615047996772498)  2025-09-18T09:56Z [----] followers, 49.8K engagements


"finally chat the book is over. took around [--] days but it was not that hard tbh. got introduced to alot of new shit also its not that now i'll never re-read this obv i will but next time would be much more faster & pinpointed. That's how i approach books. now what's next"  
[X Link](https://x.com/anyuser/status/1968664193684459982)  2025-09-18T13:11Z [----] followers, [----] engagements


"@NimbuSeth maine tag nhi kiya kya ek baar comment kr dena koi kisi ke sath team up ho jyage shayad :)"  
[X Link](https://x.com/tm23twt/status/1968700289868759519)  2025-09-18T15:35Z [----] followers, [--] engagements


"ml goals for the day chat - need to wrap up a project in a week : spans yolo rag & class detection - explore dataset of exoplanets : lets see where this takes us - finish working with text data : if time allows that's for me what abt u chat"  
[X Link](https://x.com/tm23twt/status/1971096409236775253)  2025-09-25T06:16Z [----] followers, [----] engagements


"@herooffjustice wth Take care recover soon & bounce back king :)"  
[X Link](https://x.com/tm23twt/status/1971543113211781355)  2025-09-26T11:51Z [----] followers, [---] engagements


"nothing just building ml project at [--] : [--] am :)"  
[X Link](https://x.com/anyuser/status/1971713973071569092)  2025-09-26T23:10Z [----] followers, [----] engagements


"@ninzo121 letss gooo congo bhai :)"  
[X Link](https://x.com/tm23twt/status/1972567575096913936)  2025-09-29T07:42Z [----] followers, [--] engagements


"transformer in electrical eng 💀 transformer in machine learning 💀"  
[X Link](https://x.com/tm23twt/status/1972583509689835615)  2025-09-29T08:45Z [----] followers, [---] engagements


"im cooked 🥀 - have to deploy my app - have to use docker - have to add models to docker image"  
[X Link](https://x.com/tm23twt/status/1973055524989018377)  2025-09-30T16:01Z [----] followers, [----] engagements


"@TheGlobalMinima inexperience is the reason ig i dislike it btw bhai aapne mera project dekha kya like basic h hi but review dena app :)"  
[X Link](https://x.com/tm23twt/status/1973066870426075507)  2025-09-30T16:46Z [----] followers, [--] engagements


"im so done with docker i just built a docker image of size [--] fcking gb broo that too in a failed run yeah now my mood is fcked fr"  
[X Link](https://x.com/tm23twt/status/1973375189049094577)  2025-10-01T13:11Z [----] followers, [----] engagements


"@TheGlobalMinima Yeah bro this was my mistake now since my weights are pretrained I'm planning to use hugging face api to fetch. Thank you for your help though 🤍"  
[X Link](https://x.com/tm23twt/status/1973426692850065801)  2025-10-01T16:36Z [----] followers, [--] engagements


"@maidishahoon bounce back the next day :)"  
[X Link](https://x.com/tm23twt/status/1974497432655503706)  2025-10-04T15:30Z [----] followers, [--] engagements


"yea so im pulling an all nighter ig need to make some changes in the project before presentation ml model is good though. i just want to give my best shot to it we're still in a good position but why not target the best yk anyone up rn reply me when ru seeing this post :)"  
[X Link](https://x.com/tm23twt/status/1974581349341364296)  2025-10-04T21:04Z [----] followers, [----] engagements


"finally its over bro we could have made deployment in time but still we built a good shit. comparing to the fcks that happened im pretty much satisfied with the outcome. bro im on the screen for like whole day just 3hr of sleep with ded back. crazy experience fr :)"  
[X Link](https://x.com/tm23twt/status/1974915806183203258)  2025-10-05T19:13Z [----] followers, [----] engagements


"something always bugs me something reminds me that you're not great work work just work"  
[X Link](https://x.com/anyuser/status/1975055919223332874)  2025-10-06T04:30Z [----] followers, [----] engagements


"@TimmermansSam I think after sometime work doesn't teaches u much Is it right bro"  
[X Link](https://x.com/tm23twt/status/1975139495218729185)  2025-10-06T10:02Z [----] followers, [--] engagements


"this won't change for years :)"  
[X Link](https://x.com/tm23twt/status/1975945082026029198)  2025-10-08T15:23Z [----] followers, [---] engagements


"@OzyBehindScreen thank u bruh :)"  
[X Link](https://x.com/tm23twt/status/1976227035279413279)  2025-10-09T10:03Z [----] followers, [--] engagements


