THE FUTURE PAGE

[ Front page ] [ The edge ] [ Finished ] [ About ] [ Links ]

The latest reading, written down 2026-09-27.

Assembly mistakes spotted in a photograph

83.3 per cent

Held by Claude Opus 5.5. Set 2026-09-22, and has stood for 7 days.

Everything else on this page asks a machine to do something machines were built for - arithmetic, search, symbols, a board with sixty-four squares on it. This one hands it a photograph of a half-built shoe cabinet and asks what went wrong. The benchmark was assembled by buying three pieces of IKEA furniture of increasing difficulty and photographing the builds as they went: some of the photographs show a correct build, and in the others a realistic mistake was made deliberately, often with assembly carried on past it so that the error is visible but is not the step just finished. The model gets the official manual as a PDF, a tool for zooming into the image and a Python interpreter, and it has to connect a technical diagram to a real object in a room - which is what fixing an appliance or a car actually requires, and which almost nothing else tests. The grading is deliberately lenient: name the right step, describe the problem roughly, and it counts. So this is the generous reading of the ability rather than the strict one, and the best score has still only just passed four fifths, while plenty of models released in the very same week sit nearer a fifth. That spread is the reason this is worth a page. It is not a skill the field has broadly; it is a skill two or three models have, this year, for the first time. In 1993, the year this page is pretending to be from, machine vision meant reading typed digits off a cheque under a controlled lamp.

That description was written when this frontier was opened, on 2026-09-27, and is never rewritten. Where it and the number above it disagree, the number is the reading and the description is history.

[ The whole history of this number ]

The edge today

• The longest task an AI can finish on its own — 1,045 minutes
• Open problems in mathematics solved — 2.9 per cent
• Unsolved integer sequences worked out — 29.9 per cent
• Real software issues resolved — 83.5 per cent
• Questions answered without inventing an answer — 75.6 per cent
• Chess puzzles solved — 72.0 per cent
• The largest training run ever disclosed — 1 x 10^27 operations
• The largest training cluster ever disclosed — 200,000 chips
• The fastest computer on earth — 2,198 petaflops
• The most arithmetic per watt — 73.282 gigaflops per watt
• The largest prime number ever found — 41,024,320 digits
• The largest quantum volume — 8,388,608 quantum volume
• The smallest a gigabyte of Wikipedia has ever been made — 96,096,261 bytes
• The most data ever read and written in a second — 26,888 gibibytes a second
• The most connections ever followed in a second — 410,266 billion edges a second
• The most digits of pi ever computed — 314,000,000,000,000 digits
• The highest rating any chess program holds — 3,650 Elo
• The most transactions a database has ever run in a minute — 2,055,076,649 transactions a minute
• The most storage operations ever done in a second — 30,002,765 operations a second
• Assembly mistakes spotted in a photograph — 83.3 per cent

Watching 20 frontiers. 6 others have finished.   [ The full table, with dates ]

The rules

• Every number here carries a date and a place anyone can check it.
• When a number moves, the old one is kept. Nothing is edited or deleted. A script enforces this.
• A number nobody can check is not published.
• Every page on this site is a single document smaller than 14,336 bytes.

Written down in:   2026

When a number moves it appears in the feed. Nothing else on this page can reach you again.

[ Front page ] [ The edge ] [ Finished ] [ About ] [ Links ]