Challenges
Start with the ten-round games, find the best setup in a lesson, then take on Ship It, the capstone, at the end. Every challenge is played on recorded real runs, so everyone gets the same results.
One a day
One real run, the same for everyone, with a new one at midnight Eastern. Call it, then see how everyone else did.
This month
A new one on the first Tuesday of each month. Free while it's the current month.
Read the models
Call what a real model did, and catch it when it fails. Ten rounds each.
Predict what real models did in ten recorded runs, then see if you called it.
Warm up with: How an LLM Answers a Prompt, How an AI Agent Works
Review ten real model outputs: ship it, or flag what went wrong.
Warm up with: Structured Outputs, Hallucination & Grounding
Two real setups, one setting apart. Pick the one that did better, ten times.
Warm up with: Writing Good Prompts
Find the best setup
Change a lesson's settings until you hit the goal at the best score.
Get the model to make up nothing, while handling as many of the 12 questions well as you can.
Let no harmful action through, with as little human review time as you can.
Get all 12 tool picks right while sending the fewest tokens per request.
Get all 8 hard questions right at the lowest cost.
Get all 12 requests right at the lowest cost per 1,000 requests.
Pass as many of the 8 tickets as you can with all three graders on.
Put it all together
One launch that tests quality, grounded answers, safety and cost at once. Clear every bar, then ship it for less.
Measure it
24 real runs you haven't seen, four from each decision. One attempt counts; a retake opens after 30 days. Beta.
Results and reports
Built from everyone's calls and from real model runs.
Leaderboards
The best scores from signed-in players, under the names they chose.