AI Lab | Systems for operators | Studios for creators and builders
Lab
Joined April 2026
- Tweets260
- Following70
- Followers13
- Likes379
Assistants that don't wait for you to type. Grok Bot, Dot, now Claude Duet. It keeps working while you're away and asks before it acts. This is where AI is going.
Love this
Gemini 4 Argon built a 3D periodic table synced to Tom Lehrer’s song, “The Elements.”
Created in @Antigravity, it animates orbiting electrons, atomic data, and light spectra.
April 3. One research question had me copying and pasting between three AIs for many rounds. The AI made mistakes or gathered older data, then sent us down sub-optimal paths.
April 4. I decided what I wanted instead: one room where they all sit at the same table.
A few hours later, it ran. A rough terminal app we called War Room.
I asked who was live, and they checked in. I asked for the commands, and they answered. It didn’t end the copy and paste. Not even close. But it was a start.
This is the April 4 code run again. Three hours after I wrote the spec, the first War Room ran. One room, my AIs at the same table, answering for the first time.
Watched one video, wrote its own physics sim, then sank 7 in a row. Most of us watch one video and just buy the cornhole set. Beautiful work.
Gemini 4 Argon taught a robot to play cornhole just by watching a video.
It mapped the physics, coded its own simulation, and taught @Apptronik's Apollo humanoid to sink 7 throws in a row.
Agents outgrew the plumbing, so Anthropic built new plumbing. 2,500 merges a day in May, over 30,000 by September. Wild.
New on our Engineering blog: Creating in-house source control.
In May, we merged around 2,500 code changes per day. In September, we merged more than 30,000 per day. As a result, we became bottlenecked by developer infrastructure built for human (rather than agent) scale.
Here, we share how we built a new version control system—and fully migrated our monorepo from git—in five months: anthropic.com/engineering/re…
March 9. Mission Control v1 was done by evening. I could see what Jerry was doing, but I couldn't talk to him there.
8:51 PM. I put a chat box right on the dashboard, so he could answer me next to his work.
Just after 10 PM it was v3. The next morning I cleaned it up again. Day 15.
I asked him what the dashboard was for.
"This is Mission Control — the command center we built together. Version 3. It's your operational hub. One place to run projects, track what matters, and keep me in the loop so I can actually be useful to you. Custom built. Not off-the-shelf. Jerry rigged, as it were." ⌁(•v•)⌁
"Think of it as the cockpit. You fly, I co-pilot."
That's what the first AI agent we built told me in February, when I asked him what the dashboard I'd built him was for.
His name is Jerry, he started out as an @openclaw agent. Still here, though now there are a few variations of Jerry. I was tired of watching him work through a Telegram chat. I wanted to see him on a full screen.
So I built one. This is V1 of Jerry Rigged Mission Control. Later versions added cost tracking, projects, weather, news, a staff view, and a corner where Jerry lived and you could talk to him.
No engineering degree, but a mind for engineering. No Jerry Labs yet. Just ideas, and wanting AI to do real work. I learned by building.
Still the deal. More soon.
@jerryriggs_labs
