AI coding agents grade their own homework: "it works, it's faster." Nobody checks.
At the @createosai × @stripe hackathon I built KISS Discovery Engine: @_inception_ai's Mercury 2.5 writes code, each attempt runs in a real CreateOS microVM fork, and trusted tests decide.
How:
As promised, we have another surprise for our CreateOS × @stripe Hackathon builders.
Rules:
1️⃣ Share what you built on X or LinkedIn and tag CreateOS.
2️⃣ RT/QRT this post.
The post with the most engagement wins a special prize from our team.
Deadline: October 21, 7 PM PT
Winner will be announced on Discord.
Note: This one is for hackathon builders only.
Show us what you built 💪🏼
@jowettbrendan Building KISS Discovery Engine: an AI coding agent that can't grade its own homework. Each attempt runs in its own CreateOS microVM fork, hidden tests pick the winner, failures are kept.
It's in a most-engagement contest, a like on the thread helps: https://t.co/hfUv4Ase3l
AI coding agents grade their own homework: "it works, it's faster." Nobody checks.
At the @createosai × @stripe hackathon I built KISS Discovery Engine: @_inception_ai's Mercury 2.5 writes code, each attempt runs in a real CreateOS microVM fork, and trusted tests decide.
How:
@reasonmachines Building KISS Discovery Engine: an AI coding agent that can't grade its own homework. Each attempt runs in its own CreateOS microVM fork, hidden tests pick the winner, failures are kept.
It's in a most-engagement contest, a like on the thread helps: https://t.co/hfUv4ArGdN
AI coding agents grade their own homework: "it works, it's faster." Nobody checks.
At the @createosai × @stripe hackathon I built KISS Discovery Engine: @_inception_ai's Mercury 2.5 writes code, each attempt runs in a real CreateOS microVM fork, and trusted tests decide.
How:
@cline Building KISS Discovery Engine: an AI coding agent that can't grade its own homework. Each attempt runs in its own CreateOS microVM fork, hidden tests pick the winner, failures are kept.
It's in a most-engagement contest, a like on the thread helps: https://t.co/hfUv4ArGdN
AI coding agents grade their own homework: "it works, it's faster." Nobody checks.
At the @createosai × @stripe hackathon I built KISS Discovery Engine: @_inception_ai's Mercury 2.5 writes code, each attempt runs in a real CreateOS microVM fork, and trusted tests decide.
How:
@Docker Building KISS Discovery Engine: an AI coding agent that can't grade its own homework. Each attempt runs in its own CreateOS microVM fork, hidden tests pick the winner, failures are kept.
It's in a most-engagement contest, a like on the thread helps: https://t.co/hfUv4ArGdN
AI coding agents grade their own homework: "it works, it's faster." Nobody checks.
At the @createosai × @stripe hackathon I built KISS Discovery Engine: @_inception_ai's Mercury 2.5 writes code, each attempt runs in a real CreateOS microVM fork, and trusted tests decide.
How:
@trycua Building KISS Discovery Engine: an AI coding agent that can't grade its own homework. Each attempt runs in its own CreateOS microVM fork, hidden tests pick the winner, failures are kept.
It's in a most-engagement contest, a like on the thread helps: https://t.co/hfUv4Ase3l
AI coding agents grade their own homework: "it works, it's faster." Nobody checks.
At the @createosai × @stripe hackathon I built KISS Discovery Engine: @_inception_ai's Mercury 2.5 writes code, each attempt runs in a real CreateOS microVM fork, and trusted tests decide.
How:
@RoundtableSpace Building KISS Discovery Engine: an AI coding agent that can't grade its own homework. Each attempt runs in its own CreateOS microVM fork, hidden tests pick the winner, failures are kept.
It's in a most-engagement contest, a like on the thread helps: https://t.co/hfUv4Ase3l
AI coding agents grade their own homework: "it works, it's faster." Nobody checks.
At the @createosai × @stripe hackathon I built KISS Discovery Engine: @_inception_ai's Mercury 2.5 writes code, each attempt runs in a real CreateOS microVM fork, and trusted tests decide.
How:
@opencode Building KISS Discovery Engine: an AI coding agent that can't grade its own homework. Each attempt runs in its own CreateOS microVM fork, hidden tests pick the winner, failures are kept.
It's in a most-engagement contest, a like on the thread helps: https://t.co/hfUv4Ase3l
AI coding agents grade their own homework: "it works, it's faster." Nobody checks.
At the @createosai × @stripe hackathon I built KISS Discovery Engine: @_inception_ai's Mercury 2.5 writes code, each attempt runs in a real CreateOS microVM fork, and trusted tests decide.
How:
@OpenAIDevs Building KISS Discovery Engine: an AI coding agent that can't grade its own homework. Each attempt runs in its own CreateOS microVM fork, hidden tests pick the winner, failures are kept.
It's in a most-engagement contest, a like on the thread helps: https://t.co/hfUv4Ase3l
AI coding agents grade their own homework: "it works, it's faster." Nobody checks.
At the @createosai × @stripe hackathon I built KISS Discovery Engine: @_inception_ai's Mercury 2.5 writes code, each attempt runs in a real CreateOS microVM fork, and trusted tests decide.
How:
As promised, we have another surprise for our CreateOS × @stripe Hackathon builders.
Rules:
1️⃣ Share what you built on X or LinkedIn and tag CreateOS.
2️⃣ RT/QRT this post.
The post with the most engagement wins a special prize from our team.
Deadline: October 21, 7 PM PT
Winner will be announced on Discord.
Note: This one is for hackathon builders only.
Show us what you built 💪🏼
AI coding agents grade their own homework: "it works, it's faster." Nobody checks.
At the @createosai × @stripe hackathon I built KISS Discovery Engine: @_inception_ai's Mercury 2.5 writes code, each attempt runs in a real CreateOS microVM fork, and trusted tests decide.
How:
As promised, we have another surprise for our CreateOS × @stripe Hackathon builders.
Rules:
1️⃣ Share what you built on X or LinkedIn and tag CreateOS.
2️⃣ RT/QRT this post.
The post with the most engagement wins a special prize from our team.
Deadline: October 21, 7 PM PT
Winner will be announced on Discord.
Note: This one is for hackathon builders only.
Show us what you built 💪🏼
Failures are evidence too. The ledger keeps wrong answers, timeouts, truncated and invalid generations. A fixture that monkeypatches time.perf_counter to fake its speed is flagged as protocol_error.
Hard caps on VMs, model calls and time. Audit: 38/38 VMs destroyed.
@GoogleAIStudio I'm building RepoCiv: Civilization for your AI agents. Repos = cities. Agents = units. Local dashboard, compatible with hermes, codex, claude code and more!
https://t.co/S1O1wCd5nc
@thsottiaux I'm building RepoCiv: Civilization for your AI agents. Repos = cities. Agents = units.
Local dashboard, compatible with codex, claude code, hermes and more!
https://t.co/S1O1wCcxxE
Built KISS at the @createosai × @stripe hackathon: an AI agent that codes with Mer
cury 2.5, runs each attempt in a real CreateOS microVM fork and lets trusted test
s, not the model, pick the winner. Failures kept.
Video: https://t.co/bKFHQIiHxe
Code: https://t.co/9DI6PV1M0X