weightythoughts.com
|
ksl
|
|
James Wang, a GP at Creative Ventures and former Google X exec, wrote a practical breakdown of building AI agents without writing code. The core argument is that the real bottleneck isn’t technical skill – it’s being precise about what you want done. He walks through three specific workflows: a language learning chat using Claude Projects, a morning briefing that pulls from Gmail and Google Calendar, and a meeting summary pipeline that spawns 38 parallel subagents via Claude Code to handle transcription and reporting. That last example is telling – useful agent work often requires chaining narrow tasks rather than trusting one long context window. The piece arrives as both Anthropic and OpenAI have been steadily pushing their “Projects” features toward exactly this kind of personal automation.
