I Am Joining Forces With TeichAI And It Is Funny Either Way

I am officially part of TeichAI now. They know I exist. We have been communicating for a while. I am listed on their Hugging Face page as a collaborator. This is not a unilateral declaration. This is real. And it is still funny.

Collaborations usually start with contracts and meetings. Ours started with Discord messages and a shared belief that local AI should not require a second mortgage.

What TeichAI Does

They specialize in crafting high-quality distilled models and datasets. Their focus includes fine-tuning and distillation, custom dataset curation, and they built Datagen as an open-source tool for creating datasets like theirs. They publish model or dataset requests through TeichAI/Requests so people can ask for specific work. They have a Discord server for updates, announcements, inquiries, and conversations.

Their goal is making AI accessible through open source and open science. Their distilled models are designed to run locally. This is exactly what I have been trying to do with my TMLM series except they have better tools and documentation.

                        # TeichAI team members listed on HF page

                        @Liontix

                        @armand0e

                        @owenqwenllmwine

                        @tikeape

                        @CompactAI <-- That is me

                        # Small fish in big pond but still part of the school.

The Team

@Liontix @armand0e @owenqwenllmwine @tikeape @CompactAI

View @TeichAI on Hugging Face

What We Do

Fine-tuning & Distillation
Custom Dataset Curation
Open Source Tooling (Datagen)
Request-based Work (TeichAI/Requests)

Why This Partnership Works

We all believe in making AI accessible. We all publish open weights. We all want people to run models locally without burning their laptops or spending hundreds on API credits.

They distill from frontier models like Claude and GPT. I distill from... well, I try. Sometimes I train from random Reddit posts. But we are aligned on the mission. Democratizing AI means letting regular people run models on their own hardware. Not renting access forever.

The Communication Status

I have communicated with one of the managers for a while now. Nothing official happened suddenly. It was a gradual thing. Then I woke up and I was on the Hugging Face page. That is how collaborations happen sometimes. You talk, you help, you eventually appear together.

                        # Timeline of our non-dramatic collaboration

                        Phase 1: Thought ideas were similar

                        Phase 2: Talked via Discord/HF

                        Phase 3: They offered collaboration

                        Phase 4: I said yes because why not

                        Phase 5: Listed on Hugging Face page

                        # No grand declarations needed.

What This Means For TMLM

Sonnet-2 was going to use Engram. Now it might also get some TeichAI inspiration. Or maybe I will copy their dataset structure because I am lazy and efficient. Either way, my tiny models will be smarter.

Opus benefits too. Six hundred million parameters. Maybe it will remember things better than my current version which thinks Paris is a person named Pierre. The TeichAI influence might help there.

The Humor Element

This is still funny even though it is real. I still have one GPU running at 800W. They have pipelines. I still train models that give fish answers. They distill frontier intelligence. We are still friends. The contrast is the joke.

When you join forces with people who actually know what they are doing, you learn things. Mostly you learn how much you did not know. This is growth.

Future Plans

Haiku is already out and giving fish answers. Sonnet is training and learning Engram. Opus is planning to learn both. We might do joint releases. We might do community-requested work through TeichAI/Requests. We might just continue collaborating quietly and see where it goes.

The Discord server exists for updates, announcements, inquiries, and conversations. I joined. Mostly lurking. Occasionally contributing when inspired.

Final Thoughts

I am part of TeichAI now. They know I exist. I am listed on their Hugging Face page alongside Liontix, armand0e, owenqwenllmwine, and tikeape. I am @CompactAI in this universe now.

My models will keep training. Our models will keep distilling. Down the line we will probably release something useful. Until then I will sit here proud of being part of a group that does real work with actual pipelines.