Sad that Conductor got bought by Meta. It was a polished product that brought a lot of small delights to my daily work. My two favorites are: the ChooChoo sound , giving each git worktree a city name. It shows characters and care from the team.
Large-scale RL has a problem: weight updates from trainers to samplers are slow and expensive
We’re sharing weight teleportation: faster weight updates than existing intra-cluster techniques with internet speeds, enabling samplers anywhere (maybe even space!)
Weight teleportation is so efficient (<0.23s sampler pause time, ~3s non-blocking compression + prepare), the bottleneck ends up being the GPU hardware’s decompression ASIC
We unlock these gains by using GPUs for massively parallel compression of model weight changes, engineering a heterogeneous decompression and patching pipeline, along with other optimizations and overlaps. These techniques are co-designed with hardware-native features in NVIDIA’s Blackwell generation of chips
We’ve open-sourced weight teleportation, so everyone can use it today
Introducing Ampersand Tutor (&Tutor) - we trained a model to learn like people, then trained &Tutor to teach it
On a widely-used tutor model eval, &Tutor performed best, showing how better models of people can help build AI for humans
Available through our API - try it!
What a difference one year makes
Last Halloween, it is all about K-pop demon hunters and hundreds of coding agents
This Halloween, there is clear winners in coding agents (Claude/Codex/Devin). The new battle ground seems to be hundreds of personal agents (Muse/Instinct/Dot/GrokBot/ClaudeDuet). I suspect this requires very different company DNAs so it is still everyone’s game?! Guess we will know by the next Halloween.
We accelerated one of the most optimized NVFP4 decode kernels while also making it more accurate
Fewer bits typically mean more speed but less accuracy. We did the opposite: more bits, but up to 10% faster generation
Already used in Ampersand’s RL and Boba serving
Introducing Boba - the social intelligence that understands your goals and your community. Text Boba today or message it online at https://t.co/ZrF5yBrp2g
Year 2000, I ordered a book from Amazon and it showed up.
Year 2026, I ordered a boba from DoorDash using personal agent and it showed up.
Both are mind blowing
No one changes the world alone
Introducing Ampersand - the first model built for us. Ampersand is a collaborator trained to help people achieve their individual and collective goals. In multi-month simulations, users collaborating with Ampersand accomplished more than with Fable and other frontier models
Available to everyone through our API
Comparing main chat + side chat UI in Codex and Claude Code:
* one of them has the input box aligned
* the other doesn't
I prefer the aligned version, you?
“OpenAI enterprise customers can now use existing OpenAI commitments for open models served by Baseten within Codex or via the Responses API.”
This went even further than I thought - https://t.co/3hEEbqH01O
If things play out well, this might meaningfully accelerate tokens consumed per day (TPD) through OpenAI and hopefully improves margin as well