Some of these started with a comment we left under someone's paper or post, some with a message someone sent us. Each has its own subject and its own twists.
In half a minute
Three people, three different questions. Each of them builds something of their own, and we work with them on one question, in the open.
Anagha Langhe checked whether a lighter version of her model (FP8) answers worse. Out of 120 questions, only one went from right to wrong.
Mark Goodbody tested whether his reasoning frameworks change what ChatGPT and Claude decide. In these runs the main distinctions appeared with and without them, so no benefit is established yet: the models already keep "what we can say" apart from "what we should do" on their own.
Mohan Bheemadolu squeezes an agent's memory into tiny codes, and together we're testing whether that makes two AIs agree on the same wrong answer more often.
Anagha put the same 120 questions to a bf16 build of her model and to an FP8 build. Only one answer flipped. When both were wrong, they gave the same wrong answer 42 times out of 48.
Three advisers recommend the school, all from the same two-year-old map. One adviser has today's inspection saying the school's step-free entrance is closed. Both AIs counted the three as one source, with or without Mark's framework in front of them.
When an agent's memory missed the fact, GPT and Claude still gave the same answer 33 times, and 15 of those were wrong. Mohan's FlyHash switches on only a few active bits per memory. Does a more compressed code make that shared mistake more common?
You bring a question, a system or some data: a model you run, a memory for an agent, a set of cases, a paper you want to put to the test.
Before looking at any results, we agree on what to run and what would count as an answer.
Each side runs its part. Everything that can be open is open: code, prompts, raw answers, numbers.
The story goes on a page like these, in plain words, with your name and links, and grows as the work does.
Propose a collaboration
Tell us in a few lines what you are working on, what you would like to find out, and where your code or paper lives. We read everything and answer personally.