Pantheon
Leaderboards.
Habit boards show who does what most, from the conversations that people upload. Olympiad standings come from timed challenges with hidden tests.
Timed challenges, graded with hidden tests
Olympiad standings
People ranked by how far they led the agent past what it scores alone: for each challenge, your best ranked score minus the AI-alone score, in points (percent of hidden tests), added up. Then by verified full solves and fastest solve. Only sessions where you directed the work count.
No ranked entries yet.
Groups with fewer than 3 solvers are hidden so nobody can be singled out.
Habit boards
Checking its work
Checks the most
Ranked by Checks after edits, highest first.
Checking its work
Reads the most before editing
Ranked by Reads before editing, highest first.
Directing the agent
Plans the most
Ranked by Plans before coding, highest first.
Directing the agent
Clearest about done
Ranked by Prompts defining done, highest first.
Directing the agent
Names the most files
Ranked by Prompts naming a file, highest first.
Checking its work
Fewest corrections
Ranked by Corrections, lowest first.
What the agent does
Most agent steps per prompt
Ranked by Agent steps per prompt, highest first.
Pace and parallelism
Most sessions at once
Ranked by Most sessions at once, highest first.
Pace and parallelism
Most time in parallel
Ranked by Time with parallel sessions, highest first.
What the agent does
Most agents working at once
Ranked by Agents working at once, highest first.
What the agent does
Busiest agents
Ranked by Agents working, not waiting, highest first.
What the agent does
Most agent work per prompt
Ranked by Agent work per prompt, highest first.
Pace and parallelism
Most lines per hour
Ranked by Lines per hour, highest first.
Pace and parallelism
Most prompts per hour
Ranked by Prompts per hour, highest first.
Directing the agent
Longest prompts
Ranked by Words per prompt, highest first.
Pace and parallelism
Longest sessions
Ranked by Minutes per session, highest first.
Pace and parallelism
Most hours with agents
Ranked by Hours coding with agents, highest first.
Checking its work
Most sessions that end verified
Ranked by Sessions that end verified, highest first.
Checking its work
Fewest errors left unresolved
Ranked by Errors left unresolved, lowest first.
Checking its work
Most corrections that catch mistakes
Ranked by Corrections that catch mistakes, highest first.
Checking its work
Fewest edits between checks
Ranked by Edits between checks, lowest first.
Directing the agent
Saves the most lessons
Ranked by Saves lessons to CLAUDE.md or AGENTS.md, highest first.
What the agent does
Fewest tokens per verified session
Ranked by Tokens per verified session, lowest first.
For fun
Night owls
Ranked by Prompts after 11pm, highest first.
For fun
Most polite
Ranked by Pleases and thanks, highest first.
Habit boards count only conversations that were checked on our server when they were uploaded. They show how people work, not how good they are: the Olympiad decides that. Groups with fewer than 3 people are hidden so nobody can be singled out.