By engineers, for engineers
Oxynote started as a small, slightly stubborn attempt to fix something that kept annoying us.
Everyone we know codes with AI now, and almost everyone is doing it the hard way: re-typing the same context every session, watching the same prompt produce different code on different days, fixing AI-generated code that almost matched what was meant. The writing that feeds the AI is doing real work, but it's living in chat boxes and scratch files.
Engineers have been writing software in English for years now, and the tooling has barely caught up. The two open source projects growing fastest in the space, Spec Kit and OpenSpec, are conventions for "where to put the words your AI reads". They've pulled over 130k GitHub stars between them. The signal is loud: more teams are reaching for real tooling around the writing.
The problem is the tooling so far is a folder of markdown files. There's no editor built for it, no proper review system, and no link back to the code that came out the other side. English is being treated as a new programming language, but without any of the things that make programming languages workable at team scale.
So this is what Oxynote is:
Everyone we know codes with AI now, and almost everyone is doing it the hard way: re-typing the same context every session, watching the same prompt produce different code on different days, fixing AI-generated code that almost matched what was meant. The writing that feeds the AI is doing real work, but it's living in chat boxes and scratch files.
Engineers have been writing software in English for years now, and the tooling has barely caught up. The two open source projects growing fastest in the space, Spec Kit and OpenSpec, are conventions for "where to put the words your AI reads". They've pulled over 130k GitHub stars between them. The signal is loud: more teams are reaching for real tooling around the writing.
The problem is the tooling so far is a folder of markdown files. There's no editor built for it, no proper review system, and no link back to the code that came out the other side. English is being treated as a new programming language, but without any of the things that make programming languages workable at team scale.
So this is what Oxynote is:
- An IDE for what you tell the AI to build. A real editor for the writing, not a markdown file in a git folder. Structured for AI to read, formatted for humans to review.
- A review and diff view. Click through the changes, preview the generated code, approve the version that ships.
- Live link between the writing and the code. Click a block, see the code it produced. Click a function in the repo, see the writing behind it. Both stay linked as either side changes.
- Hooks into the world the code runs in. Production metrics, third-party APIs, container images, external pages. When something out there changes, the writing it depends on gets flagged before the AI ships against the old version.
We're keeping the whole thing intentionally simple. You shouldn't need a training session to use it. You open it, it behaves the way an editor should, and it gets out of your way.
Who we are
We’re Simon (GitHub: swithek) and Dovydas (GitHub: davseby) — engineers who like building tools we’d want to use ourselves.
Over the years we’ve created and contributed to a bunch of open source projects, mostly in the "make developer life less annoying" category. One example is ttlcache, an open-source in-memory data storage and caching library for Go: https://github.com/jellydator/ttlcache. It’s used widely enough that you’ll spot it being pulled into public repos from teams like Microsoft, AWS, Datadog, and Tailscale, among many others.
Oxynote is us applying the same mindset to the writing behind AI-generated code, so that it could stay structured, reviewable, and linked to what ships.
If you want to say hello, you can reach us at hello@oxynote.io
Over the years we’ve created and contributed to a bunch of open source projects, mostly in the "make developer life less annoying" category. One example is ttlcache, an open-source in-memory data storage and caching library for Go: https://github.com/jellydator/ttlcache. It’s used widely enough that you’ll spot it being pulled into public repos from teams like Microsoft, AWS, Datadog, and Tailscale, among many others.
Oxynote is us applying the same mindset to the writing behind AI-generated code, so that it could stay structured, reviewable, and linked to what ships.
If you want to say hello, you can reach us at hello@oxynote.io
©2026 Oxynote. All rights reserved
Crafted with care in Lithuania & France
Closed beta. Request access above.
Crafted with care in Lithuania & France
Closed beta. Request access above.
Say hello: hello@oxynote.io