"ml goals for the day chat - implementing a GPT model from scratch to generate text - maths we have norm generated norms orthogonality & gram schmidt - not that productive days tbh also if time allows then we finish our blog that's abt me what ru learning building bruh :)"  
[X Link](https://x.com/tm23twt/status/1976554160095957476)  2025-10-10T07:43Z [----] followers, [---] engagements


"first step to become a cracked dev is to use linux i failed here :("  
[X Link](https://x.com/anyuser/status/1976648547090219158)  2025-10-10T13:58Z [----] followers, [----] engagements


"so the only thing left is training the model we'll use the fine web dataset and then our gpt_2_small model will be ready to generate text yayy. quite fascinating that small model has [---] million parameters & size 650mb. also im addicted to kaggle code looks so cool in this :)"  
[X Link](https://x.com/anyuser/status/1976979276055593387)  2025-10-11T11:52Z [----] followers, 22.2K engagements


"@cneuralnetwork reason i don't like this book that much is cause i alr covered llm in a good depth already so this feels like extra effort but yeah book is really good :)"  
[X Link](https://x.com/tm23twt/status/1977105725156737326)  2025-10-11T20:15Z [----] followers, [---] engagements


""the most certain way to succeed is always to try just one more time""  
[X Link](https://x.com/anyuser/status/1977230245221499359)  2025-10-12T04:30Z [----] followers, [----] engagements


"ml goals for the day chat - today we'll train our gpt model on fineweb which has 18T tokens demn - in maths we're still at norms inner product orthogonality cause they didn't made much sense - final review & editing of the blog what abt you chat hanging out or locked in :)"  
[X Link](https://x.com/tm23twt/status/1977277962966671769)  2025-10-12T07:39Z [----] followers, [----] engagements


"@ANITUBER6 nice broo great that u joined X in 1st yr i regret doing it soo late"  
[X Link](https://x.com/tm23twt/status/1977440187836969262)  2025-10-12T18:24Z [----] followers, [--] engagements


"@Anonymanasensei san francisco am i right or am i right"  
[X Link](https://x.com/tm23twt/status/1978530806852145162)  2025-10-15T18:37Z [----] followers, [--] engagements


"f-ML💢 Day [--] Around 8-9 hrs Chapter [--] : CNN 🥳 CNN architectures : GoogleNet VGGNet ResNet Xception SeNet (fcking exhausted 😵💫) Residual unit SE block Inception mod doing a react native beginner tutorial + project 😎 next up more other architectures & code"  
[X Link](https://x.com/tm23twt/status/1949918846975897785)  2025-07-28T19:44Z [----] followers, [---] engagements


"f-ML💢 Day [--] Around 8-9 hrs Chapter [--] : CNN 🥳 CNN architectures : GoogleNet VGGNet ResNet Xception SeNet (fcking exhausted 😵💫) Residual unit SE block Inception mod doing a react native beginner tutorial + project 😎 next up more other architectures & code"  
[X Link](https://x.com/tm23twt/status/1949918922016162040)  2025-07-28T19:44Z [----] followers, [---] engagements


"f-ML💢 Day [--] Around 8-9 hrs Chapter [--] : CNN 🥳 CNN architectures : GoogleNet VGGNet ResNet Xception SeNet DenseNet CSPNet EfficientNet 🤯 Implemented ResNet [--] CNN using Keras Exploring PreTrained Model Supabase with react native 🤢 next Transfer learning"  
[X Link](https://x.com/tm23twt/status/1950278612294537649)  2025-07-29T19:33Z [----] followers, [---] engagements


"f-ML💢 Day [--] Around 8-9 hrs Chapter [--] : CNN 🥳 CNN architectures : GoogleNet VGGNet ResNet Xception SeNet DenseNet CSPNet EfficientNet 🤯 Implemented ResNet [--] CNN using Keras Exploring PreTrained Model Supabase with react native 🤢 next Transfer learning"  
[X Link](https://x.com/tm23twt/status/1950278702647984188)  2025-07-29T19:34Z [----] followers, [---] engagements


"Implemented pretrained Xception Model CNN architecture with TensorFlow 🥳 performed Data augmentation (flip rotation contrast) used Functional API freezing weights 🥱 total [--------] (million) parameters 🤯 accuracy [----] here is gist 👇 https://gist.github.com/TM23-sanji/a96bea5c32723e19b2316350d0f2564c https://gist.github.com/TM23-sanji/a96bea5c32723e19b2316350d0f2564c"  
[X Link](https://x.com/tm23twt/status/1950610838823502018)  2025-07-30T17:33Z [----] followers, [---] engagements


