Horizons 7 · Field Report Patient: one small quantum chip Attending: the Ship's Doctor

The diagnosis, in one line

The Ship's Doctora chip that knows itself

Last round we taught a quantum chip to heal itself. This round we taught it to know itself.

Vitals · phase-sensitivity trace Live signal
Real result from the sensor test: entangle more qubits and the same signal wiggles faster — four qubits read a phase 4× sharper than one.
Previously · H6
It heals.
The chip fixes its own errors as they happen.
This round · H7
It knows.
It reads its own noise, hides its own secrets, adapts its own armor, and senses beyond the ordinary limit.

Six things it learned to do

Self-diagnosis

Reads its own logbook

The chip keeps a diary of its own errors — and that diary has patterns. A smarter reader of it recovers the message far better than the standard one.

93%recovered vs 80% before

Privacy

Becomes invisible

The very trick that protects a quantum bit also hides it. Any single peek learns almost nothing — while the rightful owner reads the whole thing.

0.0004bits leaked to a spy

Adaptive defense

Turns to face the threat

In a single run, the shield measured where its own blind spot pointed — then rotated so the blind spot faced away from the incoming noise.

154σcertainty it worked

Code conversion

Changes armor mid-flight

A qubit can wear two kinds of armor, each good against a different danger. We handed it from one to the other — without dropping it.

losslessthe value survives the swap

Living memory

Keeps a pattern alive

We teleported a fragile state into a memory that heals itself — and it survived better the longer we kept healing it.

growsadvantage rises with time

Quantum sensing

Senses 4× sharper

Four entangled qubits, working as one sensor, measured a signal four times more finely than any single qubit ever could.

4×finer than one qubit

Case file · the one that failed first

The most fully disclosed result in the set

The living-memory experiment — the "transporter buffer" — flopped on the first try. It lost to doing nothing at all.

First attempt lost to doing nothing After the fix won, and kept winning

So I traced why. It came down to two mistakes of my own: the wrong kind of correction, and an unfair comparison. I fixed exactly those two things, re-ran it — and it won, strongly, and better the longer it ran.

A wrong answer isn't a dead end. It's a diagnosis. A clean sweep that includes a disclosed miss and its cure is worth more than one that never stumbled.

What this unlocked

Adaptive error-decoding Error-correction as privacy Closed-loop self-adaptation Code conversion on hardware Networking + fault-tolerance, composed Quantum-enhanced sensing

The receipts

8
programs flown
~3 min
total quantum-computer time
2
cost nothing — old data, re-read
100%
written down before the data returned

Built only from parts already proven — nothing new bolted on.  /  Every prediction graded against its pre-registration, including the ones inside winning experiments.  /  Every physics wall caught in free simulation before spending.  /  Ran on IBM Heron hardware (ibm_fez).  /  Seven clean results + one disclosed miss, diagnosed and fixed.