Maaz ‎

Maaz ‎Sep 26, 2026

another update for Agent Orchestrator. ao is accepted into Vercel’s Open Source Program. big thanks to Guillermo Rauch for personally helping me with this. Try Orchestrator.inc

Maaz ‎Sep 28, 2026

AO Mobile v2.0.0 is out now. You can orchestrate Agent Orchestrator from your phone > dictate prompts using voice input > choose from 28 agent harnesses and select the model that fits your task > switch between Terminal UI and Chat UI at any time > delegate tasks to agents through the orchestrator > get notified when judgment is needed > add images and files to prompts > send, queue, promote, steer or cancel messages > monitor workers by state - needs you, review, ready, building or validating > approve actions and answer agent questions > open terminals, worktree shells and live previews > track PRs, CI, reviews and merge blockers Looking good in your pocket. - https://lnkd.in/g2upiaBv (App Store) - https://lnkd.in/gaB5H5Dm (Play Store) gg Prasad Ware

Maaz ‎Oct 14, 2026

"this would have taken a whole team of developers a year or two." Stephen Toub, a senior engineer at Microsoft, wrote that about the work he just finished alone in 14 weeks. he ported the GitHub Copilot runtime from TypeScript to Rust. May 12 to Aug 21. 128 PRs into main. 135 releases. 832,378 lines of production Rust, 468,689 lines of tests, ~$120K in tokens. about 3 weeks of his own actual work. the few things that broke got fixed within a day. he typed ~2,600 messages to the agents. 63% were review, pushing back on test results, challenging design decisions, or telling an agent it stopped too early. he was not writing Rust. he was running a loop. nobody read 800K lines. reviewer agents on three different models compared old TS against new Rust until every pass came back clean. humans only looked at architecture, API contracts and risk. and here's the thing about that first sentence. a team, a year or two. that project never gets approved. it loses to feature work every time. it exists now only because it became cheap enough to not need a team. the uncomfortable part is that almost nobody works like this. Jellyfish looked at ~19M rows of Claude Code telemetry. 84% of users run one or two agents at a time. 2.5% of companies run a fleet. OpenAI measured Codex usage in June. outside OpenAI, two thirds of users never run sessions in parallel. inside, 28.6% peaked at 5 or more agents at once. same model. different workflow. and 128 PRs in a quarter isn't the ceiling. Lauren Tan (poteto) at Cursor shipped 2,500 PRs last month. one coordinator agent, cloud workers doing the work, a swarm of verifier agents fuzzing every PR, agents merging their own code while she sleeps. she reviews the commit log in the morning. her words, the limit is no longer the model, it's the human standing between the agent and everything it needs. so the gap isn't model access. it's the loop. isolated workspaces, CI and review feedback routed back to the agent that owns the branch, multi-model review, one place to see what needs you. they all hand-built theirs. we built that exact loop, that's what Orchestrator.inc is. models can already talk to each other and fix their own CI. the part nobody had solved is where the human sits. the thesis isn't full autonomy, it's optimized human in the loop. prateek k runs 10 to 20 agents at once through it. the bottleneck was never the model, it was switching tabs to see which agent is doing what, and copy pasting CI failures into terminals like 99% of engineers still do, you just need an orchestrator to push agent limits at this point! a few years ago this was one engineer, one PR a week if you were good. now it's 128 PRs in a quarter at Microsoft, 2,500 a month at Cursor, because the engineer stopped being the one who types. the question isn't whether you use agents. it's how many you can trust at once. nobody knows the ceiling yet because almost nobody has tried. Full write-up by github in comments.