Proof
Our AI Employee has already contributed to scientific discovery and to building AI-native start-ups.
As a co-scientist, it helped design a state-of-the-art algorithm for one of computer science's most difficult problems. Timenix, an AI-native start-up, builds its agents with the same technology. And every day, ours helps run our own work.
Scientific discovery
It helped design a state-of-the-art algorithm for one of computer science's most difficult problems.
The algorithm solves the Hamiltonian cycle problem: find a route through a network that visits every point exactly once. The possible routes multiply faster than any computer can check them.
Co-scientist
Our AI Employee's part in the work: planning, design and implementation, alongside our mathematicians and developers.
See it for yourself.
In the live demo you can run a standard benchmark and twenty deliberately difficult graphs, or upload your own of up to 2,000 points. On a difficult graph of 1,002 points, our solver finds a route in seconds, in 177,706 steps. Beside it, the demo shows the best-known estimate for quantum search: about 1045 steps.
Try the live demoAn AI Employee worked as a co-scientist with our mathematicians and developers, above all on the planning, and helped them design the algorithm. The algorithm has scored 100% on all existing benchmarks for the problem, a success rate no other algorithm has reached yet.* It also made our solver, already considered state of the art, up to 1,000× faster.
* A 100% score does not prove that the algorithm always succeeds. It is a score on the benchmarks that exist today. The results are not yet published; you can test benchmark graphs yourself in the live demo.
Why it worked
Memory, not a blank slate. It started from what our mathematicians and developers already knew, and from the documentation they wrote, instead of starting from nothing like a generic agent.
Self-improvement loops. It tried better and better ideas, on its own and in discussion with the team, until one worked. Then it did it again, and the gains compounded.
Learning from experience. It got better as it did the work and as our scientists worked with it: it corrected its mistakes and remembered how things work, building a knowledge base from its own experience.
It was built the way Brainoid IO's AI Employee is built: on memory, with self-improvement loops.
AI-native start-ups
It helps a start-up build its agents.
Timenix, a start-up from Accelerate Cambridge at Cambridge Judge Business School, builds a digital protection agent. It is designed with ADHD and attention challenges in mind, and it helps people protect their attention from unwanted pulls. Our partner uses the technology behind the AI Employee to build its technology stack and its agents.
Our partner
Timenix
A digital protection agent, from Accelerate Cambridge.
Everyday work
What ours do for us, every day.
Automation with judgement, built on memory.
Opportunities
- Watches for grants and scholarships, and flags the ones worth our time
- Drafts the proposals with us
Advice
- Creates clones of great thinkers and business leaders, and asks what they would do in our place
- Points out the blind spots in our plans
Personal
- Tracks our health intelligently, across everything we record, not just one app's numbers
- Plans each day around what matters most, and updates the calendar to match
Records
- Writes up our product's progress as it happens
- Keeps a record of every piece of work done
What made the difference: memory, self-improvement and experience.
The co-scientist started from what the team already knew, improved its ideas in self-improvement loops, and got better as it worked: it corrected its mistakes and remembered how things work. Brainoid IO gives you the same: your knowledge as a memory on your computer, and an AI Employee that learns as it works and keeps getting better.
The app is free for personal use (local models are free; optional cloud AI is paid for separately).
Limited alpha 0.0.1 · Apple silicon, macOS 26 or later