Awesome interview @bitronix https://t.co/0SedBvWLtx Lot of insights about JVM performance, transactions & distributed systems. Fun days of @terracottatech .

cool paper studying the differences between SFT and DPO.

We’re releasing Bloom, an open-source tool for generating behavioral misalignment evals for frontier AI models. Bloom lets researchers specify a behavior and then quantify its frequency and severity across automatically generated scenarios. Learn more:

@signulll Here’s a significantly longer post if you want to read the specifics

Introducing Latent-X2 — AI-generated antibodies with drug-like developability and low immunogenicity in human panels, zero-shot. Technical report: https://t.co/9uU6GQCdM7 Blog: https://t.co/TFyCzlZ4qN Apply for access: partnerships@latentlabs.com

Announcing a16z Build: A dinner series and community for founders, technologists, and operators figuring out what they want to build next — and who they want to build with. Apply or nominate today: https://t.co/urDzCOP96t

First big project i've been been cooking up at @a16z... excited to share: a16z Build A dinner series and community for founders, technologists, and operators figuring out what they want to build next — and who they want to build it with. Great startups don’t come from ideas

In yet another example of how important reinforcement learning has become in the AI industry, inference provider Baseten has acquired RL startup Parsed. We get into the deal in this morning's AI Agenda: https://t.co/N6GRDzt60T

LLMs are claimed to reach PhD intelligence, but still fail mundane tasks. To understand this challenge, Databricks launched OfficeQA, a benchmark of useful tasks that require reliability&diligence, not specialized knowledge. We're also doing a competition!

My takeaways from Neurips 1. Continual learning. To support this next frontier, we’re going to need new architectures, new reward functions, new data sources, and new revenue models. 2. Neolabs. Frontier research for risky bets is being shared across multiple companies now 3.

Thank you to everyone who made the inaugural Virtual Cell Challenge a success. Over 5,000 participants from 114 countries competed to build AI models that predict cellular responses to genetic perturbations. Today we're announcing the winners and reflecting on what we learned.

https://t.co/eRVcFX8leD Kefir C17/C23 independent C compiler by Jevgenij Protopopov; Q&A. Verification, quickcheck, determinism, inlining , unrolling, validating peepholes, Graph coloring register allocator, Advent of code, Simple modules

Thank you to all the builders who joined us at re:Invent. We created The Kernel—a newspaper by builders, for builders. I hope it sparks new ideas as you head into the new year. Now go build! https://t.co/LKAtAZOOEZ

Presented the Innovation Award to @antirez last Friday. His work is the definition of Blue Ocean innovation, earning him this well-deserved recognition for his significant global impact on the digital ecosystem. https://t.co/5U5LpQwGpe

Join us for Jon Kleinberg's Theoretically Speaking lecture, 12/9! This is an in-person event, and registration is required. https://t.co/uU1y7Co9zU

Honeycomb published a blog post called "The End of Observability as We Know It." I've been using observability tools for years now. Heck, I wrote a book about CloudWatch ⛅ And honestly? I like the take of this post. Let's see a typical daily observability issue: - You have 5

Bf-trees are a high-performance alternative to the classic B-tree. The key contribution: Decoupling the size of on-disk pages from the size of memory cache entries. I'm curious if/when MySQL and Postgres will adopt modern versions of classic indexes.

The world of compilers is fascinating. Learn about flattening ASTs, how it's implemented, and associated performance benefits - better locality, simpler lifetimes, and speedups. https://t.co/YcIFFIFj46

We have 14 survey lectures for our @SimonsFdn Collaboration on the Physics of Learning and Neural Computation! All videos available at: https://t.co/miM7HjZeO7 Here is the list: @zdeborova: Attention-based models and how to solve them using tools from quadratic networks and

Talk at Ray Summit on "Building Cursor Composer." Overview of the work from our research team. https://t.co/9a5yeC3IT8