"f-ML💢 Day [--] Around 8-9 hrs Chapter [--] : CNN 🥳 Implemented Pretrained Xception Model🤯(thanks to colab for handling [--] million params) Classification & Localization Objectness score IoU Progress in React native (1hr) next up is FCN Semantic Segmentation 🔥"  
[X Link](https://x.com/tm23twt/status/1950645628138385671)  2025-07-30T19:52Z [----] followers, [----] engagements


"f-ML💢 Day [--] Around 8-9 hrs Chapter [--] : CNN 🥳 Implemented Pretrained Xception Model🤯(thanks to colab for handling [--] million params) Classification & Localization Objectness score IoU Progress in React native (1hr) next up is FCN Semantic Segmentation 🔥"  
[X Link](https://x.com/tm23twt/status/1950645693242359990)  2025-07-30T19:52Z [----] followers, [---] engagements


"just take a look at this shit like wtf bro these mathematical notations makes even the easier shit hard. I just got the understanding of absolute relative & rotatory positional embd. but look at this paper chat. this is what i mean when i say ml is hard fck this shit :("  
[X Link](https://x.com/tm23twt/status/1964962744526119174)  2025-09-08T08:03Z [----] followers, [----] engagements


"+  :) fck it chat we coded rotary absolute & relative positional embd from scratch. for time being Byte pair & [--] papers are on hold will deal with them pretty soon. they're on the hit list chat. link 👇"  
[X Link](https://x.com/tm23twt/status/1965042811067310420)  2025-09-08T13:21Z [----] followers, [----] engagements


"onto chapter [--] dawg the time we spent on embd & transformers is helping now book - hollm :)"  
[X Link](https://x.com/tm23twt/status/1965308089172898076)  2025-09-09T06:55Z [----] followers, [----] engagements


"ml grind - 09/13 - read around [---] pages in [--] days - got introduced to alot of shit - vision transformers multimodality CLIP & BLIP - image captioning & multimodal chat based prompting - embd models contrastive learning SBERT now only [---] more pages to go :)"  
[X Link](https://x.com/tm23twt/status/1966891564137111590)  2025-09-13T15:47Z [----] followers, [----] engagements


"@VijayPr27567352 clg chill h bhai"  
[X Link](https://x.com/tm23twt/status/1969041459245375780)  2025-09-19T14:10Z [----] followers, [--] engagements


"yo chat so we got the embd using clip model without fine tuning for captions & image. now we going to store this in pinecone & will try to see if for input image or text we get the [--] best or similar images :)"  
[X Link](https://x.com/tm23twt/status/1971580380160602617)  2025-09-26T14:19Z [----] followers, [----] engagements


"ml grind - 09/26 - coded BPE from scratch - new project going on 10% completed img & context embd are stored in pinecone around 900k - read article on ml methods for exoplanets - currently too busy with a project & hackathon so books are at pause quite some shit going on :)"  
[X Link](https://x.com/tm23twt/status/1971624740747202831)  2025-09-26T17:15Z [----] followers, [----] engagements


"small change in plan for image captioning part we're moving with flan-t5-small rather than blip model. since our clip model has gen embd for both img & caption in same space so we just need to retrieve that & pass it to flan-t5. need to adjust backend now 🥀"  
[X Link](https://x.com/tm23twt/status/1972222576119447598)  2025-09-28T08:51Z [----] followers, [---] engagements


"ml grind - 09/30 - project 80% completed it has clip flan t5 yolov9m models & pinecone for embd - cooking [--] blogs : one is technical & other one is chill - also working on exoplanets dataset this is exciting ngl - for books we'll resume it from next week ahh need a break🥀"  
[X Link](https://x.com/tm23twt/status/1973000841662808135)  2025-09-30T12:23Z [----] followers, [----] engagements


"@vasumanmoza this will be a thread sir (not a premium user) 🧵 so i wanted to build a project that spans multiple domains like rag img segmentation class detection image captioning. now i'll explain the models i used & the challenges i faced & why i paused this project (almost done) demo"  
[X Link](https://x.com/tm23twt/status/1976351495412154591)  2025-10-09T18:18Z [----] followers, [---] engagements


"@vasumanmoza so the dataset we picked was coco dataset since the clip yolov9m flan-t5-small performed good on this we didn't go towards fine tuning. We used pinecone for storing the embd generated by clip for images & captions. now when user enters prompt we retrieve top images why paused"  
[X Link](https://x.com/tm23twt/status/1976352262005063856)  2025-10-09T18:21Z [----] followers, [---] engagements


