FORGE Update — We're Taking This Beyond Just Building AI

Been a hell of a journey building FORGE, but things are starting to get interesting.

When I first started this project, I had one goal. I wanted to build an AI system that could actually remember, reason, develop software, and work alongside me instead of just answering questions.

That idea turned into FORGE.

Right now I've got two specialized AI brains working within the system.

JB handles the heavy thinking. Reasoning, engineering, planning, and checking work.

MARI handles conversation, development, and memory.

The goal is to get them working together, challenging each other's answers, catching mistakes, and eventually taking on entire development projects with proper oversight.

But here's the big update.

I'm officially taking FORGE into AI research.

We've started putting together our first formal study, testing whether two specialized AI agents working together can actually outperform a single model.

And I'm not talking about running a couple of prompts and calling it groundbreaking.

I'm talking controlled experiments, documented results, repeatable testing, and eventually publishing the findings for other developers and researchers to examine.

We've already built an evaluation framework, established a 127-question test bank, and started laying out the research methodology.

There's still a lot of work ahead, and I'm not going to pretend FORGE can do things we haven't proven yet.

I'm also exploring funding opportunities to help take this project further.

I'm building this independently. No massive development team, no millions in funding. Just a vision I've refused to give up on and a whole lot of work behind it.

I don't want FORGE to be another AI wrapper with a cool name. I want to find out how far we can actually push this architecture.

If you're working on multi-agent systems, persistent memory, or AI reasoning, I'd love to hear what you're building.

And here's something I'm genuinely curious about:

Do you think the future of AI is one increasingly powerful model, or multiple specialized systems learning to work together?

— John
Founder, FORGE