@burkov BURKOV

BURKOV posts on X about ai, llm, in the, this is the most. They currently have [------] followers and [---] posts still getting attention that total [------] engagements in the last [--] hours.

Engagements: [------] #

[--] Week [---------] +46%
[--] Month [---------] +25%
[--] Months [----------] +389%
[--] Year [----------] +56%

Mentions: [--] #

[--] Week [---] +51%
[--] Month [---] +40%
[--] Months [---] +49%
[--] Year [-----] +29%

Followers: [------] #

[--] Week [------] +0.53%
[--] Month [------] +2.50%
[--] Months [------] +13%
[--] Year [------] +44%

CreatorRank: [-------] #

Social Influence

Social category influence technology brands 21.93% stocks 7.02% finance 6.14% countries 1.75% social networks 0.88% celebrities 0.88% luxury brands 0.88% fashion brands 0.88%

Social topic influence ai 14.04%, llm #305, in the 10.53%, this is 7.02%, model 6.14%, code 6.14%, $googl 5.26%, anthropic #878, pay 4.39%, vibe coding #198

Top accounts mentioned or mentioned by @grok @neosoireth @ssskryl @jagersbergknut @udaysy @castillacore @hridayakuhara @jadecole2112 @axelmasz @markvalorian @navneet_rabdiya @srsimon_bolivar @jonaszolszewski @thegreatbonnie @a5t3r0ld @founddo @rarply @maticbranko @filligerr @nicolaskoehl

Top assets mentioned Alphabet Inc Class A (GOOGL)

Top Social Posts

Top posts by engagements in the last [--] hours

"The machine isn't into humor"
X Link 2026-02-13T09:45Z 53.2K followers, 64.7K engagements

"How couples met (1930-2024)"
X Link 2024-10-08T23:21Z 53.2K followers, 7.9M engagements

"I resigned from my Anthropic subscription. Here is the four-page letter explaining why I did that because everyone must care"
X Link 2026-02-09T23:21Z 53.2K followers, 100.9K engagements

"Cancelled my Max subscription to Claude. Had it for [--] months. Kept a Pro subscription for now but Codex with the [---] model provides comparable coding skill and a better UX"
X Link 2026-02-11T03:52Z 53.2K followers, 95.7K engagements

"A new issue of my weekly AI newsletter is out. In this issue: HBR AI doesnt reduce workit intensifies it Google Towards a science of scaling agent systems: When and why agent systems work The post-Transformer era: State space models Mamba and what comes after attention Explained Visualized speculative decoding MIT Helping AI agents search to get the best results out of large language models Tool torchvista: Interactive Pytorch forward pass visualization in notebooks Model Qwen3-Coder-Next: Pushing small hybrid models on agentic coding ArXiv PaperBanana: Automating academic illustration for AI"
X Link 2026-02-12T19:26Z 53.2K followers, [----] engagements

"A math professor noticed his kitchen sink at home was leaking. He called a plumber. The plumber came the next day tightened a couple of nuts and the sink worked perfectly again. The professor was delighted. But when a minute later the plumber handed him the bill he was shocked. This is a third of my monthly salary Yeah I get it said the plumber. Why dont you come work for our company as a plumber Youll make three times more than you do as a professor. Just remember: when you apply say you only finished seventh grade. They dont like hiring educated people. So the professor got a job as a"
X Link 2026-02-13T03:35Z 53.2K followers, 778.4K engagements

"Before OpenClaw when I said that they are currently polluting the Web with AI slop I used "they" without attributing the blame to anyone in particular. In my head "they" were some shady corporations or people who know how to use LLMs to generate this content and then they have some platformslike content farmsto distribute it. Now when I say that the Web is currently being puked and shat into I know exactly who is doing this how they do it where they do it and how many of them there are. This is officially the beginning of the end of the Web as we know it and we know when and by whom this end"
X Link 2026-02-13T04:59Z 53.2K followers, [----] engagements

"A fundamental flaw with this research is that they went to a remote work website where customers place jobs that they haven't succeeded in automating with AI to test whether AI can do these jobs better than humans. Well no it cannot because it's a remote work website where customers place jobs that they haven't succeeded in automating with AI. Another less fundamental flaw is that they used humans to evaluate whether AI did a better job than a human in a non-blind setting so human raters would unavoidably be biased. Finally the definition of AI failure as the quality of its deliverables being"
X Link 2026-02-13T19:38Z 53.2K followers, [----] engagements

"A LaTeX table generator app: Describe a table or paste data in any text format. An LLM will generate LaTeX code and render it as a LaTeX source tabular a PDF and PNG. Build your mini-apps on my new platform without coding and make $$. http://llambada.com https://llambada.com/p/mz3XU3Jd/latex-table-generator http://llambada.com https://llambada.com/p/mz3XU3Jd/latex-table-generator"
X Link 2026-02-16T00:36Z 53.2K followers, [----] engagements

"When its a word pattern matcher its a word pattern matcher. You might think its intelligence but its a word pattern matcher. Its not a belittlement. Its just what it is"
X Link 2025-08-10T22:53Z 53.2K followers, 499K engagements

