The Problem

Today, many came to a realisation that simply scaling to ever larger LLMs will not bring us to superintelligence or AGI. Larger models do get smarter, but also much slower. But all of them share a critical flaw - imperfect reliability, which forces us all to choose between keeping a human in the loop and taking a chance of producing slop.

At Lab86 we believe existing models (both open- and closed-weight) are often already smart enough, and our focus is on making them more reliable (especially with long context). This is the missing step required for AI to be trusted to consistently complete practical tasks in the real world without supervision.

Our vision of the future of AI and humanity

Our goal is to enable AI tools that can get the job done so that no one has to verify (and take responsibility for) the results. Humans are not very good at spotting rare mistakes anyway. The human is not getting redundant - his role is to ask things to be done and collaborate on the requirements. So as AI gets more and more done, humans will have to learn to want more.

Company & contacts

Lab86 AG is a privately funded startup in Zug, Switzerland, where we enjoy a very favorable business climate and gorgeous nature.

General inquiries: [email protected]
Business development: [email protected]
Investor relations: [email protected]
Interested in working with us?

Lab86 AG
Bahnhofstrasse 19
6300 Zug, Switzerland