Indie dev says he built non-autoregressive decision models a year before TypeSafe AI's Jev
HowDevelop · x · 2026-09-22
Nandakishor Mukkunnoth, founder of ConvAI Innovations, published a long post claiming his prior work anticipated the hype around Jev, the non-autoregressive decision model from TypeSafe AI (founded by ChatGPT co-inventor Diogo Almeida).
- In March 2025 he released an arXiv paper (2503.23303) on RL-based sequence conversion, along with open weights, a SaaS-sales conversation dataset, and a PyPI package
- A second paper (2510.01237, Sept 2025) formalized schema-based decisions guided by reinforcement learning — the core was RL, not embeddings or autoregressive LLMs
- In Sept 2026, TypeSafe AI launched Jev proposing the same non-autoregressive decision concept as a breakthrough — with no papers or open weights
He has now released Laya: an open-weight System 1 decision engine with sub-35ms latency, RLCD training, routing across 100+ languages, and state-of-the-art calibration on shared benchmarks.
More from Models
- Google's Physical-Engineering's Last Exam: all 14 frontier models score below 50% — dl_weekly · 2026-11-12
- Commit leak reveals unreleased Gemini 4 variants: Argon, Hydrogen and Helix (unconfirmed) — BlackHC · 2026-11-11
- Rewiring just 89 of 442,368 neurons in Qwen3-8B matches full SFT for a persona — maxsloef · 2026-11-11
- Unsloth ships Qwen3.8-27B Dynamic V3.1 GGUFs with 5-10% accuracy gains — danielhanchen · 2026-11-11
- Zvi: being forced back to an older model is the best reminder of AI progress — TheZvi · 2026-11-11
- Screenshots Made a 4B Browser Agent Worse: A Postmortem With Real Numbers — Stunning-Sherbet1853 · 2026-11-11