Survival strategies involving offline AI.
NeuroCognica is a research initiative and software company. We build Sentinel-governed systems for emergency readiness, local creative studios, games, archetypal profiling, practice and durable memory.
The mission
NeuroCognica is a research initiative. Its first research lane is survival: how offline AI can help a person prepare before an emergency, stay calm during one, and keep useful knowledge available when ordinary networks fail.
Survival is the first product priority. Art studios, games, reflection and study matter because life is worth protecting. Sentinel Online and EOAI-MGS exist so readiness is not an afterthought: water, food, power, shelter, communication, health, lawful coordination, and calm decision-making before comfort.
Our engineering direction is built around durable memory: Rust where speed and stability matter, BLAKE3 hashes, Merkle Mountain Range structures and RocksDB-backed stores where they fit the product. Some of that is already implemented in witnessed systems; some remains research direction. The rule is simple: say what is real, build what is missing, and never confuse vision with shipped behavior.
Get ready first. The Sentinel checklist is the front door because preparedness is not decorative. You can create, play, reflect and train better once the basics are handled.
The argument
The industry has settled into a trade it rarely states out loud: you get capable AI, and in exchange your work leaves your machine, your data trains someone else's model, and the service can be revoked. Safety is asserted in language engineered so that no test could ever prove it wrong.
Every part of that trade is negotiable — and the only honest way to argue it is to build the counterexample and let people check it.
The Shaolin thread running through NeuroCognica is not decoration. It is a build discipline: practice before claim, restraint before power, correction before pride. The Dao and the Middle Path show up here as engineering pressure — systems should act with balance, refuse excess, keep records, and leave the human more sovereign than they were before.
Kung fu is life in the practical sense: disciplined practice under pressure. Breath, posture, attention and steadiness are survival tools before they are aesthetic ones. Chirox exists in that lane: not as medical advice or certified combat training, but as a practice surface for strengthening, stretching, reading, recording and learning to remain calm enough to help.
Governed
Actions are checked before they execute, refusals are recorded, and the policy is published. Not "responsible AI" as a slogan — a written taxonomy you can read and a measured result you can check.
Private
Local-first wherever it can be. Your prompts and your work stay on your machine, and there is no server for us to read them from even if we wanted to.
Efficient
Running on one desktop, not a datacentre. If it needs a rack of GPUs to be useful, it is not sovereign — it is a subscription with extra steps.
Honest
We publish the limitations, the measured failure rates, and the findings we have not fixed yet. A claim nobody could falsify is not a claim.
The standard we hold ourselves to. Our safety system was measured against an adversarial set and the numbers published — including a regression we introduced while fixing something else, and two findings still open. Most vendors publish neither.
How the Sentinel works · The full protocol, with the numbers
Use something now
These are the working public surfaces. Some are commercial, some are free, and some are experimental, but each gives a visitor something real to do on the site instead of asking them to believe a roadmap.
Sentinel Online and EOAI-MGS — readiness
The survival-first lane: signed-in checklist, private inventory, offline sync, public-source monitoring, and a Windows app meant to remain useful when normal services fail.
ChronoSophia — available
A local art studio for building editable 3D scenes on your own Windows computer from a sentence. The first thing NeuroCognica took end to end: designed, built, governed, documented, and sold.
Archetypes — game research
A Rust/Bevy council game where the player occupies the Witness seat, speaks with archetypal agents, and can receive Chronos-generated images directly in the experience when the local pipeline is available.
Mirrorborn — live quiz
An archetypal profiling and reflection system that reads what you write about yourself and builds an explainable five-station profile, keeping your history so you can see what changes. Not a clinical tool, and it says so first.
Chirox — live dojo
A signed-in Cloud Dojo for books, read-aloud practice, bookmarks, personal Dojo Records, Master guidance, and camera-based form reflection. It is training support, not a substitute for a human teacher.
The Sentinel — running
The governance layer underneath everything: authority, content policy, and execution control as three separate mechanisms. It refuses before it generates, fails closed, and seals every decision.
Research — ongoing
Archetype structure, cognitive sovereignty, Shaolin virtue, and what a constitution for a machine would actually have to say. The thinking the products are built out of.
What runs through all of it
One rule: a capability is not finished until it works from the interface a person actually touches. A passing test is not a shipped feature, and a green tick with no path behind it is a lie.
Who this is
NeuroCognica is one person — Michael Holt, a registered sole proprietorship in McLean County, Illinois. Not a lab, not a funded startup, not a team hiding behind a plural pronoun.
That is disclosed rather than buried, because it changes what you should expect: support is a direct email to a human, releases arrive when they are ready, and none of this is propped up by a marketing department. It also changes what the work is worth — every claim on this site has one person's name against it.