Moustafa Banbouk @KnowledgeRep

SpaceX just acquired Cursor for $60B > be Cursor > 4 MIT students start a side project in 2022 > build the AI coding tool developers love > hit a $10B valuation > decide copilots aren’t enough > move into models > need massive compute to compete with OpenAI, Anthropic & Google > meanwhile xAI is losing the coding race > realizes catching up could take years > skips the line > buys Cursor for $60B > Cursor gets compute > xAI gets the coding leader > founders become multi-billionaires

Andrej Karpathy spent 2h showing how he actually uses AI day to day he's a co-founder of OpenAI and led AI at Tesla, so when he shows how he works, it’s worth watching and the whole session is just him telling the machine what he wants in simple terms, like he's briefing a coworker watch what's actually happening the entire time: > he describes the task in normal words > it goes off and does the work > he glances at the result and nudges it with one more sentence that's the whole skill, and you've had it since you learned to talk the only gap between that and a worker that runs on its own is handing that sentence a schedule and the tools to act check his work, then build the version that keeps working when you stop

Chrome @0xchromium

2 weeks ago

x.com/i/article/2061…

12 93 701 1.8M 3K

LLM Knowledge Bases Something I'm finding very useful recently: using LLMs to build personal knowledge bases for various topics of research interest. In this way, a large fraction of my recent token throughput is going less into manipulating code, and more into manipulating knowledge (stored as markdown and images). The latest LLMs are quite good at it. So: Data ingest: I index source documents (articles, papers, repos, datasets, images, etc.) into a raw/ directory, then I use an LLM to incrementally "compile" a wiki, which is just a collection of .md files in a directory structure. The wiki includes summaries of all the data in raw/, backlinks, and then it categorizes data into concepts, writes articles for them, and links them all. To convert web articles into .md files I like to use the Obsidian Web Clipper extension, and then I also use a hotkey to download all the related images to local so that my LLM can easily reference them. IDE: I use Obsidian as the IDE "frontend" where I can view the raw data, the the compiled wiki, and the derived visualizations. Important to note that the LLM writes and maintains all of the data of the wiki, I rarely touch it directly. I've played with a few Obsidian plugins to render and view data in other ways (e.g. Marp for slides). Q&A: Where things get interesting is that once your wiki is big enough (e.g. mine on some recent research is ~100 articles and ~400K words), you can ask your LLM agent all kinds of complex questions against the wiki, and it will go off, research the answers, etc. I thought I had to reach for fancy RAG, but the LLM has been pretty good about auto-maintaining index files and brief summaries of all the documents and it reads all the important related data fairly easily at this ~small scale. Output: Instead of getting answers in text/terminal, I like to have it render markdown files for me, or slide shows (Marp format), or matplotlib images, all of which I then view again in Obsidian. You can imagine many other visual output formats depending on the query. Often, I end up "filing" the outputs back into the wiki to enhance it for further queries. So my own explorations and queries always "add up" in the knowledge base. Linting: I've run some LLM "health checks" over the wiki to e.g. find inconsistent data, impute missing data (with web searchers), find interesting connections for new article candidates, etc., to incrementally clean up the wiki and enhance its overall data integrity. The LLMs are quite good at suggesting further questions to ask and look into. Extra tools: I find myself developing additional tools to process the data, e.g. I vibe coded a small and naive search engine over the wiki, which I both use directly (in a web ui), but more often I want to hand it off to an LLM via CLI as a tool for larger queries. Further explorations: As the repo grows, the natural desire is to also think about synthetic data generation + finetuning to have your LLM "know" the data in its weights instead of just context windows. TLDR: raw data from a given number of sources is collected, then compiled by an LLM into a .md wiki, then operated on by various CLIs by the LLM to do Q&A and to incrementally enhance the wiki, and all of it viewable in Obsidian. You rarely ever write or edit the wiki manually, it's the domain of the LLM. I think there is room here for an incredible new product instead of a hacky collection of scripts.

How did we go from vanilla RAG to agentic RAG SO fast? Just some years ago, we were excited about basic retrieval-augmented generation. Now we're building full AI agents that can reason, plan, and use multiple tools autonomously. Here's what changed everything: 𝗩𝗮𝗻𝗶𝗹𝗹𝗮 𝗥𝗔𝗚 had some serious limitations: • Only one external knowledge source • One-shot retrieval with no validation • No reasoning about retrieval quality 𝗔𝗴𝗲𝗻𝘁𝗶𝗰 𝗥𝗔𝗚 changes the game completely. Instead of a simple retrieve-and-generate pipeline, we now have AI agents that can: • Route queries to specialized knowledge sources • Validate retrieved context before using it • Access multiple tools (web search, calculators, APIs) • Perform multi-step retrieval and reasoning The technical shift happened through two main approaches: 1️⃣ 𝗙𝘂𝗻𝗰𝘁𝗶𝗼𝗻 𝗰𝗮𝗹𝗹𝗶𝗻𝗴 - LLMs can now reliably connect to external tools and APIs 2️⃣ 𝗔𝗴𝗲𝗻𝘁 𝗳𝗿𝗮𝗺𝗲𝘄𝗼𝗿𝗸𝘀 - DSPy, LangChain, CrewAI making it easier to build complex multi-agent systems 3️⃣ E𝗻𝘁𝗲𝗿𝗽𝗿𝗶𝘀𝗲 𝗮𝗱𝗼𝗽𝘁𝗶𝗼𝗻 - companies like Replit and Microsoft started shipping agents in production But here's the thing - this rapid evolution brings both exciting possibilities and new challenges. While agentic RAG produces more accurate responses, it also introduces added latency and potential failure modes. The speed of this transformation is honestly crazy. We went from basic retrieval to full reasoning agents in what feels like months, not years. Read more in this blog post: weaviate.io/blog/what-is-a…

