We asked GPT-6 Astra to edit Qwen3.5-0.8B weights to improve math.
It directly changed only 0.016% of weights, fixing 0.8B's reasoning loops & improving math up to 6x and even instruction following 3x, without gradients or training.
We call this Weight-Space Programming: frontier LLMs can now natively read/write neural network weights! 🧵
Blog: rikiyat.github.io/blog/weigh…
Agents should get better with experience. We teamed up with @tolokaAI to measure that.
Toloka evaluated xmemory on closed datasets. No peeking at the data. No tuning to the benchmark. Just learning from failures and measuring what improves.
📈 The results: 26% lower token costs when xmemory replaced document search as the policy store. And adding lessons from past runs lifted scores by 4.8 percentage points on held-out tasks.
💡 It means with memory, a mid-size open-weight model scored 0.75 – beating Sonnet 5 at 0.68 and GPT-5.6 Terra at 0.69 without memory.
This experiment is just the beginning of what memory can unlock. We see even bigger gains on longer-horizon tasks, where every lesson has more chances to pay off.
Read our blog post below!
We launched 1M+ cloud agents. Now we’re sharing how you can build them for your company.
We spent a year battle-testing flexible components for diverse use cases and security needs.
Unreal Agent is one of those components. Today, we’re discussing the broader stack 🧵:
An Uber-style app with iOS 27 Siri support, fully built in Rork
Ask "where's my driver?" and Siri tells you the driver, the car and the plate
Get your app ready for iOS 27 in Rork
Unreal Agent - open-source harness with frontier cost efficiency on long-horizon tasks
48% cheaper than Codex+Sol 6.1 xhigh on SWE Marathon while maintaining performance
Now with compaction, TUI and Anthropic support. More in 🧵
Introducing Unreal Agent:
An open-source harness with state-of-the-art cost efficiency
39% cheaper than Codex+Astra on Terminal-Bench 4.0 while maintaining performance
meet @mostik_ai!
what happens when you put 12 PhDs in one room for four months? first place on the ARC-AGI leaderboard, which I can't say much about while the competition is still running. and this, which I can.
everyone's arguing about whether open models will catch up to frontier models. we think it's the wrong question. here's the one we pose: why does a frontier model have to generate your answer at all, when the only thing you need from it is the reasoning?
we do this by enabling models to communicate in latent space. through our protocol, hidden states pass straight from a frontier model into a small one running on your infrastructure -- no text between them, and neither model is fine-tuned. two models from different families, sharing reasoning, both left untouched.
how do we know it works? we tested it on a setup where a 753B model reads the problem, and a 4B edge-class model writes the answer. with this approach, we get results 80% as accurate as the frontier model, but at 20x faster performance.
we're committed to preventing frontier model lock-in and are already partnering with inference providers to accelerate open-weight adoption. we've done this between 15 of us, in four months, 12 PhDs and a Fields medalist, backed by @generalcatalyst
WIRED has the first external account of the company and the work: wired.com/story/russian-star…
full writeup, the setup, and all the numbers: mostik.ai/read-more
I have been looking into GADTs in OCaml for a while and found some interesting real-life use cases for them. Decided to focus on one, "OCaml GADTs for Authentication Tokens".
Huge thanks to @ChShersh for reviewing it!
Full blogpost below: