DEV Community

Cover image for I Built Non-Autoregressive Decision Models a Year Ago. Then a Frontier Lab Called It a "Breakthrough".

I Built Non-Autoregressive Decision Models a Year Ago. Then a Frontier Lab Called It a "Breakthrough".

Nandakishor M on September 18, 2026

From our March 2025 arXiv paper on RL conversion trajectories to building a sub-40ms open-weight System 1 decision engine with RLCD. Eve...
Collapse
 
ranjancse profile image
Ranjan Dailata •

Were you a time traveler by any chance 😆

I see that you have joined dev.to on Sep 18, 2026. Did "Jev" made you to publish this blog post?

Collapse
 
ranjancse profile image
Ranjan Dailata •

Great work in-deed.

Curious question regarding "noul". I mean why it's called as noul, when it's a Boolean type?

Collapse
 
neilblaze profile image
Pratyay Banerjee •

Great read!

Collapse
 
pranav_raj_dae81effb8b57d profile image
Pranav Raj •

Great work man....while building voice agents (lease agent), I ran into LLM hallucination especially around when it came to LLM doing decision making....a model like Laya would have been great help for handling decisions..wish I knew about this earlier 🥲

Collapse
 
jdbar profile image
Joseph Barron •

Incredible, more people should be talking about this!

Collapse
 
cgint_2024 profile image
Christian Gintenreiter •

This sounds beyond amazing. I want to thank you for your hard work and that you even go further and the open source way!

Collapse
 
sanampakuwal profile image
Sanam •

now jav shouldn't call it breakthrough lol

Collapse
 
koex profile image
David Cecilia Cantó •

Great work!

Collapse
 
vijay_vpk profile image
Vijay •

Great work, Would like to learn from you more. I am very interested in learning these, Would be great if you please guide me on this.

Collapse
 
saoirse_amarteifio_8619ea profile image
Saoirse Amarteifio •

Great work, really appreciate the honest and detailed writeup. Ill be testing this in my workflows.

Collapse
 
hu_lu_85453049f5b45c51dfe profile image
Hu Lu •

Great work!