I guess however this is a harness and it needs to connect to the API?
Dear YC, please force the startups to dedicate some of the 500k funding for standard web design.
Cmd+Option+I > Console > 'allow pasting'
const onboarding = document.querySelector('#onboarding'); const app = document.querySelector('#app'); onboarding.style.setProperty('display', 'none', 'important'); app.inert = false; app.removeAttribute('aria-hidden'); document.querySelector('#prompt')?.focus();
Also, warning:
# Share chats with Bullet — helps us improve model routing and answer quality
Enabled by default.
Is it even hidden when your AI ends up tagging it with `aria-label="Hidden secret code"`? Lol. Fun mini game.
I recently discovered this same lesson the hard way. I was trying to get a multi-agent system I was building to improve upon GPQA Diamond scores (system here: http://pellmell.ai). No matter how hard I tried, I could not get any lift. When Fable 5 dropped, it also did not improve upon Opus, and I realized my mistake. The benchmark was saturated!
Now, looking at the result here, I see a similar pattern. Fable is not better than Opus, and the score is ~95%. Notably, this post omits which subagent is being used. Why? An intellectually honest way to tell if this thing really works would be to run that agent and report its score and cost as well.
Going back to my GPQA Diamond lesson, you can see here how a saturated leaderboard behaves https://artificialanalysis.ai/evaluations/gpqa-diamond. Fable gets 92.6% for $0.22 per task while several models score higher for $0.01. I could easily publish a router that “enhances Fable on GPQA Diamond” showing improved score for lower cost, just by implementing a router that picks the model at random!
"The bullets" are not any different than other new agent projects. New agent projects pop up and die constantly and it's not from a lack of funding.
Looks like "Techno music design of the 90s" or similar :-D
On the saturation point, we agree that a 95.8% result on a mature benchmark isn't the main proof, which is why we're currently running against harder, less saturated benchmarks like Terminal-Bench, CursorBench, and SlopCodeBench (going to publish results on these hard benchmarks shortly). Apart from current user experiences, that will show the value of our harness.
Hell, I could publish better score at 87.5% time reduction by having the router always pick Flash!
Bulletproof.
B–001 / INTRODUCTION
MACOS · LINUX · PRIVATE BETA
FREE ACCESS
SCROLL TO EXPLORE
01
THE PROBLEM / LATENCY
We were burning hours waiting on agent runs. Not because the models were incapable. The machinery around them was simply heavier than it needed to be.
02
THE BULLET SYSTEM
Same model → tools → results pattern.
A tighter loop around it.
PROTOCOL 01ROUTE / ESCALATE
TASK IN
FASTSTRAIGHTFORWARD
DEEPCOMPLEX
REASONHARDEST
Right moment.
Route straightforward work to fast models. Escalate only when the mission demands it.
PROTOCOL 02SEARCH / ACQUIRE
⌕auth middleware
01src/router.ts94.2
02src/auth.ts88.7
REPOSITORY CONTEXT08%
Ignore the noise.
Targeted search and file reads find relevant code without embedding the whole repo.
PROTOCOL 03PARALLEL EXECUTION
SEARCHREADVERIFY
ELAPSED00:02.313 OPERATIONS / SIMULTANEOUS
Independent tool calls execute in parallel. Duplicate calls and stuck loops are intercepted before they waste another second.
NO DEAD TIME / NO DEAD TIME / NO DEAD TIME /
BUILT FORMOMENTUMNOT MAGIC
03
FIELD TEST / 001
Real prompt. Real repository.
No cinematic shortcuts.
04
GET BULLET / CLI
Same router, tools, and agent loop.
Now it lives in your shell.
zsh
$ npm install -g @trybullet/cli
$ bullet
macOS & Linux · Node 18+ · free, no key required to start
05
FOUNDERS' NOTE
At our company, we were burning hours waiting on agent runs. We were building with Claude Code every day, watching capable models move through needlessly slow machinery.
So Bullet began as a passion project: route simple work faster, read only what matters, run independent tools together, and stop loops before they spiral.
We use it internally now. It saved us a headache. We thought it might save you one too.
DON'T LET YOUR CODE
HOLD YOU HOSTAGE.
PRIVATE BETA / FREE ACCESS
Free to use. No subscriptions.
Just a faster way to ship.