Great course, great notes - and so much coverage of matching and combinatorial optimization! Matching is the best. Everyone should study matching. Life is matching. Love is matching. All hail matching.

$> whois alignmentalignmentalignment dot ai # whois.nic.ai Domain not found. -- Hate to compete within Mozilla, but Louis Barclay I can see the next YC company coming in to de-throne you 😉. Have to build that moat! Align more!

Software engineers shouldn't fear being replaced by AI. They should fear being asked to maintain the sprawling mess of AI-generated legacy code their employer's systems will soon run on. Because that one will actually happen.

.Marc Andreessen 🇺🇸: "My guess is we are going to live in a world in which most aggregate AI is going to be executed probably on smaller form factors and probably most of that is going to be open source"

The VLDB paper "SQLite: Past, Present and Future" is an excellent deep dive into SQLite's architecture and design philosophy. Worth reading whether you're just getting started or already shipping SQLite in production. vldb.org/pvldb/vol15/p3…

Go go Relace - one of *the* most dialed-in, talented teams out there building the models that power coding agents. Congrats on the raise Preston Zh Eitan Borgnia and the whole team, great to see this support from a16z and Matrix.

.@HackAPrompt team on top as always - evaluations are hard!

some VC firm should rebrand associates stalking founders at events to 'forward deployed capitalists'

Relace Repos is here! It's GitHub for agents with semantic search out-of-the-box. Just use standard git operations without worrying about rate limits. After each push, we auto-index the updated files and use a reranker to make sure each search tool call is fresh. Retrieval is

"Researchers need open source. Developers need open source. Companies around the world, we need open source. Open source models is really, really important." Thanks NVIDIA for the amazing contributions to open source AI, everyone should follow hf.co/nvidia

Why do computer scientists put word count limits on recommendation letters, when we know full well that: * We write in LaTeX; and * Word counting in a PDF is the hardest problem in all of computer science, harder than AGI, harder than ASI. Why do we do this to ourselves.

Join apuchitnis, John P Dickerson (of mozilla.ai) and I for our session at Mozilla Festival on the collective land grab for user context reshaping the internet. Join the discussion if you'll be there: schedule.mozillafestival.org/session/230 Mozilla x koodos labs💥

#MozFest 2025 🎉   From November 7–9, creators and builders are coming together to rethink how tech can better serve people.   Peter, Javier, and Mario from Mozilla.ai are attending — and our fearless leader, John P Dickerson, is co-hosting “Context is All You

Introducing any-llm-gateway 🏗️ An open-source FastAPI gateway to manage LLM costs, budgets, and access across providers. Deploy, monitor, and control usage with confidence. Read the full post: blog.mozilla.ai/control-llm-sp…

AI research developments are translated into real tools faster than ever, but what’s coming next? Join Sasha Rush (Cursor/Cornell Tech), Soumith Chintala (Creator of PyTorch), and Jonathan Frankle (Chief Scientist at Databricks) for a discussion at South Park Commons NYC next week. ⬇️

3 reasons we’re excited about local LLMs: 1️⃣ Full control 2️⃣ Privacy-first 3️⃣ Works everywhere (macOS, Linux, Windows) That’s llamafile. 🦙 Learn more: github.com/mozilla-ai/lla…

Too many LLM APIs can slow you down. any-llm (v1.0) lets you run cloud and local models through one interface. Try it here: github.com/mozilla-ai/any…

Completely unsurprising for everyone's go-to small vision language model. Awesome family of models, looking forward to the full moondream3.

Not a head-scratcher.

It's an absolute pleasure that Meta / AI at Meta release open-weight multi-modal models like SAM 3. Put this in the hands of the people! Heck yes.