Ship loops, not code.
The architecture of agentic feedback loops that converge and correct. That's the work now. Not the output. Not the code the system writes. The loop itself: how you steer it, how you improve it, how it changes what you're capable of building.
1,524
research papers scanned
1,736
agentic setups tested
Weekly
videos
Biannual
sitreps
> Latest Video
Click to load video
loads third-party content from youtube-nocookie.com
Why AI Coding Tools Misread Your Diagrams (The Claude Code Fix)
AI tools think in text. Diagrams, screenshots, and whiteboard photos force a translation step where information gets lost.
> Full page with takeaways and transcript →> Intelligence
> Sitrep
A ground-truth reading on agentic development, every six months.
What shifted. What didn't. What to do about it.
> Open
> SIGINT
Every claim in agentic AI checked against the primary source.
1,524 papers scored on Rigor, Transparency, Claims, and Integrity. Sortable by year, topic, and dimension. Only 4.1% meet full reproducibility standards.
> Open
> Agent Lab
Sweeping the jagged frontier for the best agentic configurations.
Same task, one variable at a time. 1,736 experiments across 11 models and 23 configuration axes. Tornado charts, efficiency frontiers, per-run breakdowns.
> Open