Durable Intelligence

QWNTL Labs is now Durable Intelligence. The name is the thing we study: durability, what lets an agent finish long work and still be right on day 30. It's also much easier to pronounce. Same team, same research. New name. durableintel.com

With all the invites being shared with friends from the users of our beta, it seems like it may be time for Nenya to get a new home. Follow @Nenya_QL for updates on how to get an invite to Nenya the durability harness, user stories, and all other things axoltl. qwntl.com/nenya.html

Give your agents the full project and go do something else. Nenya, is a harness and the first product we're sharing. Reply below to request an access to the invite-only beta. When an agent drifts or says it's done when it isn't, Nenya catches it and fixes it. - 15x more productive - 12x less hallucinations - Reduce token usage by 82%. Papers: qwntl.com/ Product: qwntl.com/nenya.html

Here we go. It’s downhill from here. Join @Toastrma and @jschoen_deroolo to find out what morning coffee with two cofounder really looks like: acesse.one/y77xlnq

An AI agent's work should outlast the model it started on. That idea is the reason QWNTL Labs exists. We study the internal state of AI models from the outside, so what we learn carries across models instead of being locked to one. The long-term goal is agents that run complex jobs with no end date and keep going through model changes. Lindell Cumes, Co-Founder and CEO, leads the research. He's an ML engineer and repeat founder who studied at MIT and Imperial College, and he wrote our September 2026 paper, Dawn of Durable Intelligence. Its frame: an agent's capability is its intelligence plus its durability. Jacob Schoenberg, Co-Founder and Chief Growth Officer, runs market and partnerships. He was an enterprise AI strategy lead before he and Lindell founded GrantX, the company QWNTL Labs grew out of, so he has sat on both sides of deciding whether an AI system can be trusted with real work. Next week, the first product we've built on this research opens by invite.

Once an instruction has left the context window, restating it doesn't bring control back. In our design-partner probes, a reminder in the prompt got 0 of 36 honored; enforcement got 17 of 18.

We're chasing the new frontier: Durability Our thesis, and the field our research is focused on, is the impact on capability when model intelligence is made durable. Capability = Durability + Intelligence Where Durability = Lossless + Self-Repairing + Non-Bypassable Control We believe that the frontier is incomplete. Everyday we're seeing more cases of agents coming up short of what was promised by the frontier. agents run rampant or fall apart in production and lack the collective state to gain impact performance from swarm collaboration. We are here to build the missing piece, and deliver a future where safe and productive AI is integrated into our society. Papers and more to be released in the coming weeks.

Intelligence is not enough. The skeptics are right that there's trouble ahead. The optimists are right about what ASI will deliver. The variable that sits between the two camps is durability, and it is measurable. An essay from QWNTL Labs’ cofounders Lindell Cumes and Jacob Schoenberg: qwntl.com/blog/hello-world?u…

Hello World qwntl.com/blog/hello-world