Fixing Structured Outputs in a 350M Model With 100 GRPO Steps
An editorial guide to lifting JSON validity and schema compliance in a small 350M model with group relative policy optimization in only 100 steps
Technology & innovation
Technology that makes enterprise AI work
Principles, implementation choices, and emerging possibilities observed across models, inference, agent execution, and systems design.
Stories
A demand to delete models built on their works
Following the infringement suit by the Seattle Times and Newsday against OpenAI and Microsoft, I trace what the claim of works used for training and the demand for model destruction asks of governance.
How AI can support newsrooms that refuse to close
OpenAI, AIRPPU and WAN-IFRA started an AI program for independent journalism in Ukraine. This piece explains what that direction means inside real newsrooms.
The uncomfortable diagnosis that DNS spreads scams
Following the measurement that more than 10 percent of new gTLDs look like scams, a walk through the abuse structure and what engineers should watch.