TVs are still controlled like it’s 2005.
Buttons. Plastic remotes. Searching letter by letter.
I’m building Drift: an AI interface that lets you control your TV without a remote.
This is my first log of DriftV1 from this week.
Hardware coming soon!
@mitchellh If GitHub was built for the 2010 workflow, I’m curious what a code forge designed for AI collaboration from day one would look like.
That one invites conversation instead of declaring.
@ajambrosino The fact that roles now include agent orchestration says a lot about where things are headed.
Feels like we’re moving from building tools to building operators.
@ycombinator@bcherny Claude Code feels less like a coding tool and more like a shift in how we spend attention as devs.
If AI handles the repetitive parts, humans get to think bigger.
@NSSpain@steipete Most people wait for permission when realistically the internet already removed most gatekeepers. The real bottleneck now is just starting.
@steipete@openclaw The hardest part of open source agent projects won’t be code quality. It will be clarity of direction. A strong vision doc is what prevents fragmentation when momentum hits.
We’re entering the “AI execution layer” era.
Not just models.
Not just chatbots.
Agents that act.
Infra that scales.
Robots that execute.
The next wave isn’t smarter AI.
It’s AI that does things.
3 patterns I’m seeing:
• Agents are going mainstream (OpenClaw, SurrealDB, Vapi)
• AI infrastructure money is accelerating (Neysa, Anthropic)
• Robotics is re-entering the capital cycle (Apptronik)
This isn’t random hype.
It’s capital + distribution + execution converging.