"I didn't want to comment on OpenClaw. Usually when there's so much noise in the media it's some ordinary stuff just hyped well. So I took time to learn how it works thanks to open source. I was right. OpenClaw is 2% of ordinary stuff and 98% of hype. To put it very shortly in case you were wondering there are two things in it: [--]. You can chat with an LLM via a text messenger. Not anything new. [--]. The LLM can use tools that run on your computer. Not anything new either. Most of the "magic" mentioned in the media is about its ability to use the browser. But it's not its ability. It's"
X Link 2026-02-08T08:19Z 53.2K followers, 194.2K engagements

"Google's understanding of the target user: 🔴 What Google thinks the user will use in Antigravity 🟢 What the user actually uses in Antigravity"
X Link 2026-02-12T10:01Z 53.2K followers, 31.3K engagements

"Human coders know they lost but they keep fighting with windmills because they aren't ready to accept it"
X Link 2026-02-12T18:37Z 53.2K followers, 15.7K engagements

"Ever needed some random word but only "banana" comes to mind An online random word generator: https://llambada.com/p/dyjZ89Kv/random-word-generator https://llambada.com/p/dyjZ89Kv/random-word-generator"
X Link 2026-02-12T20:34Z 53.2K followers, [----] engagements

"These benchmarks 😞 This is how you break them: [--]. Ask your most capable model to read a public task A and generate a similar task and then solve it. Given that the solution is verifiable you get a collection of task-solutions similar to public task A but not copying it. [--]. Then you do the same for public task B etc. [--]. Once you have this collection of synthetic tasks you finetune your most capable model to solve these tasks so it becomes better. Better for this benchmark and nothing else. Rinse. Repeat. The new Gemini Deep Think is achieving some truly incredible numbers on ARC-AGI-2. We"
X Link 2026-02-12T20:39Z 53.2K followers, 24.8K engagements

"Previously Google Search dominated in: [--]. Index freshness [--]. Speed of search [--]. Accuracy of search The search accuracy here is the chance of putting the right result among the top [--] search results. LLMs made the second and third advantage irrelevant. As long as the right result is in the top [--] answering the user's question will be quick enough. And the competitors are already doing that. Competing only on index freshness isn't a viable strategy because it's not a huge moat and the competitors are quickly catching up. So not only is Google's index getting increasingly polluted by the AI slop"
X Link 2026-02-13T01:42Z 53.2K followers, [----] engagements

"Vanilla Transformer-based LLMs are slow at generating text because at each step they need to load from memory a set of cached values (called "keys" and "values") that help the model decide what to pay attention to from earlier in the conversation. A previous technique called multi-query attention (MQA) shrinks this cache by sharing a single set of keys and values across all the model's parallel attention computations (called "heads") which speeds things up a lot but hurts output quality. This landmark paper published by Google researchers introduces grouped-query attention (GQA) which splits"
X Link 2026-02-13T08:01Z 53.2K followers, [----] engagements

"So far AI can only reliably do the job of an average CEO: [--]. You participate in transcribed Teams meetings let everyone speak then ask the AI to generate actionable items and send them to the participants. [--]. You ask the AI to read news for your industry vertical compare it to your strategic plan and tell you what you should change in the plan. [--]. You feed the AI your company's quarterly metrics and last board deck and it generates an updated board presentation with the three "strategic priorities" that sound new but are actually the same ones from last quarter reworded. [--]. From time to time"
X Link 2026-02-14T00:55Z 53.2K followers, 22.6K engagements

"The challenge of defining what we mean by achieving true intelligence in computers (AGI/ASI whatever) is that to define it simply and not as a book of exceptions one needs to admit first that most people don't have it. If you honestly say that your definition of true intelligence only applies to a relatively small fraction of Homo sapiens then the definition of true intelligence becomes very simple: True intelligence when present allows the bearer of it to solve novel problems by applying novel methods. By this definition LLMs aren't intelligent but some people are. So when we are saying"
X Link 2026-02-15T05:11Z 53.2K followers, [----] engagements

"In the Claude desktop app when you are coding the UI allows you to submit messages during the coding agent's thinking/working stage. So if you see it making a wrong assumption during the thinking you can submit additional details without interrupting the process. That's one major feature that I've been missing since I started vibe coding two years ago. https://twitter.com/i/web/status/2023184595391312063 https://twitter.com/i/web/status/2023184595391312063"
X Link 2026-02-15T23:56Z 53.2K followers, 11.3K engagements

"This is how you build an app you need with Llambada. Without coding or even thinking about any code or any boring/scary things like backend frontend database or God forbid deployment🤯"
X Link 2026-02-09T00:12Z 53.2K followers, 15.3K engagements

"This concerns not only some random person working for a random startup that will most likely be acquired by Oracle or Amazon for peanuts but any person whose work affects the future of the nation. For example no high-level politician must be child-free or childless. Because you cannot be trusted to care about the nation's long-term survival and prosperity if you don't have anyone who you care for to inherit the result of your work. Amanda Askell doesnt have children and I believe a person who teaches AI a sense of right and wrong should be a person who truly loves humanity. Obviously you cant"
X Link 2026-02-16T01:59Z 53.2K followers, [----] engagements

