Agentic React · BenchmarkPost 06
Structured UI context used 53.6% fewer tokens
A reproducible Agentic React benchmark comparing screenshot input with structured UI context for React source location, including token use, cost, and search behavior.
- Token economyContext on demand
- Agent memoryRecall, not preload
- Open ecosystemShared building blocks
LatestPost 06
Ideas in practice
Current blogs
Agent architecturePost 04
Memory should be a skill, not a file loaded at startup
A blog post on replacing static, prompt-loaded agent memory with granular memories discovered and recalled when the work calls for them.
Builder practicePost 03
From product intent to software a team can review
A blog post on preserving the reason behind a product while turning it into concrete screens, flows, and technical choices.
Open-source practicePost 02
Open source where it creates leverage
A blog post on choosing open-source boundaries around inspectability, interoperability, trust, and shared technical progress.
Product practicePost 01
Useful AI becomes real when the work becomes visible
A blog post on turning an abstract promise about AI into a concrete surface people can inspect, question, and improve.
