Show HN: I built an AI that uses a metacognitive loop 2 solve invention problems
8 WiseRob 3 9/3/2025, 6:46:00 AM robw1se.substack.com ↗
Hi HN,
I'm the creator of this project. For the past months, I've been working on building an AI agent that could move beyond simple generation and tackle inventive challenges autonomously.
The core idea was to create a system with a "metacognitive loop"—the ability to recognize when it's stuck on a fundamental problem and then launch a sub-mission to solve that specific bottleneck before continuing.
The linked article is a deeper introduction to the system's architecture and a snapshot from a recent run.
I tried to design it to be evidence-grounded and self-critical to avoid the pitfalls of hallucination.
Happy to answer questions
RW
Comments (3)
adastra22 · 1d ago
I’ve been finding success with similar cognitive loops for vibe coding large projects. How generalizable is your implementation? To other scientific field? To other industries entirely?
WiseRob · 1d ago
The code I have written is domain agnostic - its intention is to mine the knowledge within LLMs encoders etc to "invent" novel solutions to problems. Essentually you give it the problem and a desired solution description and the thing works away using inventive steps to create a forward trajectory toward the goal solution. I have a "red team" function that checks the validity of the steps forward against what is real and negotiates bottleneck issues or kills the idea - causing the system to try something else - it can also call itself to try to solve that bottleneck problem if the system feels it is not an intractable problem. It is all on autopilot so the user can set and forget and allow it to work away.
adastra22 · 1d ago
We all seem to be writing the same thing. You have plans to open-source it? I feel like there is potential here to stop wasting time replicating work, and to create a single framework that everyone benefits from.