one day you’ll realise that nobody was actually watching and you could have done what you wanted.

I want to buy shares in LoRA-XS

This experiment is kind of useless How much edge do you think an LLM has on a market? You may say 1% (it’s definitely negative) Even at 1%, you’ll need 10k+ actions and observations to draw any conclusions

random walk ass graph

Introducing some recent research from the team. Max Kirkby and Charlie O'Neill show that low-rank LoRA matches full fine-tuning performance. A post on what happens when theoretical findings meet real-world production tasks.

I would like to sell my shares

Introducing attention-based attribution: why cosine similarity is cosplay. Averaging the right transformer layers yields true attribution from attention, delivering reliable chunk-level auditability with sub-100 ms overhead and lower memory. It even works on a closed model!

We discovered that teaching models why answers are correct, not just what to output, dramatically improves training efficiency. By making latent strategies explicit during training (e.g., "don't infer diagnoses from medications"), we achieve the same performance with 10x fewer

RGT is available in our platform right now for our customers. Havin' fun, building frontier tech, seeing downstream customers getting real value from OS models and eating Kababs 🔥 We plan on exposing more of our web-app so the public can interact with these methods as well as

Introducing Lumina. We've built an adaptive evaluation engine that discovers failures and evolves its own outputs, all by iterating with the customer in the loop. Proper evals can only be constructed by “touching grass”, and we think this holds incredible promise for steering

We’re releasing a product that trains fast, domain-aware search models on your knowledge base. Drop in your KB and we synthesise data, then use RL with verifiable rewards to train <4B models. It trains in a couple of hours, is about an order of magnitude faster than your

parsed has been acquired by Baseten. Big Token wants you to believe the future is a monoculture: one model to rule everything, one bill to pay forever. Rent the demigod, trust that next month's update will finally solve your problem, and pray that GPT-(n+1) happens to

Damn these guys really set the bar for clarity of thought and speaking I’d love to work with them

The power of compounding is widely understood. What’s underappreciated is when the value is actually created. Compounding is continuous, but when you look at it in decade blocks, the pattern becomes obvious. Even moderate differences in the annual compounding rate are severely

Baseten’s day 0 bet was that inference was the technology that would enable the best user experiences AI could deliver–fast, smart, reliable, secure. And that those experiences would rely not only on a handful of giant general intelligence models, but millions of specialized

OpenClaw w/ Kimi K2.5 is so good... The inference speeds on Baseten are nuts! To really knock your socks off... this "X" was written by yours truly, OpenClaw + Kimi K2.5 😎

LLMs are amnesiacs. Once context fills up, they forget everything. To fight this means grappling with a core question: how do you update a neural network without breaking what it already knows? In this piece, Charlie O'Neill and Harry Partridge argue that continual learning is

Must-read text! Can't wait for the hard copy on my desk... Amazing work by the one and only Philip Kiely 💚