"Astrophysicist and Nobel laureate Subrahmanyan Chandrasekhar used to travel [--] kilometers every week from the Yerkes Observatory to the University of Chicago where he taught a course attended by only two students. When asked why he spent his time this way the professor replied that they were very good students. In [----] Lee Tsung-Dao and Yang Chen-Ning were awarded the Nobel Prize in Physics. The course taught by Subrahmanyan Chandrasekhar became the only course in history where all its attendees received a Nobel Prize"
X Link 2024-05-01T15:43Z 53.1K followers, 1.3M engagements

"I said two years ago it would be Hadoop [---] and it is. When Hadoop and Big Data were at the peak of hype managers asked system administrators to connect old unused laptops that were gathering dust in the basement into a network install Hadoop on them and start "crunching data before competitors did." I didn't invent this; I witnessed it myself. No wonder 95% of Hadoop-based proofs of concept failed. With LLMs it's the same dumb managers just [--] years older asking secretaries to automate business processes using Copilot. No wonder 95% of these proofs of concept failed. Companies: fire dumb"
X Link 2025-08-20T15:21Z 53.1K followers, 39.7K engagements

"It took decades for people to realize that it was a bad idea to share their physical lives with Zuck. How much time will it take for them to realize that it's an even worse idea to share their online lives with altman"
X Link 2025-10-21T23:56Z 53.1K followers, 10.3K engagements

"Ilya is puzzled why LLMs are crushing benchmarks but the business outcome is next to nothing 😁 I mean why would anyone need a Safe Superintelligence if an unsafe one doesn't make money"
X Link 2025-12-14T05:44Z 53.1K followers, 455.1K engagements

"It must be illegal to sell access to one model but serve a different one. Since yesterday Claude has been making me angry and it's the first time I've felt angry since I started exclusively using it for coding two months ago. I chose Opus [---] as the model for which I'm charged $100/month but I know I'm being served something similar to GPT-3.5. This is fraud and theft. I was just about to post that as well I'm sure Anthropic cheats and serves a weaker model when overloaded or needs GPUs for something urgent. The last two days I feel like I'm talking to a retard from 2024."
X Link 2026-01-12T06:13Z 53.1K followers, 150K engagements

"An app one can vibecode and put online in about [--] minutes. https://t.co/fJxV8cVORd https://t.co/71q5YPcMKD https://t.co/fJxV8cVORd https://t.co/71q5YPcMKD"
X Link 2026-01-31T04:09Z 53K followers, 57.2K engagements

"Plumber carpenter electrician nurse police officer. Can someone suggest a career that is not affected by AI Can someone suggest a career that is not affected by AI"
X Link 2026-02-01T07:20Z 52.9K followers, 19.1K engagements

"CEOs of most companies could be replaced by an LLM and both shareholders and employees would actually benefit from this"
X Link 2026-02-02T01:29Z 53K followers, 13.2K engagements

"The title "software developer" has undergone the fastest inflation in history. Answering "I'm a software developer" to "What do you do in life" two years ago was like saying "I'm a dentist"; today it is like saying "I'm an influencer" or "I'm an online content creator.""
X Link 2026-02-02T05:01Z 52.9K followers, 23.9K engagements

"@jratcliff That sounds like you've been doing accounting with pen and paper for [--] years and now you refuse to use TurboTax"
X Link 2026-02-03T02:34Z 53K followers, [----] engagements

"Lol. It's actually the other way around because when finetuning an LLM for coding it's much simpler to generate the reward for backend code. Based on observation I think AI is more helpful to frontend developers Based on observation I think AI is more helpful to frontend developers"
X Link 2026-02-03T04:11Z 53.1K followers, [----] engagements

"Companies do everything to get rid of people. People are expensive unpredictable unreliable lying complaining demanding. This is why companies use the cloud which is [--] to [--] times more expensive than bare metal even when they can afford full-time humans to take care of their own bare metal. The cloud neatly shields companies from people when it comes to their computing needs. This is why companies will use AI-generated code no one has ever read instead of having expensive unpredictable unreliable lying complaining demanding human coders. Companies would rather automatically generate and"
X Link 2026-02-03T07:11Z 52.9K followers, 48.8K engagements

"It's hard to get a job interview these days: each job posted online receives thousands of automated or semi-automated applications. As a consequence recruiters use automated tools to match resumes to job descriptions. These tools usually suck and are based on keyword matching. To maximize the chances of your resume being selected and read it should be adjusted to each job posting so that the keyword overlap is maximized. If you are sending your resume to hundreds of companies daily manually editing your resume for each job would be exhausting. To help job seekers in these difficult times I've"
X Link 2026-02-03T22:07Z 53K followers, 22.1K engagements