"@sujaltaneja_ @grok explain my bro what is imaginary axis make it easy for him i want him to get that reference"  
[X Link](https://x.com/tm23twt/status/1977028278075769320)  2025-10-11T15:07Z [----] followers, [--] engagements


"@RamaniRagh763 @sujaltaneja_ bro i never tried pca on a dataset with real & complex num how would they behave like i know this is silly still"  
[X Link](https://x.com/tm23twt/status/1977047376150823127)  2025-10-11T16:23Z [----] followers, [--] engagements


"ml goals for the day chat - will finish with fine tuning for classification ch [--] - making notes in maths was a good decision no cap - will go through proof of cauchy schwarz then orthogonality - in maths we're going slow cause it can be too much confusing what abt u chat :)"  
[X Link](https://x.com/tm23twt/status/1978006351138652510)  2025-10-14T07:53Z [----] followers, [---] engagements


"how tf will you debug this"  
[X Link](https://x.com/tm23twt/status/1978706220916654159)  2025-10-16T06:15Z [----] followers, [---] engagements


"@X_Ibyte Unicorn startup 😁"  
[X Link](https://x.com/tm23twt/status/1978831224635290055)  2025-10-16T14:31Z [----] followers, [--] engagements


"Happy diwali everyone :) dw just give this a heart no need to reply back"  
[X Link](https://x.com/tm23twt/status/1980201262328209780)  2025-10-20T09:15Z [----] followers, [----] engagements


"@ninzo121 Happy diwali ninzo bhai 🪔"  
[X Link](https://x.com/tm23twt/status/1980357123482734713)  2025-10-20T19:35Z [----] followers, [--] engagements


"@MRK__786 Haan bro ye maine spiderman movie m dekha tha andrew garfield wali m"  
[X Link](https://x.com/tm23twt/status/1980580680284479643)  2025-10-21T10:23Z [----] followers, [--] engagements


"@maidishahoon @sharpeye_wnl Need to reflect back on the work and my direction. Btw belated happy diwali broo 🪔"  
[X Link](https://x.com/tm23twt/status/1980584148189475072)  2025-10-21T10:37Z [----] followers, [--] engagements


"@imbearguy It's something I got inspired by spiderman movie 😁"  
[X Link](https://x.com/tm23twt/status/1980712960684355816)  2025-10-21T19:09Z [----] followers, [--] engagements


"so we started this book read first chapter which was basically abt introduction & all. some things that are on the list are - training our gpt model (124M params) : currently planning to use only 250M tokens of fineweb dataset this will take around 2hrs. will add kv caching low rank adapters and test out diff things on this model itself to compare things. - pipeline for face detection : got inspired by spiderman movie in which he build an automatic door unlocker will use the main landmarks and will compare it with input landmarks if admin allows that then door will unlock. its not hard tbh"  
[X Link](https://x.com/anyuser/status/1982100093751775430)  2025-10-25T15:01Z [----] followers, [----] engagements


"4 am & just realized idk shit abt these topics - torchscript / onnx for deployment - torch.distributed ddp mixed precision training - torch.profiler nvprof register_backward_hook still alot to learn need to explore pytorch docs atp :("  
[X Link](https://x.com/tm23twt/status/1982947860149608449)  2025-10-27T23:09Z [----] followers, [----] engagements


"@nothiingf4 lfgggg congo himanshu bhai :)"  
[X Link](https://x.com/tm23twt/status/1983108384196047178)  2025-10-28T09:47Z [----] followers, [--] engagements


"ml goals for the day chat - working on gpt model : ik this is taking some time reason i feel is that our dataset has 10B tokens & we will just use 250M tokens which makes sense for a 124M params model. now our dataloader is working fine first thing we need to tie the weights to make it from [---] to [---] then we need to fix that cuda out of memory shit. after on a small sample we'll check for checkpoints optim gradScaler autoGrad & if everything is working fine. also need to create a validation dataset cause after 250M token we shouldn't stop rather move it into valid dataset. - read abt model"  
[X Link](https://x.com/tm23twt/status/1983118109192945748)  2025-10-28T10:26Z [----] followers, [----] engagements


