We Analysed 40 AI Launches of 2026: 3 Findings
3 findings from measuring every major AI project launched in 2026 — the fastest star growth ever recorded, a licence pattern nobody talks about, and a language shift that says where AI is heading.
We Analysed 40 AI Launches of 2026: 3 Findings
These numbers are ours. We pulled 120 notable AI repositories from the GitHub API on 1 September 2026, kept the 40 created this year, and measured them. Nobody reported these figures to us — which also means you can check them yourself, and the method is at the bottom.
1. The fastest star growth we can find a record of
DeepSeek Harness was 19 days old and had 207,069 stars. That is roughly 10,900 stars per day — an order of magnitude faster than the next project on the list.
| Project | Stars/day | Age |
|---|---|---|
| deepseek-ai/deepseek-harness | ~10,900 | 19 days |
| DietrichGebert/ponytail | ~1,459 | 82 days |
| ultraworkers/claw-code | ~1,267 | 154 days |
| mattpocock/skills | ~1,160 | 210 days |
What this says: a frontier lab open-sourcing its own agent harness pulls attention no independent project can match, however good it is.
2. Two out of three chose the most permissive licence
Of the 40 projects launched this year, 27 are MIT licensed — 68%. Six are Apache-2.0. Only three chose AGPL, the licence that restricts building commercial services on top.
The practical reading for anyone building: most new AI tooling is safe to build a business on. The exceptions are worth spotting early — an AGPL dependency discovered after launch is an expensive surprise.
- MIT and Apache-2.0: build freely, including commercially.
- AGPL: fine for internal use, obligations if you offer it as a service.
- No licence stated: the riskiest of all — you have no granted rights at all.
3. AI tooling is moving from Python to TypeScript
This is the finding we did not expect. Among projects created before 2026, Python leads TypeScript by 34 to 13. Among those created in 2026, it is 13 to 13 — a dead heat.
The reason is visible in what the new projects are. The old ones are research and model code, where Python is unquestioned. The new ones are agent harnesses, developer tools and interfaces — products, which live where the web lives.
If you are choosing what to learn: Python remains essential for models. But the tools people actually install are increasingly TypeScript, and that is a change worth noticing early.
How we measured this
Collected the data
GitHub API search across ai, llm, machine-learning, ai-agents and generative-ai topics, 1 September 2026.
Cleaned it
Removed archived repositories, forks and anything without a real description. 120 remained.
Split by creation date
40 created in 2026; 80 before.
Computed rates
Stars divided by days since creation. A snapshot, so it favours anything with a recent spike.
Counted licences and languages
Straight counts on the API's own fields, not inferred.
Are these numbers current?
They are a snapshot from 1 September 2026. Star counts move daily, so treat them as a picture of that date rather than a live figure.
Why does stars-per-day favour new projects?
Because it does — a project that trends for a week scores enormously on this measure. It is useful for spotting momentum, not for judging quality or durability.
Does a star mean someone uses the project?
No. A star is a bookmark. It measures attention, which is worth knowing about, and nothing at all about whether the code is good.
Can I reproduce this?
Yes — the method is above and the GitHub search API is public. That is rather the point of publishing how it was done.
Save this summary as an image or share it.
AICreatorHub Team
The AICreatorHub editorial team is a group of hands-on AI practitioners, writers and developers based in India. We test AI tools and models ourselves, track official releases from OpenAI, Anthropic, Google, Meta and xAI, and translate them into simple, India-first guides in English and Hindi. Every article is written for real Indian use cases — pricing in rupees, free-tier tips and practical, tested steps — so you get accurate, up-to-date and genuinely useful AI information.