4 min read
OpenAI Codex Sites turns a prompt into…
Codex Sites builds, deploys, and hosts lightweight web apps from plain English. Here's who gets access, what actually ships, and when I'd use it on client work.
Blog 385 posts
Notes from the trenches on AI engineering, LLM apps, and the full-stack work that holds it all together.
3 min read
Surya OCR 2 packs document parsing into…
Datalab's Surya 2 scores 83.3% on olmOCR-bench with a 650M VLM. One model for OCR, layout, tables, and reading order. Runs on GPU or Apple Silicon.
4 min read
JetBrains Mellum2: a 12B coding MoE that…
Mellum2 is JetBrains' open 12B MoE with 2.5B active parameters per token, 131K context, and Apache 2.0 weights. Here is when it beats bigger dense models for routing, RAG, and agent sub-calls.
4 min read
Life-Harness: 88.5% agent gains without retraining the…
Life-Harness adapts the runtime wrapper around frozen LLM agents, not model weights. Across 18 backbones it reports 88.5% average relative lift. Here is what that means for production harness design.
4 min read
tau0-WM: one world model that acts and…
tau0-WM unifies video prediction, action generation, and candidate scoring in one 5.5B open model trained on 27,300 hours of robot and human video. Here is how test-time imagination changes manipulation policy design.
7 min read
Codex Computer Use on Windows: foreground desktop…
OpenAI shipped Codex Computer Use on Windows with ChatGPT mobile remote control. Here is what foreground takeover means for testing, security, and how I would wire it into a real agent workflow.
6 min read
Codex computer use on Windows: what changes…
OpenAI shipped Codex computer use on Windows with mobile steering. Here is what foreground takeover means for QA, privacy, and how I would wire it into a real dev loop.
5 min read
Cursor Auto-review: fewer approval prompts without going…
Cursor 3.6 shipped Auto-review with a three-stage filter and a classifier subagent. Here is how it cuts terminal prompts by roughly 84% and what I configure on client machines.