Post

all these posts on how ppl gave llms “memory” or w/e completely vibe coded based on some vibedea i love that ppl are building stuff but tbh what they should be building is an mcp server for doing actual science, with like data collection + hygiene, controlled experiments, systematic evaluation

in retrospect my post was a little snarky, but i meant it lightheartedly. i actually built pretty much exactly what i was suggesting recently for my own project, & it’s been kind of incredible to use. i give it an experimental protocol, it validates it, runs experiments, & analyzes the results.. <,<

interesting post for a number of reasons. the oss section is a sticky wicket. but for this thread in particular, i've found it useful, similarly to you, i think, to put in thoughtful work to consistently constrain the llm to a "3 foot world"