"ml goals for the day chat :) - add flash attn to model : without this our speed was [----] tok/s training on 250M tokens will take [--] hours. Chinchilla-optimal for 124M params is 2.5B tokens with flash attn that will take [--] hours & 250M tokens will take 3hrs. so now we will add this flash shit to our model & train it. - introduction to markov process mrp mdp & policy : basically ch1 stuff there is a great video by Veritasium ig so maybe will refer it too. cause this is the fundamental & starting point of rl. - maths : why not it's been a long time idk how much we will read but atleast read"  
[X Link](https://x.com/tm23twt/status/1983483407087661304)  2025-10-29T10:37Z [----] followers, [----] engagements


"ml goals for the day chat :) - openai gym api : explore openai gym api & gymnasium cartpole session will target to finish ch2. then pure-python implementation of the random RL agent (no policy) - more abt profilers : token speed of sdpa is slow so we go deeper there mixed precision training & also after 4M tokens why it paused. - explore smol-training-playbook : this was released yesterday & it looks a good resource so maybe we will explore the first section training compass. for how tf long will medium not open man plz tell me if you got any quick fixes :("  
[X Link](https://x.com/tm23twt/status/1984160764442415389)  2025-10-31T07:29Z [----] followers, [----] engagements


"Medium is finally working chat :) now we will write more cool shit maybe some technical ones as well. you can checkout my super helpful ml blogs at https://tmwork.vercel.app/blog.html https://tmwork.vercel.app/blog.html"  
[X Link](https://x.com/tm23twt/status/1984257319090393178)  2025-10-31T13:53Z [----] followers, [----] engagements


"Pytorch version of the book 'Hands on machine learning' just came out. those reading the tensorflow & keras version do check it out :) link to original post 👇"  
[X Link](https://x.com/anyuser/status/1984598358527525031)  2025-11-01T12:28Z [----] followers, 66.3K engagements


"Noticed a pattern here that super cracked devs are either : [--]. terminally online super active on X reply to every comment flex their knowledge or [--]. opens this app post some super technical content reply to only useful comments & log out"  
[X Link](https://x.com/tm23twt/status/1985397126155665609)  2025-11-03T17:22Z [----] followers, [----] engagements


"Cowards shrink from challenges weaklings flee from them but warriors wink at them"  
[X Link](https://x.com/tm23twt/status/1985557615984984536)  2025-11-04T04:00Z [----] followers, [---] engagements


"@kalomaze i test ran my gan on atari images and gen loss shoots up to [--] 😁"  
[X Link](https://x.com/tm23twt/status/1985656027954753914)  2025-11-04T10:31Z [----] followers, [---] engagements


"ml goals for the day chat :) - we got a good project idea that we'll build on top of our gpt2 model will share more details on this. so we going to run this today till 200M yesterday idk why but model didn't checkpointed. After run we will save it & then i need to deploy it. for today we will train it and save the checkpoint locally. - explore the rl notations sometimes i get lost like what do even policy means what was the formula symbol. so we'll revise that once & also we'll explore KL divergence & cross entropy which is such a beautiful concept like i highly recommend you to watch yt"  
[X Link](https://x.com/anyuser/status/1986380994208014475)  2025-11-06T10:31Z [----] followers, 36K engagements


"🚀 Mixed Precision Training is one of those superhero upgrades that's completely redefined deep learning. It's how billion-parameter models fit on your GPU and train in a flashall while keeping that high-fidelity accuracy 👀 For a [--] Billion parameter model in pure fp32: Total: [----] GB per GPU Most models wont even fit on a single 24GB card 💡 The Simple Idea: Half-Precision Full Power Run most of your model's operations in half-precision (fp16 or bfloat16) but keep the most critical parts (like weight updates) in full-precision (fp32). The Result ⚡ [--] times faster training speed. 💾 [--] times"  
[X Link](https://x.com/tm23twt/status/1986404506285003219)  2025-11-06T12:05Z [----] followers, [----] engagements


"@TensorTwerker gpu support kaise check krna h docs se kaisa h omarchy bhai"  
[X Link](https://x.com/tm23twt/status/1986755959537250702)  2025-11-07T11:21Z [----] followers, [----] engagements


"@__toughware__ nice broo so you full time on linux or dual boot"  
[X Link](https://x.com/tm23twt/status/1986758851761463407)  2025-11-07T11:33Z [----] followers, [---] engagements