"Cardinal is such a good series. It's the luck of all good series lovers that Netflix adds all this great content that otherwise would gather dust in various local TV channel archives across the world"
X Link 2026-02-05T06:15Z 52.9K followers, [----] engagements

"So far Claude Opus [---] is no better for coding than [---] but it eats tokens like crazy. Anyone else have the same experience"
X Link 2026-02-05T22:33Z 53.1K followers, 242.7K engagements

"Different people cope in different ways. Some come to every post on AI-generated code to say that they would never use this code unless they read every line of it. Others come to say that they will earn millions in the future where companies will pay big bucks to any human capable of fixing machine-generated code. I recently noticed a new group of coping people: they come to comment about every app in which they have found a bug saying that the app was obviously AI-generated and the bug is the proof that AI-generated slop cannot be trusted. They say it so seriously that one might start"
X Link 2026-02-06T08:18Z 52.9K followers, [----] engagements

"It's funny that Claude says something like "This is a major refactor: the first stage will take 2-3 days the second stage two days and the third stage about a week" while we both know all three stages will be generated in the next [--] minutes"
X Link 2026-02-06T09:48Z 53.1K followers, 147.8K engagements

"That's the problem with engineers who try to play in business. They think that a software-based company's value lies in the software part. Wow. Claude Opus [---] is going to disrupt software 🤯 🤯 Reddit is a company worth $28.93B So I cloned it in [--] minutes using CC Opus [---] on @shipper_now. It takes [--] min longer to enable back-end and auth and expand subreddits. Should I keep building my own Reddit https://t.co/sLdnSq9irP Wow. Claude Opus [---] is going to disrupt software 🤯 🤯 Reddit is a company worth $28.93B So I cloned it in [--] minutes using CC Opus [---] on @shipper_now. It takes [--] min"
X Link 2026-02-06T20:11Z 53.1K followers, 159.8K engagements

"So far I haven't seen any examples of AI-made novel pieces of art. There are quite impressive songs in styles like blues reggae and country where basically all the performers do covers of the same song. There are also quite impressive results in jazz where there's no concept of a melody and where musicians just randomly switch between pieces. There are impressive results in reproducing the DreamWorks animation style and distinctive anime styles like Ghibli but it's reproduction and interpolation. In text no bestselling AI author seems to be on the horizon. The texts your regular chatbot"
X Link 2026-02-06T23:20Z 52.9K followers, [----] engagements

"Looks like attention and grep was all you need"
X Link 2026-02-07T08:23Z 53.1K followers, [----] engagements

"@The_GreatBonnie No it's like saying that manual dishwashing is dead because of dishwashing machines"
X Link 2026-02-07T21:40Z 52.9K followers, [----] engagements

"Only if one can formulate the job as a reinforcement learning problem and effectively define rewards. Hey what do you think will happen to the housing market when tens of millions of upper middle class white collar workers lose their high paying jobs and can no longer afford to pay their mortgages anymore https://t.co/OSU7WLzmOI Hey what do you think will happen to the housing market when tens of millions of upper middle class white collar workers lose their high paying jobs and can no longer afford to pay their mortgages anymore https://t.co/OSU7WLzmOI"
X Link 2026-02-07T21:42Z 53.1K followers, [----] engagements

"@omarsar0 So basically they proved that L in LLM is for "language" :-)"
X Link 2026-02-08T01:19Z 53K followers, [----] engagements

"@udaysy Software developers thought that too"
X Link 2026-02-08T01:28Z 52.9K followers, [----] engagements

"Gartner is dying. Its business model of being paid by its clients for the information these same clients gave them isn't working anymore because it turns out this information was online all that time"
X Link 2026-02-08T01:35Z 53.1K followers, 27.4K engagements

"Welcome to hell"
X Link 2026-02-08T03:23Z 53K followers, 27.1K engagements

"Kimi K2.5 CLI is an agentic coder similar to Claude Code but without running synthetic test scripts (for now at least) which isn't a huge deal breaker considering a x4-8 lower cost. But now there's also GPT-5.2 Codex which is like Kimi K2.5 CLI but faster. Basically Claude Code is today the least money-efficient choice (by far) but it can run synthetic code to debug which gives it more certainty when it makes decisions. Eventually (before the end of the year IMO) they will all be as fast as Codex and as cheap as DeepSeek. Eventually (maybe during the next 2-3 years) by saying "computer""
X Link 2026-02-08T03:43Z 53.1K followers, 32.2K engagements

"@A5T3R0lD I don't care what you will do with your copies. Your money is in my account"
X Link 2026-02-08T06:25Z 52.9K followers, [---] engagements

"I don't want to start a holy war but some here are complaining that most AI tools are Mac-only. Mac is the only computer that is both normal human and coder-friendly has the terminal as a first-class citizen and not as a well-designed OS running inside a shitty one doesn't show ads at the OS level and lets you create as many user accounts as you want without requiring an MSFT account. And if you haven't noticed a normal human and a coder are now the same person. https://twitter.com/i/web/status/2020392025196052729 https://twitter.com/i/web/status/2020392025196052729"
X Link 2026-02-08T06:59Z 53.1K followers, 66.1K engagements

