AIM: Efficient Local Agents Hackathon · Small model. Smart harness
Same small language model and budget for every team, plus Jev 1.14 by TypeSafe AI for instant typed decisions. The best harness wins.
⚠️ IMPORTANT: APPLICATION VIA AIRTABLE
To participate in the Hackathon, please apply through the Airtable form below.
👉 Apply here: https://airtable.com/appSzS1Qf9d7I2nKm/pagnJd0Rr5uA6l8ba/form
No registration on Luma is required.
Seats are limited (~30) and allocated through an application process. Please apply with your CV or LinkedIn profile.
👉 Application deadline: Saturday, November 7, 23:59 CET
👉 Confirmations sent by: Monday, November 9, 23:59 CET
THE CHALLENGE
• All teams share the same small language model and token budget. The model is announced at the opening.
• Jev 1.14 by TypeSafe AI, a System One model for cheap typed decisions ("supported or not?", "retry or continue?")
• You engineer the harness: planning, tools, memory, retrieval, verification
• Build against our live benchmark with instant feedback per task
• 17:00: freeze your harness; your agent runs once, on its own, on an unseen final scenario
• Score: 50% accuracy (right answers and right sources), 50% cost
• Coding assistants are welcome while building; at runtime only the two provided models
SCHEDULE (subject to change)
09:00 Doors open
09:15 Opening & challenge reveal
12:00 Pizza lunch
17:00 Final evaluation
18:00 Leaderboard reveal, solution presentations & award ceremony
19:30+ Drinks, food & networking
PRIZES
IBM credits & goodies for the top teams. Certificates for winners and all participants.
Looks great on your CV and LinkedIn, and the evening is for networking with our partners.
PRACTICAL
• Teams of up to 4: everyone applies on their own with the same team name, or picks "alone" / "assign"
• Bring your laptop with an IDE and dev tools set up. IBM emails Wi-Fi access links.
• Food provided all day
• Photos and videos may be taken. Tell board@ai-mission.eu in advance if you'd rather not appear.
Organised by AI Impact Mission (AIM) in collaboration with IBM, Klartext AI and Texterous. Questions: board@ai-mission.eu