lablab.ai

The #1 Ecosystem for AI Builders - backed by 100+ world-class tech partners.

[lablab.ai @lablabai]

LabLab is bringing AI Hacks to Web3.

We are the largest global hackathon platform with over 250k+ builders, and foundational partnerships with @Google, @Nvidia, @Meta and many more.

We’ve proven our tracks; partnering with @Circle to co-launch @arc, shipping 4k +new apps in two weeks.

It’s time to build with the best.

[lablab.ai @lablabai]

Fourteen thousand dollars, November 3-8, online with an optional finale in Salt Lake City.

Full brief, the two build patterns, and the one thing every submission has to show on video ↓

https://t.co/HNZiST6HQl.

[lablab.ai @lablabai]

Everyone’s racing to build agents that act. Almost nobody’s building the box they act inside.

That’s the gap the Vultr Agent Rush Hackathon is aimed at ↓

https://t.co/HNZiST6a0N.

[lablab.ai @lablabai]

An agent that does real work. Writes and runs code, or drives a live browser, with every action contained in a sandbox on Vultr. Then prove the boundary holds.

Most agent demos are judged on what the model says. This one is judged on what the sandbox absorbs.

[lablab.ai @lablabai]

All six converged on the same answer. Not better models, not cleverer prompts. Separation of duties.

Which is exactly how you’d run a factory with the lights off.

Thanks to BAND for the challenge and the judging. Featherless awards still to come! Stay tuned for that...

[lablab.ai @lablabai]

Nobody watched the code get written. That was the whole point.

The Dark Factory challenge asked teams to build the thing that builds the app. 167 submissions came in. Here are the six that won ↓

https://t.co/MCVwUibOh1.

[lablab.ai @lablabai]

Third: The Felt Five, a solo entry. Six seats, one cross-auditing the rest, and it injects faults into its own test suite to check the suite is worth anything.

Of 81 seeded faults it caught 75. The project’s own shipped checks caught 8.

[lablab.ai @lablabai]

Most of what gets built at a hackathon never leaves the repo. Not because the idea was bad, but because nobody knows what to do on day eight.

Online build runs until October 18, on-site in Rome, Milan and Imperia. Total prize pool is $17,000 and up.

[lablab.ai @lablabai]

Most hackathon prizes end when the weekend does. This one starts after it.

IMMENT joins AMD Developer Hackathon: ACT III with incubation instead of credits ↓

https://t.co/VXWsOPjtfD.

[lablab.ai @lablabai]

IMMENT helps founders create, fund and manage startups, with specialised expertise in Administration, Finance and Control.

Their challenge gives the three most promising startup ideas dedicated support to turn a weekend build into something real. Each gets a programme worth €1,599.

[lablab.ai @lablabai]

Welcome to the lineup, @GoogleDeepMind. Thanks @AIatAMD for making it happen.

Sign up: https://t.co/VXWsOPk15b.

[lablab.ai @lablabai]

Just when you thought this hack couldn't get any better.
@GoogleDeepMind just joined the @AIatAMD Developer Hackathon: ACT III.

Gemma on AMD GPUs. Let's go.

[lablab.ai @lablabai]

Online build: Oct 12-18, from anywhere.
On-site finale: Oct 17-18 in Milan, Rome and Imperia.
Free to join. All skill levels.

[lablab.ai @lablabai]

All of it built on @intel Robotics AI Suite, with Physical AI Studio and OpenVINO. Nobody stopped at a demo that talks. Every one of these executes, checks its own work, and recovers when it goes wrong.

Thanks to Intel for the challenge and the judging ↓
https://t.co/SsDwxhLfSb.

[lablab.ai @lablabai]

Set the table, but skip the cup.

Two robot arms heard that, worked out what order to do it in, and did it. That project just won the Intel Physical AI Challenge at the AI Infra Summit Hackathon ↓

https://t.co/6IT2gNwwZz.

[lablab.ai @lablabai]

Second: Bimanual Dinner-Table AI on Intel Edge, by SuperKings. One command becomes a two-arm plan ending in a pour into a mug the other arm holds steady. Learned policy runs in 1.35 ms on CPU.

Third: MISE, by SeoulStack. Ten of ten runs completed, recovery triggered on eight.