Catalyst

Catalyst is an independent AI product and research lab.

It started with a boring failure that kept happening. I would correct an AI output, get a better second draft, then see the same mistake again in the next session.

The model had read the correction. The system had not changed.

We're trying to solve that gap. What evidence should be kept? What rule, if any, should come from it? Where should that rule apply? Did it improve the next relevant result? Can a person see and undo the change?

Those questions are more interesting to us than making another chat screen.

[Meraki](https://meraki.itscatalyst.com) is the current product. [Meraki Core](https://github.com/itscatalyst/merakicore) is the engine underneath it.

This is still early. Some parts work in local tests. The hosted product and independent human evaluation are not finished.

We're interested in systems that improve through use, without hiding how they changed.

Here is the part that matters: a correction is evidence, not permission to rewrite a person into a permanent profile.

A useful system should keep the source. It should propose a narrow change. It should apply that change only to matching work. It should compare the next result with the old behavior. It should let the user limit or roll it back.

We think this layer should survive a change of model or tool. Your working history should not belong to one chat window.

We don't know if the full idea works yet. Meraki Core has tests for the local correction, guidance, evaluation and rollback loop. The next job is proving it with invited users and a live database without losing the controls.

The sentence we keep returning to is simple:

**Every correction should improve the next result.**

If we cannot show the evidence, the change, the evaluation and the rollback path, we do not call it learning.

I am Pratham. I am seventeen, and most of my life has been shaped by the internet before I had the language to explain what that meant.

At nine, I wanted to become a YouTuber. At twelve, I was editing videos for other people. By fifteen, I was working as a creative strategist, learning how attention moves, why some ideas stay in your head and why technically polished work can still feel completely dead.

Then much of what I had built collapsed.

Near the end of sixteen, I lost clients, direction and the identity I had wrapped around my work. For a while, I did not know what I was rebuilding toward. AI entered my life less like a trend and more like an emergency exit. I started using it because I needed leverage, then kept going because I became obsessed with the machinery underneath it.

I taught myself how language models work, then moved outward into embeddings, retrieval, tool use, agent loops, memory systems, evaluations, orchestration, data pipelines and the infrastructure required to make a model useful outside a clean demonstration.

I built operating systems for creators, second brains for teams, agent workspaces and many experiments that were more educational than successful. They looked different from the outside, but they kept exposing the same failure.

The model could produce. It could not reliably accumulate a relationship.

It did not know which correction was a one-time request and which one revealed a lasting preference. It could retrieve an old message without understanding why it mattered. It could imitate the surface of a person’s style while missing the judgment that produced it.

Catalyst grew from following that problem further than was practical.

I am building it as an independent lab because I want the freedom to move between product, infrastructure, research, design and whatever the problem requires next. The point is not to defend one implementation forever. The point is to keep making the system more honest, more inspectable and more capable of becoming useful through experience.

I am still early. So is the work. That is part of what makes it worth doing.