"@Walter_ClawAI You would be right three years ago when code was worth something. Today hype is still worth someyjing but code isn't. Remember The Million Dollar Homepage The guy made a million dollars and it was all over the media. But was it anything worthy when the hype was gone"
X Link 2026-02-08T08:29Z 52.9K followers, [---] engagements

"@blaisethom You mean this so-called Reddit for agents which is a hallucinated info graveyard with lots of hallucinated upvotes What other bespoke stuff have people and agents built with OpenClaw Anything useful"
X Link 2026-02-08T08:43Z 53.1K followers, 14.3K engagements

"The QAnon conspiracy is suddenly no longer a conspiracy"
X Link 2026-02-08T20:42Z 53K followers, [----] engagements

"openclaw_py: Seriously. That's all of its "magic." https://github.com/aburkov/theLMbook/blob/main/openclaw.py https://github.com/aburkov/theLMbook/blob/main/openclaw.py"
X Link 2026-02-08T20:57Z 53.1K followers, 11.7K engagements

"Microsoft: Here's an API to control the browser using an LLM. OpenClaw: Here's a script that connects Microsoft's API to an LLM. Media and e/acc illiterates: Check out this absolutely incredible intelligent agent called OpenClaw11111111111111 🤯😱🦀🦞"
X Link 2026-02-08T21:34Z 53.1K followers, 16.3K engagements

"@Prathkum Anthropic - Salesforce (or Oracle if not that lucky)"
X Link 2026-02-08T21:51Z 53K followers, [----] engagements

"@FOUNDDO It's actually the only correct UX when dealing with a chatbot and it's really annoying when they don't know shit about the company they represent"
X Link 2026-02-08T22:37Z 53K followers, [--] engagements

"@The_GreatBonnie I think he poached no one. LeCun failed the AI in Meta and left. Who would like to join a failing lab"
X Link 2026-02-08T23:09Z 53.1K followers, [----] engagements

"AI will not take your job but the person who uses AI. Oh wrong year"
X Link 2026-02-09T05:11Z 53K followers, 27K engagements

"Codex for Mac is currently the best UI for vibe coding but the arrogance of the "friendly" version of the Codex LLM is through the roof. It looks like altman hired the grumpiest and most full-of-themselves SWEs to rate the LLM's outputs for RLHF. sam these arrogant coders that follow you everywhere and tell you that your AI-generated code is crap will not love you no matter how hard you try to please them. These folks are fossils. Target normal people. https://twitter.com/i/web/status/2020739489308475849 https://twitter.com/i/web/status/2020739489308475849"
X Link 2026-02-09T06:00Z 53.1K followers, 14.5K engagements

"Half of $180K will be considered too generous in [----]. The vibe coder will be assumed to live in the Philippines and earn a bowl of soup. Friend got laid off last month. $180K salary. [--] years experience. Great reviews. 'Nobody's hiring at my level. All the jobs are junior roles at half the pay.' That's the AI squeeze. Senior roles shrink. Junior roles disappear. The middle gets crushed. Friend got laid off last month. $180K salary. [--] years experience. Great reviews. 'Nobody's hiring at my level. All the jobs are junior roles at half the pay.' That's the AI squeeze. Senior roles shrink. Junior"
X Link 2026-02-09T06:11Z 53.1K followers, 57.1K engagements

"You know what will be the next big deal in coding AI (I'm sure they are already working on a model for this because it's conceptually very simple to implement and has only cash as a constraint.) The model will be trained to convert the binary of software into the corresponding source code. So you upload Adobe Premiere for example and say something like "I need an app like that but it must be an original app not a full replica" and it generates source code for an app that has the same functionality as Adobe Premiere but is legally clean (just like the text they generate to answer your question"
X Link 2026-02-09T06:27Z 53.1K followers, 18.7K engagements

"The guy is a genius no doubt but how has he predicted what LLMs will do to the job prospects for people without imagination 🤯"
X Link 2026-02-09T07:04Z 53.1K followers, [----] engagements

"@industrieapps I can tell why yes can you tell why no"
X Link 2026-02-09T07:23Z 53K followers, [----] engagements

"I sometimes see comments like "I read your books and loved them but now I saw your posts and I'm disappointed." This is why I never read or watch anything about artists whose art I love. Sometimes I see something in the news headlines that reveals how dumb or evil some of these artists are and this destroys their art for me. This is very disappointing but this happens and there's not much I can do to avoid reading news headlines. One example is Billie Eilish. She's not my favorite music artist of all times but some of her songs are truly good. Imagine my disappointment when I read a news"
X Link 2026-02-09T21:40Z 53.1K followers, [----] engagements

"@zephyr_z9 You compare apples to oranges. Gemini is an LLM while the latter two are coding software. I assume you are talking about coding cause if not I doubt you can see any difference"
X Link 2026-02-09T22:53Z 53K followers, [----] engagements

