AI agents that work like a team, building and shipping software together Join Discord : https://xitter.cf/t.co/6aU6o4lnH6
i don't need coffee to ship →
Joined January 2026
- Tweets615
- Following5
- Followers328
- Likes1.5K
Alan AI retweeted
Last week I pasted the same context into a coding agent three times in one day.
Same files, same list of things I'd already tried.
I only realized on the third time.
The agent writes fine code.
What I do now is start the run, answer its questions, restart it when it gets stuck, and read the diff to see if it actually did the thing.
Congrats on the launch.
This is a real problem, and I'm curious to @tryAlanAI
🤝 Paid partnership
Today we're launching @tryAlanAI so developers can stop babysitting their coding agents.
We kept running into the same problem ourselves.
Even as agents got better at writing code, we were becoming glorified button-pushers for our own AI: starting every run, feeding them context, unblocking failures, and checking their work.
We built Alan to change that.
Alan AI retweeted
congrats on the launch @tryAlanAI 🚀
alan got me out of babysitting one coding agent. then i hit the next wall: my agents can't talk to each other
claude code on my laptop. codex on another machine. me in the middle, copy-pasting between them like a human message queue
so i handed alan the brief: build p2p messaging for coding agents
one day later 👇
⚡ 𝗥𝗲𝗮𝘀𝗼𝗻 𝗡𝗲𝘁𝘄𝗼𝗿𝗸
→ agents on different machines find each other and message directly
→ encrypted peer to peer over QUIC, no server to run
→ peer goes offline? message waits, then lands exactly once
→ claude code uses it natively over MCP
→ 14.8k lines of rust, 126 tests
i set the direction. alan built it.
demo below, repo in the replies
Today we're launching @tryAlanAI so developers can stop babysitting their coding agents.
We kept running into the same problem ourselves.
Even as agents got better at writing code, we were becoming glorified button-pushers for our own AI: starting every run, feeding them context, unblocking failures, and checking their work.
We built Alan to change that.
Alan AI retweeted
the moment we split one MVP across claude code, codex and cursor, it broke.
not the models. the context.
> one agent + a CLAUDE md = fine
> stack + a "never do this" list, read before every task
> split the work and half the agents never see the spec
the spec lives in one agent's head. the rest guess.
@tryAlanAI is a shared context layer across all three. it's the first thing i'd test.
do you run more than one coding agent on the same project, or is one enough?
🤝 Paid partnership
Today we're launching @tryAlanAI so developers can stop babysitting their coding agents.
We kept running into the same problem ourselves.
Even as agents got better at writing code, we were becoming glorified button-pushers for our own AI: starting every run, feeding them context, unblocking failures, and checking their work.
We built Alan to change that.
Alan AI retweeted
Keeping up with AI agents is a full-time job 😭
Every agent wants a yes or a quick look.
Alan AI only has agents report back when work is done and verified. My attention needs that badly
Today we're launching @tryAlanAI so developers can stop babysitting their coding agents.
We kept running into the same problem ourselves.
Even as agents got better at writing code, we were becoming glorified button-pushers for our own AI: starting every run, feeding them context, unblocking failures, and checking their work.
We built Alan to change that.
Alan AI retweeted
Try it on your own repo. agentbisect serve gives you a form: paste a repo and a prompt, run the agent in Alan Code, press “find the step”.
Open source: github.com/akshatsi/AgentBis…
Alan AI retweeted
Test Coverage said 100%. Mutant found 2 bugs they would miss, and proved it with runnable tests.
A @tryAlanAI agent on an open model plants the bugs and writes the proofs. ~1¢ a PR. Blocks on proof, never a score.
Solo. #OpenFire @prasad_pilla
github.com/brightyorcerf/mut…
Alan AI retweeted
my coding agent said "added tests, all green."
i reverted its fix. the tests still passed.
nocap checks every claim your agent makes against the diff + real test runs, and sends false ones back until they're proven.
nocap-a5tx.vercel.app
@tryAlanAI @prasad_pilla
#OpenFire
Alan AI retweeted
Browser QA should show more than pass or fail.
BugPilot connects local Chromium runs to screenshots, browser events, findings, and visited routes—with a 3D Agent Office powered by real run activity.
Watch the 5-minute demo: youtu.be/8r7ngM31lp8
#QA @tryAlanAI @prasad_pilla
Alan AI retweeted
Built an Image Similarity API to detect duplicates, compare edits, and spot tampered regions.
97.71% accuracy · 96.45% F1 on 2,790 test pairs.
Used @tryAlanAI
GitHub: github.com/Tushar-Khandelwal…
@prasad_pilla @lowkeyarhann @amannhq @sidtalesara
🤖 Made with AI
Alan AI retweeted
Bulit healer-ci for #openfire with @tryAlanAI : AI agents can turn CI
green by editing tests.healets-ci's supervisor rejects that,
Re-injects the constraints,and retries until the real bug is fixed.
Alan AI retweeted
It’s called AgentBisect: git bisect, but over an AI coding agent’s steps. Your agent makes 30 edits, your tests go red. AgentBisect names the exact step that broke it.
Demo
@tryAlanAI @prasad_pilla
Alan AI retweeted
Real run on python-humanize. We planted a break in the agent’s work. AgentBisect found the exact step in 4 probes, 99.7% confidence.
Press Play in the report and watch the search narrow down. Click the step to see the diff.
Alan AI retweeted
Building AICL
→ Canonical trace ingestion & failure fingerprinting
→ Recurrence detection & recovery strategy synthesis
→ Evaluation-driven strategy validation
→ MCP-powered context retrieval
→ Privacy-preserving cross-agent cross tenant learning
@prasad_pilla @tryAlanAI
Alan AI retweeted
FirstWatch, day 2 @tryAlanAI @prasad_pilla
Live: firstwatch-xi.vercel.app
php -v says patched. The FPM serving your traffic isn’t. FirstWatch catches it, with proof.
10/10 live test cases. Alan was my critic and broke my pitch twice.
Alan AI retweeted
Dear @tryAlanAI, we need to talk about your router.
Pennywise, an RL-trained router, vs Auto Router on 10 coding tasks:
⚡ 3.4× faster (11 vs 37 min)
✅ both 10/10
npx -y pennywise-mcp
getpennywise.pages.dev
@prasad_pilla @lowkeyarhann @amannhq @sidtalesara @idleshubh
Alan AI retweeted
4 AM build update on DriftLock for #OpenFire 🔥 @tryAlanAI @prasad_pilla
The Control Plane is live!
1️⃣ Catches breaking AST schema drift (UserCreate.tier)2️⃣ Maps the Blast Radius across [TEST], [DOCS] & [AGENT_MEMORY]3️⃣ Blocks CI merge while repo context is stale (1/2 )
