FULL STORY
Jev's Launch Sparks Copycat Claims
A former OpenAI employee unveiled the model Jev after two years in stealth. ConvAI's founder quickly disputed the breakthrough claims, saying an identical model shipped a year earlier.
2026-09-20 ~ 2026-09-21 · 2 episodes · 8 posts
Episode 1 · Founder Claims TypeSafe AI's Jev Was Preceded by His Own Non-Autoregressive Model (2026-09-20, 5 posts)
Around September 20, Nandakishor Mukkunnoth, founder of ConvAI Innovations, published a post questioning the "breakthrough" framing of Jev, the non-autoregressive fast decision engine released in September by frontier company TypeSafe AI (founded by ChatGPT co-inventor Diogo Almeida), claiming he had produced the same kind of work a year earlier — a post widely reshared and discussed by multiple bloggers.
Confirmed
- According to Mukkunnoth, in March 2025 he published an arXiv paper on non-autoregressive decision models (arXiv:2503.23303), followed by a second paper (arXiv:2510.01237) covering related methods guided by reinforcement learning
- He open-sourced model weights and a SaaS sales dialogue dataset on Hugging Face, and released a PyPI package (PyP…) for others to use and reproduce
- He argues TypeSafe AI's Jev is the same concept as his work, but was eclipsed by the hype of a frontier lab's launch and labeled a "breakthrough"
Unconfirmed
- The actual technical similarity between Jev and Mukkunnoth's model rests solely on Mukkunnoth's account; TypeSafe AI has not publicly responded
- The materials do not provide the exact release date of Jev or the original promotional text
Why it matters
- The incident has reignited debate over attribution and visibility imbalances between "frontier labs vs independent researchers": the same non-autoregressive decision model concept got limited traction when released by an independent researcher, yet was hailed as a breakthrough when released by a famous lab
- Mukkunnoth has open-sourced papers, weights, a dataset, and a PyPI package, providing verifiable grounds for outside independent comparison of the two
- Founder says he built non-autoregressive decision models a year before a frontier lab called the same idea a breakthrough — JiliJeanlouis · 2026-09-20
- Dev says he built non-autoregressive decision models a year before frontier lab's Jev — JFPuget · 2026-09-20
- Founder claims TypeSafe AI's Jev copied his year-old non-autoregressive decision model — pandeyparul · 2026-09-20
- Founder says he built non-autoregressive decision models a year before TypeSafe's Jev, but was ignored — Paimaamu · 2026-09-20
- Founder says frontier lab's Jev copies his 2025 non-autoregressive decision model paper and open weights — abhijithneil · 2026-09-21
Episode 2 · ChatGPT Co-Inventor Launches Decision-Only AI Model Jev (2026-09-21, 3 posts)
A self-described ChatGPT co-inventor launched Jev, a free decision-only AI model that generates no text, claiming it is 20-200x faster and 40-400x cheaper, trained via a new method called RLCD.
- ChatGPT co-inventor launches Jev, claiming 20-200x speed and 40-400x cost gains — BLUECOW009 · 2026-09-21
- ChatGPT co-inventor launches Jev, a decision-optimized model claiming 20-200x speed and 40-400x cost gains — soumitrashukla9 · 2026-09-21
- ChatGPT co-creator launches Jev, a free AI that never writes a word—only decides — PrajwalTomar_ · 2026-09-21