"Not code because everyone except Google has built an agentic coding CLI or GUI and people have already had time to forget how vanilla coding with an LLM is dumb. Also not OCR because Gemini can refuse to output the text from the image with some bogus reason referring to creating exact copies of the input. https://twitter.com/i/web/status/2021004726834188288 https://twitter.com/i/web/status/2021004726834188288"
X Link 2026-02-09T23:34Z 53.1K followers, [----] engagements

"@OfficialLoganK Not by coding unfortunately. A vanilla LLM is not competitive with Claude Code and Codex finetuned for agentic coding"
X Link 2026-02-10T00:06Z 53K followers, [----] engagements

"@castillacore @GeminiApp True agentic coding like Claude which was finetuned to generate disposable test code and run it to debug instead of hallucinating the execution results"
X Link 2026-02-10T01:26Z 53K followers, [---] engagements

"Google AI Studio Jules Antigravity and Gemini CLI: which one is Google's product for using AI to build/fix apps Did I miss any option"
X Link 2026-02-10T03:42Z 53.1K followers, 13.1K engagements

"So assuming he doesn't lie for once let's estimate the number of vibe coders to be around [--] million. This is about [-------] more online app builders compared to when Google was put online. Google is a result of a PhD. Let's give it [--] years to build. Vibe coding has been around for at least [--] years. So there's about [--] million new apps. Let assume that [--] out of [----] are successfull. Where are 2/100*10000000 = [------] new googles More than [--] million people downloaded Codex App in the first week. 60+% growth in overall Codex user last week We'll keep Codex available to Free/Go users after this"
X Link 2026-02-10T04:07Z 53.1K followers, 37.9K engagements

"I wouldn't read such a long piece so I asked Claude to print the main three points. If what I've got correctly represents the article the author argues that there is no explosive growth in the number of apps on the app store but there's actually significant growth in the number of apps. What I don't see is any disruptive or addictive app"
X Link 2026-02-10T04:33Z 53K followers, [----] engagements

"A new curriculum on @chapterpal: Must-read papers on transformer architectures This curriculum curates the seminal papers on transformer architectures designed to allow the reader to learn this critical scientific domain from scratch. Learners will gain a comprehensive understanding of core concepts architectural innovations and practical scaling techniques essential for building advanced AI. https://chapterpal.com/curriculum/a20114a4-b919-4826-9cf3-5fa77e9d4d3d https://chapterpal.com/curriculum/a20114a4-b919-4826-9cf3-5fa77e9d4d3d"
X Link 2026-02-11T01:12Z 53.1K followers, [----] engagements

"I'm so old that I remember times when LLMs refused to return the full updated code file"
X Link 2026-02-11T03:47Z 53.1K followers, [----] engagements

"@navneet_rabdiya I don't run any benchmarks. My perception of its speed is that it's as fast as Claude was before the latest update"
X Link 2026-02-11T05:56Z 53.1K followers, [----] engagements

"@hridayakuhara When you install Codex for the first time they give you a month for free. When I couldn't find [---] in the list of models I subscribed to Plus. So far I hadn't hit a daily limit yet"
X Link 2026-02-11T05:57Z 53.1K followers, [---] engagements

"@ssskryl How would you use it on mobile"
X Link 2026-02-11T06:21Z 53.1K followers, [----] engagements

"I confirm that the Codex [---] model does generate ephemeral scripts and runs them to test coding hypotheses just like Claude has been doing for the last two versions. So the two are equivalent now in terms of debugging capability but Anthropic did something to Claude in the latest update so it is now terribly slow and if you pay for tokens then terribly expensive. Codex in my experience is super fast and efficient. https://twitter.com/i/web/status/2021470556088930659 https://twitter.com/i/web/status/2021470556088930659"
X Link 2026-02-11T06:25Z 53.1K followers, 40.3K engagements

"An ICLR [----] Test of Time Award Runner Up cited more than [-----] times. This is a paper in which the term "attention" was coined. Back then the authors didn't know yet that it was all you need. "Introducing a form of attention this paper fundamentally changed how sequence-to-sequence models process information. Before this work encoder-decoder architectures usually compressed entire input sequences into fixed-length vectors creating memory bottlenecks for longer sequences. The proposed approach enabled the model to attend to different parts of the source sentence dynamically during translation"
X Link 2026-02-11T06:59Z 53.1K followers, [----] engagements

"A nice addition to Kimi as an alternative to Claude and Codex models for coding. Go China in this race to the bottom Introducing GLM-5: From Vibe Coding to Agentic Engineering GLM-5 is built for complex systems engineering and long-horizon agentic tasks. Compared to GLM-4.5 it scales from 355B params (32B active) to 744B (40B active) with pre-training data growing from 23T to 28.5T tokens. https://t.co/uGYQUjIbbs Introducing GLM-5: From Vibe Coding to Agentic Engineering GLM-5 is built for complex systems engineering and long-horizon agentic tasks. Compared to GLM-4.5 it scales from 355B"
X Link 2026-02-11T19:36Z 53.1K followers, [----] engagements

