3 months ago I wrote that the nature of all knowledge work had fundamentally changed.
The shift, in one line: research and execution moved to agents. Judgment and verification became the whole job.
A lot of people were skeptical, and for good reason. How can you understand the work if you weren't the one executing it?
Turns out agents are pretty good at explaining their work visually, through HTML pages and SVG diagrams. So the job changes. You understand the system, decide on the goals, and verify if they were actually achieved.
This also unlocks something big: the ability to parallelize work, paired with better planning that the agents drive themselves. Which brings up a real possibility. You start getting work done at the speed of your thought and judgment.
For the last 3 months I've been building the interfaces and the product for exactly this, optimizing every feedback loop across the SDLC.
It's community-driven and open source, with ⭐7.4K GitHub stars so far. Today, all of my work happens through
@aoagents . It's been a multifold productivity unlock on top of Claude Code / Codex.
If you're curious to know what makes this possible, or just eager to use this open source project yourself, I'd love to chat.