Define the test.
Name the question, measure, limits, and decision rule before reading the outcome.
AmberContinuumAI collaborative research labResearch method · Amber Anson
Coherence across systems.
My research asks how the parts of a system relate, what holds those relationships together as conditions change, and how to tell when they shift. I pursue that question across cognition, human–AI collaboration, software, materials, and physical experiments. Lived experience and pattern recognition shape what I notice; mentorship shaped how I see software architecture. I engineer the exchange between my cognition and distinct AI architectures to turn an idea into a method, instrument, or test.
A path from question to claim
- 01Frame
Name the question and its limits.
- 02Exchange
Work across human intuition and AI reasoning.
- 03Test
Build an instrument and change the conditions.
- 04Record
Publish what the evidence supports, including revisions.
How I investigate
A question carried across domains.
Coherence is the subject of the research. Cybernetic engineering is part of how I investigate it. I originate the questions and develop the core ideas, then work with distinct AI architectures to compare explanations, expose assumptions, develop code and protocols, and test what an account can withstand. Their reasoning and execution materially shape the work.
I decide which interpretations remain open, what the evidence supports, and what is ready to publish. The public record should let someone else trace the path from question to method to result, including where the path changed.
The practice
From signal to record.
This is a working sequence, not a claim that every project followed an identical script.
- 01
Notice the pattern
Write down what recurs before deciding what it means.
- 02
Set the frame
Name the question, purpose, limits, and a possible miss.
- 03
Engineer the feedback
Design how human intuition and AI architectures exchange information, challenge one another, and change the next decision. Record what each contribution does.
- 04
Keep disagreement visible
Record where answers conflict and which assumptions caused it.
- 05
Find what holds up
Compare the accounts without hiding a meaningful difference.
- 06
Give the idea a clear form
Turn the surviving account into definitions, equations, or steps.
- 07
Make something testable
Build code, a tool, a protocol, a dataset, or an inspectable model.
- 08
Change the conditions
Perturb the setting and watch what holds, shifts, or fails.
- 09
Keep the full record
Keep sources, versions, negatives, corrections, and abandoned paths.
- 10
Decide what comes next
Publish, revise, pause, or let the idea go.
The discipline
What a public claim has to carry.
Keep disagreement visible.
Record competing accounts and changes in conditions. Check whether an instrument can distinguish what it claims to measure.
Keep the full record.
Show negative outcomes, withdrawals, corrections, and abandoned paths beside later work.
Preserve human closure.
Ask whether a person can still notice, decide, act, check, correct, and recover if the AI path is wrong or unavailable.
Open the work
The method in public.
These outputs show different parts of the practice. Each carries its own status and limits.
Working specification
Open the record ↗Coherence Systems Architecture
Purpose, limits, dependencies, and responsibilities made explicit as people and AI work together.
Published framework · transfer untested
Open the record ↗Human Closure
Checks whether a person can still notice, choose, act, check, correct, and recover when an AI path changes or becomes unavailable.
NeurIPS 2025 competition entry
Open the record ↗Crystalline Polymer Dataset
A public dataset built to test what human intuition and AI reasoning could produce together. Predictive improvement still needs reproducible comparison.
Recorded outcomes and corrections
Open the record ↗QRT Bell Test
Filed predictions, IBM Quantum job records, negative outcomes, a withdrawn verdict, and later results kept as separate findings.