This 277-page PDF unlocks the secrets of Large Language Models. Here's what's inside: 🧵

🚨 Anthropic just solved the problem every AI agent engineer’s been screaming about for a year. Every agent today burns tokens like fuel every tool call, every definition, every intermediate result jammed into context. Now Anthropic’s introducing the fix: code execution with MCP. Instead of calling tools directly, agents now write code to call them and it changes everything. → 98.7% fewer tokens → 10x faster task completion → No context overload → No data leakage Think of it like this: Old agents talk about what to do. New agents just code and do it. Cloudflare calls it “Code Mode.” Anthropic just made it real. This is the turning point. The next generation of AI agents won’t prompt tools they’ll build with them.

7 Laws That Give You An Unfair Advantage: ~Thread~

Harvard just dropped a book on ~ML systems engineering~ and it's 100% free in PDF (it will be published in print by MIT Press). systems engineering is the hot skill companies die to see in candidates, but few books explain it in the context of ML. What it covers: → ML system foundations → a primer on DL → DNN architectures ⃗→ design principles and AI workflows ⃗→ data engineering → AI training (distributed, mathematics, etc.) ⃗→ performance engineering and model optimization ⃗→ AGI systems and what to expect!!! ⃗→ how to work with SLMs and VLMs the subjects cover 100% of what you need to know as a mid-level ML engineer, and the best part is that it is updated to cover what matters in 2025. get the book here: mlsysbook.ai

You can now chat with apps in ChatGPT.

Simulations are the future, & one of the main tools we’ll ultimately use to understand and predict things about the universe. This is why I’m so excited about Genie 3, our latest interactive world simulator - here are some insanely cool things you might have missed about it 🧵:

There is a book on Amazon called "Modern AI Revolution" by Geoffrey Hinton. This is a scam. It has nothing to do with me and I wish Amazon would remove it.

And coming soon to Free, Enterprise, and Edu users 🪐

OpenAI Developers @OpenAIDevs

a year ago

ChatGPT for macOS can now edit code directly in IDEs. Available to Plus, Pro, and Team users.

356 568 5K 1.4M 2K

Game over. Grok won. If you’re an investor in OpenAI, you can report the tax loss to your accountant this year.

Level Up Your Daily Tasks with These AI Tools 👌 1. Logo - Logomaster.ai 2. Images - Stockimg.ai 3. Writing - Cohesive.so 4. Forms - Al.feathery.io 5. Health - Docus.ai 6. Emails - Emailmagic.ai 7. Design - Diagram.com 8. Resume - ResumAl.com 9. Search - Gptgo.ai 10. Notes - Reflect.app 11. Blogs - Autoblogger.ai 12. Debugging - Jam.dev/jamgpt 13. Meetings - Loopinhq.com 14. Summaries - Spoke.ai 15. Video - Invideo.io 🎁 Bonus: Stay ahead in the AI revolution with my newsletter! Join 22,000 readers to get the latest AI insights, tools, tutorials, and guides. Subscribe for free: 👇 joinaiinsights.com/subscribe

RIP ChatGPT It's only been a week since Deepseek R1 dropped and people can't stop talking about it! Yes, it's that good. 10 mind-blowing examples you need to see (don't miss #6):

If you could rewrite your story, you’d do many things differently. What changes would you make today?

In 2014, Elon "gave away" Tesla's secrets to BWM. Everyone thought he was crazy. But this "act of charity" was actually the most ruthless business move in corporate history. Here's why BMW never saw it coming: 🧵

We've released a new library, fastdata, for high quality synthetic data generation! 🚀 Check out this deep dive thread and blog post from its creator explaining everything you need to know to get started:

Nathan Cooper @ncooper57

2 years ago

As a lead researcher at @StabilityAI, I worked a lot with synthetic data to train LLMs and VLMs. It is the most underrated way of boosting model performance. Now at @answerdotai I've been working to make the best practices easy—read on to learn how!

16 79 608 126K 637

This is beyond exciting! #AI’s far reaching impact is just beginning. Congratulations to @geoffreyhinton and John Hopfield! ❤️

The Nobel Prize @NobelPrize

2 years ago

BREAKING NEWS The Royal Swedish Academy of Sciences has decided to award the 2024 #NobelPrize in Physics to John J. Hopfield and Geoffrey E. Hinton “for foundational discoveries and inventions that enable machine learning with artificial neural networks.”

978 13K 32K 12.7M 3K