Loading TechWhoKnows…
Reporting, analysis and primary-source coverage on Reinforcement Learning.

Models & Research
Hugging Face launches harness adapters, small LoRA modules that adapt any model to a specific RL evaluation harness, cutting training cost by 80% and improving pass rates on coding and reasoning benchmarks.
Clara Tan

Models & Research
RadixArk releases DaisyRose-27B, built on Qwen3.8-27B with 1,200 curated RL tasks on 144 H200 GPUs, raising Terminal-Bench 4.1 pass@1 from 5.6% to 12.0%.
Clara Tan

Models & Research
Tencent's new Hy4 flash model uses hybrid attention, LatentMoE and Engram architecture to deliver 150+ tokens per second at a fraction of predecessor cost…
Clara Tan

Models & Research
DeepMind has used AlphaProof Code to construct end-to-end mathematical proofs verifying the Signal protocol machine code directly at the ARM64 assembly le…
Clara Tan

Agents & Tools
With AWS running near a $182 billion annualised pace, leadership says inference now exceeds training revenue while capacity constraints fuel external Trai…
Clara Tan

Breaches & Threats
OpenAI restricted GPT-6 Cyber to vetted defenders under its Daybreak Red programme while deploying a fail-closed Control Room service that halts frontier …
Elena Rostova

Enterprise & Cloud
Amazon released a public Trainium inference container built with RadixArk and Annapurna Labs as AWS chief Matt Garman warned inference demand has barely s…
Jin

Agents & Tools
How Cloneable deploys autonomous field agents to capture institutional engineering knowledge and automate utility grid inspections before veteran technici…
Jin