cd /news/artificial-intelligence/1-month-and-60-models-later · home topics artificial-intelligence article
[ARTICLE · art-67940] src=supcpu.com ↗ pub= topic=artificial-intelligence verified=true sentiment=↑ positive

1 month and 60 models later

Romello Goodman trained about 60 small language models in one month, releasing 11 of them under the sup computer project, each exploring a different research question such as whether a frontier model can train a small model on a laptop or whether a model can learn a game from moves. The models, with mere millions of parameters, are knowledge-light and shaped by their training data, requiring novel interfaces beyond the chat assistant paradigm to become instruments for humans and agents.

read2 min views1 publishedJul 22, 2026

About a month ago I trained my first model and caught the bug. In that time I've trained around 60 models and released 11 of them, each helping me along my path as I learn how to do AI research. Each starting with a question:

shakespeare— Can a frontier model train a small model on my laptop?gatsby— How does the dataset shape the response of the model?kenosha-kid— Can the model memorize with slight variations?daydream-chess— Can the model learn a game from the moves?glyph— Can we use hallucinations stylistically?

With each release, I understand the material more and how to sculpt it. As opposed to the knowledge-dense heavy trillion parameter models at the frontier, these contain mere millions. They're knowledge-light and shaped like the data you give them. Give it chess games and it can roughly play them. Give it Shakespeare and it roughly writes you prose. This constraint gives me a framework to work within: Brainstorm the data and shape the model.

But within each of these models, there's a mismatch between them and your idealized language model. These models aren't fun to converse with. The dominant interface for language models is the chat assistant/agent, and these models are too small to fill that role. A back and forth with them would only produce gibberish. So there is a need to invent novel interfaces for this new class of models. Each one a harness shaped to the contours and skills the model naturally has. A chessboard for daydream, a type tester for glyph.

The tie between the model, its harness and the interfaces we use can create a synergy that turns even a single purpose model into an instrument to be played. Each piece of the puzzle supports the other to surprise and delight the system using it. This is the long-term bet of sup computer: creating instruments for both humans and agents. Now that we have the models, the real work begins.

── more in #artificial-intelligence 4 stories · sorted by recency
── more on @romello goodman 3 stories trending now
sponsored brought to you by zahid.host 4,200+ EU-deployed projects
reading about agents? ship yours in a single git push.

Run your AI side-project on zahid.host

EU-based hosting, git-push deploys, automatic HTTPS, no cold starts. Free tier with a custom domain — perfect for shipping the agent you just read about.

$git push zahid main
Live at https://your-agent.zahid.host
Get free account → Pricing
from €0/mo · no card required
LIVE [news/1-month-and-60-model…] indexed:0 read:2min 2026-07-22 ·