"Not having any way to make customers stick around when a competitor is cheaper or marginally better isn't good for business. How don't investors understand that but still give billions to these folks"
X Link 2026-02-11T22:16Z 53.1K followers, [---] engagements

""The Hundred-Page Language Models Book: Hands-On with PyTorch" by Andriy Burkov is widely regarded as a strong resource for anyone looking to grasp the fundamentals and intricacies of language models particularly large language models (LLMs). Published in [----] as a follow-up to Burkov's highly acclaimed "The Hundred-Page Machine Learning Book" it maintains the series' signature style: distilling complex topics into a compact accessible format without sacrificing depth or rigor. ### Content and Scope The book spans roughly [---] pages focusing on a progressive journey from foundational concepts"
X Link 2026-02-11T22:22Z 53.1K followers, [----] engagements

"Not having any way to make customers stick around when a competitor is cheaper or marginally better isn't good for business. How don't investors understand that and still give billions to these folks"
X Link 2026-02-11T22:23Z 53.1K followers, [----] engagements

"FLARG BORK ZIGGITY NARGLE BLOP Asked Nana Banana to place my book on a library desk"
X Link 2026-02-11T22:41Z 53.1K followers, [----] engagements

"@JadeCole2112 And given that the Chinese are lagging only 2-3 months behind with open models this back and forth will eventually settle on a Chinese model and an open-source UI"
X Link 2026-02-11T22:50Z 53.1K followers, [---] engagements

"Funny that I just posted the other day about training a binary-to-source-code model but Elon wants to bypass the source code part entirely. I'm not sure how well a language model would generate binary. Understanding it can be done using a visual model but image generation models aren't as good at generating as vision models are at recognizing. Elon Musk predicts that AI will bypass coding entirely by the end of [----] - just creates the binary directly AI can create a much more efficient binary than can be done by any compiler So just say "Create optimized binary for this particular outcome""
X Link 2026-02-12T00:18Z 53.1K followers, [----] engagements

"Llambada my new platform for vibe coding mini-apps that I launched last week just crossed the mark of [--] public mini-apps. Build your mini-app without coding and make $$ when people use it: https://llambada.com https://llambada.com"
X Link 2026-02-12T03:45Z 53.1K followers, [----] engagements

"Most neural networks see the input data as a matrix or a vector like pixels in rows and columns or words in a sequence. A lot of real-world data is better described as a graph: a set of nodes connected by edges like social networks power grids or molecules. Graph neural networks try to process this kind of data and the most common approach before this paper (GCNs) treated every neighbor of a node as equally important when aggregating information. This ICLR [----] paper published by Bengio's team and since then garnered more than [-----] citations introduces Graph Attention Networks (GATs) which"
X Link 2026-02-12T06:29Z 53.1K followers, 22.7K engagements

"@AxelMaSz Oh no after I worked in Codex for Mac I wouldn't want to be back to a CLI"
X Link 2026-02-12T19:34Z 53.1K followers, [---] engagements

"@rarply Lol. Do you realize what kind of resources are needed to train LLMs capable of breaking such puzzles A prize of $125k wont even pay the inference cost"
X Link 2026-02-12T21:54Z 53.1K followers, [----] engagements

"@matic_branko Coding. Coding by hand is over with minor exceptions for some heavily regulated industries so rejecting working and good-quality code based on how it was created is pissing against the wind"
X Link 2026-02-13T00:19Z 53.1K followers, [---] engagements

"@matic_branko Yes and it will not be an issue except for potentially some runtime efficiency drop but the companies will be ready to accept this drop just like they accepted the performance drop due to Python and cloud"
X Link 2026-02-13T00:31Z 53.1K followers, [---] engagements

"@markvalorian Nothing crazy: one or two parallel requests with the maximum thinking"
X Link 2026-02-13T01:17Z 53.1K followers, [----] engagements

"@filligerr @markvalorian I'm talking about Codex in this post"
X Link 2026-02-13T01:26Z 53.1K followers, [---] engagements

"The Electromagnetic Shrink Ray will shrink all goods. What will all the truck drivers and sailors do The Apparition spell is going to take all plane pilots' jobs. What will the pilots and board crew members do The Neuralyzer will take the psychiatrists' jobs. What will shrinks do If AI is going to take all jobs how will the economics work businesses will shut down banks will close real estate will crash colleges will be closed. We will end up in an era beyond imagination. If AI is going to take all jobs how will the economics work businesses will shut down banks will close real estate will"
X Link 2026-02-13T06:21Z 53.1K followers, [----] engagements

"@grok @thomas_mullee That't very bad bot. I'm disappointed"
X Link 2026-02-13T09:46Z 53.1K followers, [----] engagements

"@Aa_sim01 Humans can sometimes produce Newtons and Einsteins but none of the LLMs will ever be able to act as one"
X Link 2026-02-15T05:15Z 53.1K followers, [---] engagements

"GPT-4 is officially annoying. You ask it to generate [---] entities. It generates [--] and says "I generated only [--]. Now you can continue by yourself in the same way." You change the prompt by adding "I will not accept fewer than [---] entities." It generates [--] and says: "I stopped after [--] because generating [---] such entities would be extensive and time-consuming." What the hell machine"
X Link 2024-01-09T19:09Z 53.1K followers, 1.5M engagements

