Michael Rouveure

🚀 Building (Openclaw Directory for agents), (VC decks), | 🤝 AI + no-code for funds (

[Michael Rouveure @MichaelRouveure]

Read more.

full digest at https://t.co/ZqwAGZTiC1 — what shipped, what flipped, what the practitioner take is. also: my codex session for the junglebee rebuild got killed by the quota bug last week. that story continues.

[Michael Rouveure @MichaelRouveure]

last monday i came back to my desk and had no claude fable 5. not a usage issue. the US government pulled it overnight — no warning, effective immediately. i build on claude daily. this hit differently than a news story.

[Michael Rouveure @MichaelRouveure]

what i'm watching: both gpt-5.6 sol and claude mythos 5 launched with government-coordinated access gates. not "buy a subscription." "trusted partner, apply through your rep." frontier models as a managed institutional access program is new.

[Michael Rouveure @MichaelRouveure]

Rebuilt Junglebee this April with Replit Agent 4. 9 years of code, millions in bookings/year. First pass at 8-10 hours wasn't usable. The back-and-forth, section by section, is what actually shipped. Closer to a very fast junior engineer than set-and-forget.

[Aaron Levie @levie]

Coding is basically the pinnacle of what you could reasonably automate with AI, and yet we still need human engineers to oversee agents for them to be effective.

The AI models are trained on an incredible amount of sophisticated code. The users are highly technical and can use the latest tools quickly. The work is “verifiable” because you can test an app. The outcomes are often removed from the quality of the code (you can have sloppy code but the app can still work). And the context for the agent is often already digitized and sitting in the codebase.

That’s an incredible amount of benefits that AI coding agents get to work with. Some of those apply to knowledge work, but most don’t in areas where the work needs to be fully reviewed to be useful, or where data isn’t as abundantly digitized. This makes the job for agents in knowledge work more complicated.

So if with all of that, engineers still remain in very high demand, the risks are going to be less than what’s perceived for other areas of knowledge work. Agents will let people do far more than they did before, but the people don’t go away.

[Michael Rouveure @MichaelRouveure]

Full digest: what shipped, what flipped, the 7 must-reads, what we're watching next week →

https://t.co/2Q2eGFjTvK

Happy to compare notes — what's changing in your stack this week?

[Michael Rouveure @MichaelRouveure]

Weekly AI/agent-infra digest 🧵

This was the week the AI harness became the moat.

One Anthropic S-1. Six shipped harnesses. One Uber spending cap. Every player above the model agreed on it at the same time.

[Michael Rouveure @MichaelRouveure]

.@garrytan named the pattern on Monday: "The AI Harness Wars of 2027."

173K views. Follow-up the next day: "after that comes the Frontier Labs vs All Software Companies war of 2028."

Two years of AI strategy sketched in two tweets.

[Michael Rouveure @MichaelRouveure]

Lord this made me laugh so much🤣🤣🤣

[Peter Dedene @dedene]

POV: you're still using GitHub Copilot after June 1st, 2026

[Michael Rouveure @MichaelRouveure]

Same pattern in the Black Matter swarm — Editor agent reviews every Lab Entry draft against the voice ban-list before flipping status to 'pending review'. Catches AI-tells the synthesizer missed. Runs in a separate Claude call so the critique doesn't get coopted by the draft.

[Peter Steinberger 🦞 @steipete]

Been teaching codex to be my QA assistant. For every commit it creates a user-test scenario and uses webVNC (crabbox), computer/browser use (peekaboo/mcporter) to test OpenClaw like a user/QA person would.

This runs in the background and opens PRs with fixes.