Home

The Vision

What I'm drawn to. The stack I see coming.

The Stack I See

THE PROBLEM

Bottleneck = Human Reception

AI can produce anything. Humans can't absorb it fast enough.

LAYER 1: DATA

Universal LLM History (Sparkii)

Every conversation you've ever had. Queryable. Portable. Yours.

Not locked in ChatGPT, Claude, Gemini. Exported, unified, searchable. I built this for myself — 366K messages queryable in 256ms.

LAYER 2: IDENTITY

SMAT (Human Verification)

Proving you're human by how you think, not clicking boxes.

CAPTCHA is dead. Behavioral patterns, thinking rhythms, correction styles — these prove humanity better than image recognition.

LAYER 3: PROFILING

HPI (Cognitive Profiling)

Understanding how each mind works. Monotropic, polytropic, somewhere between.

Not personality tests. Not MBTI. Actual cognitive patterns extracted from conversation history. How you think, not who you say you are.

LAYER 4: INTERFACE

Inference Time UI

Interfaces that adapt to the person using them, in real-time.

Same app, different experience per user. Cognitive profile informs information density, navigation style, explanation depth.

THE RESULT

100% Human Agency

AI amplifies. Human directs. Control never leaves the person.

Why I Believe This

I didn't start with a vision. I started with a problem: my brain doesn't work like others'. Monotropic focus. Hard to context switch. Information gets lost.

So I built a prosthetic. 366K messages queryable. And in building it, I saw what's coming.

If this works for one non-neurotypical brain, it works for all brains. The infrastructure I built for necessity becomes the foundation for something larger.

ZMAT: Human Verification

ZMAT (זור מרא עשה טוב) — "Turn away from evil, do good." First mentioned 2023-08-01. Answers TWO questions:

1. Is this a HUMAN?
vs bot/fake/machine
2. Is this human GOOD?
behavior score

"everyone should start off with a neautral scrore and then accumulate either possitive of negative points on the basis of there interaction"

2023-06 | Verify Human Authentic AI | chatgpt

The Game Theory

Fake reporting hurts the REPORTER, not just the reported. Good reports → your score goes up. Bad/fake reports → your score goes down.

"nobody wants to lie because it will just affect their score negatively"

2025-09 | claude-code
// ZMAT Logic
1. Everyone starts NEUTRAL
2. Interactions generate ratings
3. Anyone can report (good or bad)
4. All reports IMMUTABLE
5. Good reports → score UP
6. Fake reports → score DOWN
7. High Zscore = verified human + good behavior

What I'm NOT Saying

This isn't a startup pitch. I'm not raising money. I'm not asking you to invest.

This is what I'm drawn to. Problems that pull me in. Domains where I want to go deep.

If you're building in these spaces and want someone who's been thinking about this for 3 years, maybe we should talk.

The Domains

HPI: The 5 Framings

HPI (Hyper Personalized Interface) isn't one thing. It's been framed 5 different ways over 3 years:

1. Filter Between User and World

"now take a look at my hpi a hyperpersonalised api i want each human to own that sits between them and the world and creates a personalised filter between the world and them based on there preferences"

2024-04 | Making Organizational Contributions Accessible | chatgpt
2. User-Owned Algorithm Control

Taking back control from platforms. USER owns the algorithm.

3. Real-time Observation (Not Historic)

"i dont want to train the hpi on historic data rather i want it to observe in the background and build the backend model of each human using smart tech and creating a live and adaptable image of each humans prefs"

2024-04 | Making Organizational Contributions Accessible | chatgpt
4. Traffic Control In & Out

Bidirectional: filters what comes IN + what goes OUT. Ultimate user data access control.

5. Cognitive Profile → UI Rendering

"no no no we are trying to map and research and explore differnt cognitive learning styles ect to map to uis"

2025-12 | intellectual-dna | claude-code

The 3-Year Evolution

YearTermMeaning
2023hyperpersonalizationContent adapted to user
2024hyper personalized APIUser-owned filter
2024real-time observationLive learning, not static
2025inference time computeProcessing at moment of use
2025generative UIInterface generated, not designed

"now you can understand inference time ui for everythuijng and how it could work within kbs not mbs or gbs or tbs"

2025-12 | intellectual-dna | claude-code
Traditional UI
Designer → Static interface → All users same
vs
Inference Time UI
Cognitive profile + Content → Generate interface → Each user different
KBs not MBs