"ml goals for the day chat :) - finally our gpt2 model training is over got a loss of [----] and the outputs it gives seems good enough. so this shit is over but now we will deploy the model weights to hugging face spaces or gradio idk much abt them so will read a blog about this. also @Menace_thakur & @Pseudo_Sid26 recommended me some great ideas : so firstly will go through the paper of ShinkaEvolve which is basically about model offsprings & its a pretty recent one. @VeritusAi will assist us to get the insights of the paper. - going through a yt playlist for rl cause now we are making short"  
[X Link](https://x.com/tm23twt/status/1986776552596730319)  2025-11-07T12:43Z [----] followers, [----] engagements


"@0x6c62 @grok what are the rest of three"  
[X Link](https://x.com/tm23twt/status/1986789574119743496)  2025-11-07T13:35Z [----] followers, [---] engagements


"@mukulrai04 wait its a os never heard of it"  
[X Link](https://x.com/tm23twt/status/1986820841007624526)  2025-11-07T15:39Z [----] followers, [----] engagements


"@RamaniRagh763 not related to rl see bhai basically core idea is instead of using [----------------] use [---------] this will save the memory but mixed term refers that in some places switch back to the big one. this is a later & advanced topic u will understand it then bhai"  
[X Link](https://x.com/tm23twt/status/1986822550819836404)  2025-11-07T15:46Z [----] followers, [--] engagements


"4 am - ig its fcking 6hrs since we are switching to omarchy. me and my bois are up :)"  
[X Link](https://x.com/tm23twt/status/1986927586413564244)  2025-11-07T22:43Z [----] followers, [----] engagements


"@Walking_Snek nice bro"  
[X Link](https://x.com/tm23twt/status/1986934196787683577)  2025-11-07T23:10Z [----] followers, [--] engagements


"@BrandonTahedl @dhh @dhh installing omarchy day1 was a failure idk why but bootable usb drive is not working still we going to find a way :)"  
[X Link](https://x.com/tm23twt/status/1986956451882156281)  2025-11-08T00:38Z [----] followers, [---] engagements


"yo just woke up & realized that this month i got three setbacks already - interview fckup : this was all my mistake skill issue i would say & im working on my basics. - nasa space challenge fckup : tbvh i think we were deserving like building a hard thing should be more appreciated than building a easy slop ofc they have there own criteria so np this was a fun experience - linux fckup : idk why but our usb is not bootable like we used both rufus & etcher still it boots into win [--]. though this was so much fun maybe after 2-3 days will try this again. now we going to pull an all-nighter & tmrw"  
[X Link](https://x.com/tm23twt/status/1987188710266507489)  2025-11-08T16:01Z [----] followers, 15.4K engagements


"Fall in love with the process and the results will come. gm :)"  
[X Link](https://x.com/anyuser/status/1987329839175774583)  2025-11-09T01:22Z [----] followers, [----] engagements


"@neuroquark you too dawg :)"  
[X Link](https://x.com/tm23twt/status/1987420032029372754)  2025-11-09T07:20Z [----] followers, [--] engagements


"so we implemented llama & gpt model from scratch. the architecture is mostly same gpt was trained on fine-web edu dataset while llama isn't trained cause its a big model. i was tryna add kv caching in llama alongside RoPE and idk things got messed up so for now its without caching but we will add that pretty soon. somethings i want to try besides these implementations are : - langchain & lang graph - docker & kubernetes - fine tune some models - shift completely to cursor no more notebooks - and a project link of the repo is in desc :)"  
[X Link](https://x.com/anyuser/status/1988924897431797847)  2025-11-13T11:00Z [----] followers, 55.6K engagements


"ml goals for the day chat :) - we learning langchain & lang graph today mostly due to fomo. gonna follow a yt tutorial its around 7hrs so that would take a lot of time. idk shit abt this agentic ecosystem so if you have some exp on how to explore this shit plz share that with me. - then we'll try to finish the blog this will also take some time. anyways lets see how much we can do. btw what are you doing mate"  
[X Link](https://x.com/tm23twt/status/1989205508163809601)  2025-11-14T05:35Z [----] followers, [----] engagements


"ml goals for the day chat :) - today we gonna finish the langchain tutorial already half way through it. chains parser template retriever & now its going towards RAG & then onto lang graph. - come to think of it this is actually a good tutorial it covers : [--]. pydantic output parser [--]. structured output [--]. sequential chains [--]. runnable sequence [--]. vector store [--]. rag pipeline [--]. langgraph [--]. tool binding [--]. conditional routing [--]. document loaders i feel its a good entry point cause im aware of the theory part :)"  
[X Link](https://x.com/tm23twt/status/1989584871895805981)  2025-11-15T06:42Z [----] followers, [----] engagements


