Psyho

AI

The reported MATH-500 score rose from 7.9% to 48.1% with the same sampling settings.

18240123456789012345678901234567890123456789

TW

AK

RO

225 Sources, 1d ago, updated 1d ago

AI

Its creator says testing uses Codex or Claude Code harnesses without code generation; the game set won't be released publicly.

1801234567890123456789

PS

1 Source, 18d ago, first seen 18d ago

AI

A post says Remek is developing the model alone after hours and paying for compute himself.

1701234567890123456789

LK

PS

2 Sources, 18d ago, first seen 18d ago

AI

A user frames a shared image as what happens when you ask an AI agent to solve the impossible.

1901234567890123456789

DF

PS

2 Sources, 40d ago, first seen 40d ago

AI

The user says access to the maze’s source code removed the visual component, making the setup substantially easier than the official test. They report reaching 69 gems with Astra after a few more days on one CPU with 16GB of memory.

3601234567890123456789

PS

1 Source, 54d ago, first seen 54d ago

AI

Top competitive programmer tested Astra on hard unpublished logic puzzles.

6901234567890123456789

BP

PS

2 Sources, 60d ago, first seen 60d ago

AI

Pseudonymous commentator posts screenshot claiming GPT-6 Astra finished the game.

139012345678901234567890123456789

AC

DF

KA

13 Sources, 61d ago, first seen 61d ago

Gaming

Theo Jaffee's tweet drew replies defending speedrunning and comparing it to other pursuits like alchemy or chess.

105012345678901234567890123456789

LI

TJ

AT

14 Sources, 65d ago, first seen 65d ago