Research
Research at NeuronX.
We study the questions AI has to answer before it can be trusted with real work.
We publish what we learn.
Our writing shares the questions, the methods, and what we still don’t know.
We test in production.
Research that works ships in AutoScale, where real companies use it every day.
We work with universities.
Including a research collaboration with IIT Bombay.
Lead programs
Where we put most of our work.
NeuronX Fuse
How do AI agents get attacked, and how can they defend themselves?
Read more ActiveNeuronX ARC
How long can an agent work on its own without drifting?
Read more ActiveAGI research
Can general intelligence reason well with limited data, limited compute, and human values?
Read moreExploratory
Writing
From the lab.
Research · August 12, 2024
Bots: the invisible workforce behind half of internet traffic
Bots make up about half of web traffic. A look at good and bad bots, how CAPTCHA fights them, and what that costs in privacy.
Research · February 16, 2024
Bio-inspired vs. mathematical AGI: BICA and AIXI compared
Two roads to general intelligence. One copies how brains work; the other defines the ideal agent in math. What each offers, and why we may need both.
Collaborate
Work with us.
Professors, labs, and companies can propose a joint project. Tell us what you want to study together.