Two Yale CS seniors built a coding agent in their dorm room and shipped it to #5 on Product Hunt with 236 upvotes. Their bet: the models are already fast enough, the loop wrapped around them isn’t.
What it actually is
Bullet is a CLI coding agent — same slot as Claude Code and Codex, one selling point: latency. It claims 30-60% faster than both. Three tricks do the work. It picks the model and reasoning level per prompt instead of running everything at max effort. It fires independent searches, file reads, and shell commands in parallel while keeping dependent edits sequential. And it uses targeted code search rather than shoving your whole repo through an embedding index.
The numbers: 479/500 on SWE-bench Verified, 119 seconds average per task, 16% fewer round trips and 27% lower cost than the baseline harness. The founders are ex-AppLovin and ex-Citadel, and they’ll tell you themselves that SWE-bench is saturated.
Bring your own model
No new subscription. Bullet plugs into the Claude Code or Codex plan you already pay for, or your raw API keys — Claude, Codex, Grok, OpenCode — or a model running on your own machine. The product itself is free right now.
Why it matters
Coding agents spent two years competing on capability. Bullet is the first to compete purely on speed and still land top-three on the benchmark. That’s the tell: the harness, not the model, is where the remaining seconds live.
You Might Also Like
- Jupid hit 1 on Product Hunt During tax Season 507 Upvotes for an ai Accountant Built on Claude Code
- Zcode Zhipu z ai glm 5 2 Coding Agent hit hns 1 Spot Chinas Open Answer to Claude Code
- Blaxel Agent Drive Gives Codex and Claude Code one Filesystem to Share
- Opencomputer Agent Deploy Turns one Prompt Into a Live Agent url hit 4 on Product Hunt
- Claude Code Remote Control Just Turned my Phone Into a Coding Terminal and im Weirdly Into it
Top AI Product 
Leave a comment