Sloptris
Tetris, if the pieces were AI-generated tickets that might mutate
Emerged finds new products worth a look, daily — what’s new today
It is a working benchmark of what models still cannot do, from the co-creator of Nim, with a public leaderboard you can point at when someone claims agents have solved programming.
KillSwitch is a small programming language from Dominik Picheta, co-creator of Nim, with one design goal: be hard for large language models and coding agents to generate or edit. The spec and a Rust reference interpreter are on GitHub under AGPLv3, with a clause forbidding AI training on the code and a pledge that everything is human-written. The repo includes a FizzBuzz, tests and an anagram tool for checking identifiers, a hint at the tricks it uses. A companion site scores frontier models, including Claude Opus, on KillSwitch tasks. Install is cargo from source.
More like this
SWE-sweep
A benchmark where agents get only a repo and must find the bugs themselves
Thrum
A local message bus that lets your coding agents review each other's work
StarSkirmish Bench
Frontier models get one hour to write a Brood War bot, then fight for it
no_human
Open-source agent that takes a ticket to a reviewed pull request, with proof
DeepSeek Harness
DeepSeek's open-source agent harness where every capability is a plugin