hi botpress folks, I’m PSBigBig. indie dev. no company, no sponsor. just too many nights with AI + notebooks. last year i basically disappeared and spent ~3000+ hours building something i call the WFGY series. important: it is NOT a new model, not a fine-tune. it’s just files (PDF / docs / TXT pack) you can feed into any strong LLM (Claude, GPT, etc) to make it behave more structured. everything is open source, MIT, one repo: https://github.com/onestardao/WFGY I try to explain this from user side, not from my own theory side. for me WFGY has 3 layers, and you can pick the one that fits you. -------------------------------- WFGY 1.0 (everyday Claude users) “same model, but act smarter” one PDF. no code. no API. just normal chat UI. -------------------------------- WFGY 2.0 (builders: RAG / tools / agents) “stop saying ai is random, start naming the failure mode” so WFGY 2.0 is a map of 16 reproducible failure types (No.1..No.16), each has: symptoms, root cause patterns, and minimal repair ideas. examples in plain english: No.1 hallucination + chunk drift No.2 interpretation collapse (context is right, reading is wrong) No.3 long chain drift (step 4/5 solves a different task) No.5 semantic != embedding (cosine says close, meaning is not close) No.8 debugging black box (wrong output, no clue where it died) No.14-16 bootstrap/deadlock/pre-deploy collapse (order of services kills first run) benefit: you can say “this looks like No.3 + No.8 for us, fix those first” instead of endless trial-and-error. also you can use it in pure natural language: “you are a WFGY emergency room doctor. i describe my pipeline symptoms. map it to No.1..No.16 and give minimal repair plan.” then you paste logs, steps timeline, screenshots, whatever. -------------------------------- WFGY 3.0 (researchers + philosophy lovers) “131 cross-domain big questions written in the same structural language” this one is not for everyone, and that’s fine. it’s for people who use LLMs to think about heavy stuff at 2am: free will, consciousness, alignment, climate, economics crashes, complex systems, math structure… WFGY 3.0 is a TXT pack of 131 cross-domain questions, all written in the same structural language, so you can decompose them into smaller evaluable tasks. i’m not claiming final truths. it’s more like a new language to discuss old problems in a falsifiable, structured way. you can do: “show me the free will question in 3.0, debate with me” “take climate-related ones, propose 3 research projects” “compare your alignment story with 3.0 encoding, show conflicts” -------------------------------- why i share here i saw this channel is literally “share your bots”. so i’m dropping this as a tool for bot builders: if you build RAG bots / agent workflows, WFGY 2.0 is probably the most practical entry. if anyone wants, i can also write a short “botpress style checklist” post: how i use No.1..No.16 to debug a bot that keeps hallucinating or drifting. again, all MIT, free to use/remix/attack: https://github.com/onestardao/WFGY thanks for reading, and sorry it’s long. i just want to explain it from user benefit side. ❤️