About OpenChem
OpenChem is a collection of classic chemistry experiments and procedures — and a reference project by ATAILA. The whole site was built in a matter of hours on ATAILA's own infrastructure, to answer a question we are asked constantly: what can these models actually deliver — not in a demo video, but as a real product with a database, an admin portal and a public website?
What the machine did here — and what it did not
- The application
- Programmed by ATAILA Coder, our on-premises coding solution — an agentic coder driving open-weight models on our own GPUs, not a hosted assistant. The API, the database schema, the admin portal and this website came out of it.
- The write-ups
- Compiled by a language model running on our own servers — drawn from well-known, long-established classic demonstrations that chemistry teachers have been performing for generations. The machine did not invent new chemistry: it arranged known procedures into an edited, bilingual form.
- The images
- Not lab photographs. The illustrations were produced by a separate AI node dedicated to image generation. Nothing was sent to a public AI service, and no data left our hardware.
A “fork” in a day — from OpenChef to OpenChem
OpenChem is the sibling of openchef.hu. Same idea, same engine — structured write-ups, taxonomy, ingredients, AI imagery, admin portal, public site — except the subject is chemistry experiments instead of recipes, with a recommended age and safety notes on every one. From the idea to a site running in production took hours, not weeks.
Strictly speaking it is not even a “fork”: we did not copy the code. One shared codebase runs both sites from the same container image, and a single switch decides which brand you get. OpenChem received its own database, its own domain, its own analytics and its own content without branching a single line of shared code. That is exactly why it was this fast — and why the two sites will not slowly drift into two divergent systems.
What it took — three platform pieces
- ATAILA Factory — the landing zone
- The Factory built the new project's entire “house” on our own hardware: separate environments for development, testing and production, its own database, object storage, secrets in the vault, DNS records, certificates, container namespace and network rules. By hand that is days of work and a dozen steps you can get wrong — here it was one process that completed on the first run.
- ATAILA Coder — writing the delta
- The whole application did not need rewriting — only the difference: the new brand profile, the chemistry vocabulary (recipe becomes procedure, ingredient becomes chemical), the age and safety labelling, the legal texts, and the content of ten experiments in two languages. Our locally-running coding agent wrote it, driving open-weight models on our own GPUs.
- ATAILA Release Manager — going live
- The finished container images travelled development → testing → production from here, by version number, fully traceable. This was not “uploading something to a server”: every environment received exactly the same verified image.
The three together are the point. Without the infrastructure it takes weeks before a new project even has somewhere to live and a database to use; without the coding agent the delta is person-weeks of work; and without the release manager you are left with the gap between “it works on my machine” and “it runs in production, tested” — the exact point where most demos stall. So the experiments are not the interesting part of this site. How quickly they became a working, public product is — on our own hardware, with our own data.
Liability statement
The procedures and experiment write-ups published here are for educational and informational purposes only, and are intended only for readers with prior training in laboratory work and in the safe handling of chemicals. Every experiment must be performed only under adult supervision, at your own risk. Every experiment lists a recommended age — take it seriously.
The write-ups were compiled by AI. Although they are based on well-known, long-established classics, they have not undergone independent professional review — so before you perform anything, verify the description against a reliable source (a textbook, a teacher, an official collection of experiments). OpenChem and ATAILA accept no liability for damage arising from the use of these descriptions: whoever performs an experiment does so at their own risk.
Why it exists
The experiments are not the interesting part — the path from an idea to a running product is. This site came together in hours: the coding agent, the model tiers behind it, the image node and the platform that provisioned the whole project made that possible. We will publish that reference architecture, because it is the part you would actually be buying.
That is the claim this site exists to support: these days, if you can describe it clearly, it can be built — on your own hardware, with your data staying where it is.
Interested in something like this?
We can build it for you. One workflow, turned into a private, AI-powered app that runs on European hardware you control — your AI, your data, your hardware. OpenChem is what that looks like when the subject is classic chemistry experiments; yours would look like your business.
You imagine it. We make it real.