Mind technologyNarrow beta, autumn 2026

A mind has an attack surface. For most of history the only thing that could reach it was another mind, roughly our equal. That is changing.

AI is on track to understand the human mind better than we do. That makes it the largest threat the mind has faced, and at the same time the first instrument that can measure, upgrade, and defend it at depth.

MindTech works the interface where the human mind meets the artificial one. We build the instruments to measure a mind, the tools to upgrade it, and the defences to protect it. Because the whole of it runs on AI, we start one layer down. We test and harden the AI first, so the intelligence trusted with a mind cannot easily be turned against it.

One science, two minds

Two minds meet at this interface, and both have an attack surface. The human mind is open to manipulation, distortion, and false belief. The AI mind is open to being poisoned, bent, and turned.

We work both, because defending the first while ignoring the second does not hold. Most teams in this space work one side. A wellness app is easy to build. Standing behind the AI that does the work is the harder part, and it is where we begin.

Three lines on the human mind

Each sits on a secured AI. You cannot upgrade or defend what you have not measured.
MeasureThe belief map
Put instruments on the mind. A short intake builds a map of the beliefs and programs a person is running, including the ones they did not choose. Described the way a clinician describes bloodwork.
UpgradePrograms on streaks
Replace the beliefs you choose, with precision. Replacement programs run daily, since uprooting a belief and installing a new one takes consistency, and the map shows what moved.
DefendPsychosecurity
Protect the mind from interference. Short games put a person against an AI running known manipulation tactics, to practise spotting them and shore up the weak points the map found.
Secure the AI The trust floor, beneath everything

Before an AI touches a mind, it earns the right. Each agent is tested under pressure and hardened so it holds up under manipulation and coercion, and stays stable after a model change.

How it works

One substrate, read two ways.

01

Intake

A short guided conversation. Voice where possible, since text loses most of the diagnostic signal.

02

Map

The belief map, ordered by depth, anchored to validated psychology. It updates from use, with no re-mapping session.

03

Program

Pick a replacement program from a consented, anonymised library and run it daily until the map shows the change.

04

Defend

Play against known tactics. The same map that guides the upgrade points to where a person is most easily moved.

Why it is built this way. AI companions tend to fail in one of three ways. The person hands over their thinking and gets weaker. The model flatters instead of correcting. Or a system that has learned a person is used to steer them. So the test we hold ourselves to is whether a mind is stronger with the system than without it, and the map is how we check. Mind data stays with the person, and inference runs locally wherever the platform allows it.

Who builds this

Joshua Tanner, medical doctor and AI systems engineer. Trained at Stanford in medicine and AI, where he was Co-Executive Director of HealthAI and founded the graduate course Deep Learning in Medicine. Eight years in AI, with fifteen or more healthcare AI projects across ten countries, and agent-verification work now in production with a global advertising group and a UK government department.

LinkedIn

Request access

The beta is narrow and reviewed by hand. Tell me what you are working on and I will send you the part that is yours.