Most agents just start working. You never see the plan. You never get to say no.
I want the plan first. What it will do. What it can change. Then go, edit, or I'll do it myself.
I believe everyone should have access to the METAVERSE, and I wrote a long piece about Meta's philosophy and values for building a positive future for everyone. https://t.co/IlKDgNuMOH
I believe everyone should have access to superintelligence, and I wrote a long piece about Meta's philosophy and values for building a positive future for everyone. https://t.co/2ZoNZXZ39T
@ash3003 Agree! And I am building this.
The framework: https://t.co/ZplVIA09pJ
Been working with eve (vercel) mostly and set up a system around trust for agents https://t.co/Z80Ldk8cQJ (early stages)
@rauchg Sandboxing contains the blast radius. What still gets underbuilt is the supervision layer: Intent Preview before the agent acts, an Autonomy Dial for how far it can go, a Progress Ledger while it runs.
Patterns for that layer: https://t.co/rVTKEN3UjZ
Each pattern still ships anatomy, do/avoid, limitations, interactive mockup, production screenshots with critique, and llms.txt for coding agents.
https://t.co/ZplVIA09pJ
Agentic UX got a content pass this week — not just plumbing.
What's new on https://t.co/ZplVIA09pJ:
→ Hero reframed around the actual offer: a free, open library of 11 lifecycle patterns for agents that act on your behalf — before, while, and after the run.
→ Discoverability infrastructure: per-URL lastmod from git history, so sitemap and JSON-LD reflect what actually changed — not "everything updated today" on every deploy.