This blog collects longer-form notes on AI agents, software engineering, developer tools, and building real systems.
How claims are labeled
Analysis articles here label their important claims so you can tell sourced facts from my opinions:
- Documented: stated in official documentation or source material, linked next to the claim.
- Verified: checked directly in source code or with an experiment described well enough to rerun.
- Observed: something I noticed in use, not a controlled test.
- Opinion: my interpretation.
- Proposal: an idea or setup you should evaluate for yourself.
Tools change quickly, so articles also state the versions and dates they were checked against. If something is out of date or wrong, please let me know.
Where to start
The first in-depth article looks at Pi Agent’s architecture and its permission model. You can browse everything in the archive or subscribe to the RSS feed.