Back to the map
SuperhumanHealth AI  ·  How it compounds
Page 3 · How it compounds

Most apps decay into noise as they scale. This one gets more honest.

MIST learns on three loops at once. Here's how a single experiment ends up making the whole system smarter — and why a loud, wrong result can't hijack it.

The shape of it · widen, then narrow

Three loops. Two gates. Evidence flows in freely — and gets canonized slowly.

YOUR LOOP experiment → data → refine ↺ 1 2 3 Tuned to YOU least-invasive, improving RIGOR GATE low bar · permeable n=1 enters as honestly-weighted evidence EVERYONE'S LOOP pool → converge → strengthen ↺ 1 2 3 Shared proof stronger every cycle PROMOTION GATE high bar · selective must converge AND replicate to graduate VALIDATED FRAMEWORKS graduate → ground Kai → baseline ↺ 1 2 3 Validated grounds public Kai Validated frameworks raise everyone's starting floor — every personal loop begins smarter. New users benefit on day one.

↔ scroll the diagram sideways

Loop A · the personal loop
Your loop

MIST gets better for YOU.

Through repeated self-experiments it builds a profile unique to you and converges on the least-invasive strategy that actually works for your body — not generic advice.

Every cycle: more personalized
Every cycle: less invasive
RUN DATA TUNE YOU
Loop B · the collective loop
Everyone's loop

Your data makes everyone's MIST smarter.

Thousands of honest n=1 results, pooled, reveal patterns no single person could see. Convergence across people is what turns scattered self-experiments into real, weighted evidence.

PATTERNS converge
Loop C · the framework loop
Validated frameworks

The strongest proof becomes knowledge the system can stand behind.

When evidence converges and replicates, it graduates into a validated framework that grounds the public Kai product — and resets the baseline every future user starts from.

public Kai new canonical baseline
The safety design

We ingest freely. We canonize conservatively.

GATE 1 low bar — most enter GATE 2 high bar — few graduate framework
Gate 1 — Rigor low bar · ENTER

Anyone's experiment can enter — but only as evidence tagged with how strong it actually is. Nothing enters as "truth."

Gate 2 — Promotion high bar · GRADUATE

Only evidence that converges across many people AND replicates can graduate into a framework.

That asymmetry is the whole trick. A loud, wrong result gets in as weak evidence — and dies at the second gate. The system compounds what's true, not what's loud.
The honest part

The promotion gate is where this lives or dies.

Set that bar too low and you've built a confident misinformation engine with extra steps. Set it well and it's a moat nobody can copy. We treat the gate as a hard research problem, not a slogan — because honest uncertainty is the product.

"We win by showing the work, not by sounding certain."
The through-line of the whole series — from the weights-vs-context architecture, to one honest day, to the system that compounds it.