"Jeff Bezos is [--] connections away from me :)"  
[X Link](https://x.com/tm23twt/status/1989651583580475719)  2025-11-15T11:07Z [----] followers, [----] engagements


"ml goals for the day chat :) - finally we going to publish a new blog its been a month already. i need to increase the frequency but we write only when that urge is at peak. buckle up canon event is happening lol - we got a project idea nothing crazy after we finish with the tutorial today yea it is taking some time. we'll explore the pieces of that project less see how it goes will update abt it :)"  
[X Link](https://x.com/tm23twt/status/1989951642683183305)  2025-11-16T07:00Z [----] followers, [----] engagements


"@VeritusAi such a beautiful timelapse also im learning about langchain & langgraph :)"  
[X Link](https://x.com/tm23twt/status/1989990025422119314)  2025-11-16T09:32Z [----] followers, [---] engagements


"blog on - Learnings after coding GPT & Llama from scratch :) this contains almost everything that i learned through the process some very fundamentals yet important & subtle points that you should keep in mind. just a [--] min chill read link in replies✌"  
[X Link](https://x.com/tm23twt/status/1990017895494242430)  2025-11-16T11:23Z [----] followers, 43.1K engagements


"ml goals for the day chat :) - we done with the langchain & langgraph tutorial will share the link & review of the video. but overall it was really good. - for revisiting the basics we are considering deep learning by ian goodfellow though ik it is quite lengthy. - also we'll explore some project ideas to see which one should i pick & we need to structure things cause everything is just so messy"  
[X Link](https://x.com/tm23twt/status/1990318597353398551)  2025-11-17T07:18Z [----] followers, [----] engagements


"@fr_axvra its cursor"  
[X Link](https://x.com/tm23twt/status/1990701903056191527)  2025-11-18T08:41Z [----] followers, [--] engagements


"just finished a LangChain & LangGraph tutorial ( 7hrs) & here is my review : - theory is very less so you need to code a lot which surely helps - you don't need to know abt working of LLMs but basic knowledge abt RAG prompts embeddings will help - covers two basic projects one is on youtube transcript & other is on pdf RAG agent - learning curve is smooth uses Chroma FAISS PyDantic Gemini model - in some sections there is some background noise but it is totally manageable - in some places code doesn't work as i think langchain modules have been updated. like langchain_tools.something is now"  
[X Link](https://x.com/tm23twt/status/1990726989171286337)  2025-11-18T10:21Z [----] followers, 45.9K engagements


"ml goals for the day chat :) - gonna start by reading Deep learning by goodfellow just for revisiting the topics. - then we will start a project pretty basic : like pdf rag things we will feed it like all the book pdfs i have in ml & then maybe retrieve the chunks related to query. will post abt this ofc. - then some rl we have moved scripts acc to the env like frozen lake contains non slippery slippery cross entropy based & so the future rl algos will be under one env umbrella"  
[X Link](https://x.com/tm23twt/status/1991018853305090080)  2025-11-19T05:41Z [----] followers, 18K engagements


"wandb uv & cursor are [--] crazy upgrades what else should i add :)"  
[X Link](https://x.com/tm23twt/status/1991127750510297393)  2025-11-19T12:53Z [----] followers, [----] engagements


"Nowadays i reply to all the comments before going to bed ofc a few in between & i think this is working good for me. Also if i haven't replied to you then gimme some time i will or i might have missed it. One imp thing is that alot of my replies are somehow not reaching to the post it has happened so many times. So although my interaction is less but i interact with my homies :)"  
[X Link](https://x.com/tm23twt/status/1991208304689967478)  2025-11-19T18:13Z [----] followers, [----] engagements


"@KambojPushpit i have heard that langchain is less preferred nowadays is this true"  
[X Link](https://x.com/tm23twt/status/1991213991855788186)  2025-11-19T18:36Z [----] followers, [---] engagements


"@stewardzhuo_ sach bol rha h na"  
[X Link](https://x.com/tm23twt/status/1991215729774067806)  2025-11-19T18:43Z [----] followers, [--] engagements


"@devloper_hs i see ur alot in agentic systems & all where should i go next i have just finished langchain & langgraph tutorial & like ik nothing abt this what would u recommend"  
[X Link](https://x.com/tm23twt/status/1991218754735517806)  2025-11-19T18:55Z [----] followers, [--] engagements