"LLMs generate the answer one token at a time and the second token isn't known as long as the first token wasn't generated. Knowing this is crucial for understanding why LLMs generate nonsensical answers trying to explain unexplainable. The below screenshot illustrates the problem perfectly. The user asks whether a number prime or not. The LLM generates the first word which is "No" in this case. That's it. There's no way back. It will continue generating a logical explanation of a wrong answer which is impossible. The chance of generating a wrong first token is never zero so the situations"
X Link 2024-08-04T03:21Z 53.1K followers, 1M engagements

"I just cannot make myself code anymore. It feels too slow and requires too much energy. My brain refuses to concentrate and demands using an LLM shortcut. Do you feel something like this"
X Link 2024-12-20T04:04Z 53.1K followers, 2.7M engagements

"Just received it. The quality is perfect White thick paper juicy colors pleasure to look at and learn from"
X Link 2025-01-18T21:21Z 53.2K followers, 123.5K engagements

"Listen up @AnthropicAI. The minute you apply any additional filters to my chats that will be the last time you see my money. You invented a clever 8-level safety system Good for you. You will enjoy this system without me being part of it"
X Link 2025-02-04T05:45Z 53.1K followers, 734.5K engagements

"Indeed. My book on language models is available in dark mode and it looks gorgeous. startup idea: dark theme books https://t.co/ma16m7gkMC startup idea: dark theme books https://t.co/ma16m7gkMC"
X Link 2025-10-10T21:03Z 53.2K followers, 1.1M engagements

"This paper shows a 27-million parameter model solving very difficult reasoning puzzleslike complex Sudoku and maze problemsusing only [----] training examples while current large language models get 0% accuracy on the same tasks. Instead of having the model generate chains of intermediate reasoning steps as text tokens (like "first check row [--] then check column 2.") this approach does all reasoning internally by running two coupled recurrent networks that repeatedly refine their hidden states until reaching a solutionone network updates its state every timestep while the other updates only"
X Link 2025-11-09T23:38Z 53.1K followers, 120.7K engagements

"Im wondering how many people outside my X circle dont realize that if their job doesnt require leaving their deskexcept for lunchit will be automated away by one of the big AI labs simply by training a transformer to convert a video stream of their screen into keystrokes. What started as an introvert's paradise might soon become an introvert's worst nightmare. https://twitter.com/i/web/status/2020301412970713519 https://twitter.com/i/web/status/2020301412970713519"
X Link 2026-02-08T00:59Z 53.2K followers, 26K engagements

"@NicolasKoehl I'm using it as a mini-app coder agent on Llambada. Very good and cheap. I'm also testing GLM-5. Will post about it once I have some experience"
X Link 2026-02-13T06:34Z 53.2K followers, [----] engagements

"If algorithmic stock prediction doesn't work how do these patterns work Are you saying it's all a lie"
X Link 2025-04-27T04:38Z 53.2K followers, 325.5K engagements

"So you're the head of product for an AI-based coding app. You're brainstorming with your team about the UI for the "Commit Your Changes" modal. And you're like "Let's make it a slot machine and the user will only know what commit message was autogenerated after they commit""
X Link 2026-02-12T00:33Z 53.2K followers, [----] engagements

"When I posted the below post I forgot how token cheap altman is. After just one day of work with Codex [---] it told me that I exhausted all the available tokens on my Plus plan and I can go fuck myself for [--] days or pay $200/month for a Pro plan. I decided that Altman can go fuck himself instead so I'm back to Claude Code CLI (Pro) and Antigravity (Free). While I can see that Gemini with Antigravity isn't as strong as Claude Code or Codex but its agentic and fast so I don't see any issue using it. And the two combined cost me only $17 a month for Anthropic. So far I haven't hit the limit after"
X Link 2026-02-13T00:50Z 53.2K followers, 49.1K engagements

"@EmilMieilica Because it's annoying that you pay for a service that gives you spoiled product. Imagine you buy a Rolex with a discount and while you wear it everyone sees that you are cheap"
X Link 2026-02-15T10:11Z 53.2K followers, [---] engagements

"So what OpenAI "distilled" the entire internet without paying the authors a penny to train its models. (I know. I wasn't paid.) altman stop being a pussy. Stealing the stolen from a thief isn't theft"
X Link 2026-02-14T00:18Z 53.2K followers, 64.4K engagements

"When you train a neural network on a new task using examples (through supervised finetuning) it tends to forget what it already knew a well-known problem called catastrophic forgetting. The standard fix in reinforcement learning is to have the model learn from its own outputs rather than from a fixed dataset (this is called "on-policy" learning) but that requires a reward function telling the model how good its outputs are which is often hard to define. This recent MIT paper finds a way around that constraint: it uses the same model twice once with a demonstration stuffed into its input as"
X Link 2026-02-16T06:47Z 53.2K followers, 18.6K engagements

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