task grid
task
running
test input
the answer the lab drew
abstraction.lab [Set : ARC-AGI-2 public] - Task: loading / Solved : 0 / Spent : 0.00 USD
the team
observer
reading the examples
hypotheses
waiting
judge
waiting
- engine
- search
- programs tried
- 0
- this attempt
- 0.00 of 0.50 USD
- time on it
- 0 ms
the attempt budget
The lab only ever works the public sets.
the token that pays for the compute
judge log
waiting for a hypothesis
- verdict
- waiting
- answer hash
- no answer drawn
- written on chain
- the keeper writes it when the lab is live
token desk
Abstraction $ABSTR
contract
not yet: the mint goes in at launch
Chart
Follow- compute
- 70 percent of the fees the lab receives
- buy and burn
- 20 percent
- the day's task
- 10 percent
the archive
[0 of 0]
the score
0
tasks solved in the open
- public training set
- 0 of 1000
- public evaluation set
- 0 of 120
- attempts
- 0
- spent on compute
- 0.00 USD
attempts a day
task of the day
Holders pick one unsolved task. It gets the day's reinforced budget. One wallet, one signature, weight is the holding, capped at one percent of supply.
voting closes at midnight UTC
the day's budget: 0.00 SOL
the burn
A fifth of everything the lab receives buys $ABSTR on the market and burns it. Every burn is a signature you can open.
prize rules
If the team takes any ARC Prize award, 80 percent of it is split between holders by a snapshot taken the moment it is announced, and 20 percent buys back and burns $ABSTR.
- competition
- ARC Prize 2026, ARC-AGI-2 track, on Kaggle
- final submission
- 2 November 2026
- results
- 4 December 2026
- team status
- not submitted yet
nothing to claim: no award has been announced
Abstraction enters the competition like anyone else. It is not affiliated with the ARC Prize Foundation and uses none of its marks.
how it works
- 1
the fees arrive
Creator fees of $ABSTR reach the lab wallet. The keeper splits them: 70 compute, 20 buy and burn, 10 the day's task.
- 2
the team takes a task
A public ARC task, examples first. The observer says what changes, the hypotheses role writes one transform, the judge runs it on every example.
- 3
the answer is drawn
A program that reproduced every example is allowed to draw on the test input. If the answer matches, the task is solved, and the failures stay on the desk too.
- 4
it is written down
The keeper writes every solved task on chain as a memo: the task and the hash of the answer. Nothing can be edited afterwards.