"Docker always feels scary with terms like containers images mounts & volumes. You don't even know what command you ran & then there is a 3GB container running in the background. but it is Super imp so if you want to learn Docker commands & how it works. there is a sandbox env provided by play with docker with great tutorials. You can create multiple instances with a time limit of 2hrs highly recommended do check it out :)"  
[X Link](https://x.com/tm23twt/status/1991457697342853538)  2025-11-20T10:44Z [----] followers, [----] engagements


"ml goals for the day chat :) - we'll start with linear algebra ch2 of goodfellow it covers maths & theory both parts. - then we going to test semantic chunking on our books pdf to see how it performs. if this goes well then we'll just pick a embedding model & store vector chunks in Qdrant. - also both Q & V iteration algo are done there is just some confusion in code so we'll go through that"  
[X Link](https://x.com/tm23twt/status/1991739800886604109)  2025-11-21T05:25Z [----] followers, [----] engagements


"ml goals for the day chat :) - we'll start with theory of MOE & Switch Transformers going through the hugging face blog on mixture of experts. - in rl algo of Q & V iteration is clear but there is some confusion so will refer to hf blog & then ch [--] which is Deep Q network. - then we will read fastapi docs ofc this will take less time so yea that's it"  
[X Link](https://x.com/tm23twt/status/1992113478438633707)  2025-11-22T06:10Z [----] followers, 10.7K engagements


"HuggingFace blog on Mixture of experts model is top notch like it covers : brief history then sparsity load balancing GShard random routing expert capacity switch transformers router Z loss & also abt fine tuning of MOE & comparison on diff benchmarks. like it covers various research papers & the imp points from various sources if you want to read research papers on this topic like : - Outrageously Large Neural Networks - Switch Transformers or - Mixtral-8x7B-v0.1 then @VeritusAI is there for help :)"  
[X Link](https://x.com/tm23twt/status/1992165723838226432)  2025-11-22T09:38Z [----] followers, 59K engagements


"@Rishike89765470 @grok what is he trying to say"  
[X Link](https://x.com/tm23twt/status/1992295344378712246)  2025-11-22T18:13Z [----] followers, [--] engagements


"ml goals for the day chat :) - well start with docker tutorials till now read about volumes container images & mounts. will quickly revise them & then move ahead. - then Deep Q Network ch6 we gonna start although state & action value method took some time but got a good clarity on them. also now i think atari games will be used quite frequently as env. - then goodfellow ch2 on linear algebra some part is left so gonna cover that"  
[X Link](https://x.com/tm23twt/status/1992474158131773650)  2025-11-23T06:03Z [----] followers, [----] engagements


"cross entropy & KL divergence are very fundamental concept but quite hard to visualize. great video by artem kirsanov where he explains cross entropy from first principles. these kind of videos for some reason helps a lot like whether its of 3b1b or welch labs most of them are quite good"  
[X Link](https://x.com/tm23twt/status/1992531376776806891)  2025-11-23T09:51Z [----] followers, 93.4K engagements


"In case you don't know these legends yet : Andrej Karpathy: - Was Director of AI at Tesla focusing on computer vision and deep learning applications. - Known for his influential blog on Recurrent Neural Networks recommended by Ilya as a top work. - Provided early insight into the importance of attention mechanisms. Ilya Sutskever: - Machine learning scientist co-founder and chief scientist of OpenAI - Co-inventor of AlexNet a breakthrough deep convolutional neural network that revolutionized image recognition. - Co-developed Generative Adversarial Networks (GANs) enabling generation of"  
[X Link](https://x.com/tm23twt/status/1993014003195691233)  2025-11-24T17:49Z [----] followers, 26.6K engagements


"goals for the day chat :) - well start with linear algebra goodfellow eigen values & singular value decomposition are some main topics. will try to finish this - then gonna continue with DQN model on pong game this shit is lil confusing like the way we interpolate & transform then store so all that gets messy. - then we gonna explore some repos also a simple rag project maybe i should start. it wont take long though https://twitter.com/i/web/status/1993209545364283766 https://twitter.com/i/web/status/1993209545364283766"  
[X Link](https://x.com/tm23twt/status/1993209545364283766)  2025-11-25T06:46Z [----] followers, [----] engagements


"long time chat :) we workin on a web app that will tell us reproducibility & difficulty score of a research paper. now we gonna integrate the @VeritusAi search api that will fetch the top results for a user query. its feels good to work on web for some time like a break from monotonous ml routine & yea will deploy it soon too"  
[X Link](https://x.com/tm23twt/status/1996784754881130848)  2025-12-05T03:32Z [----] followers, [----] engagements

Limited data mode. Full metrics available with subscription: lunarcrush.com/pricing