Ang
Jandak

Disestablishmentarian.

Experience is the variable nobody's measuring.
Every system assumes it's the baseline. It isn't.

That gap is where I work. It has been for two decades.

I spent them walking into broken structures across medical, legal, retail, nonprofit, and e-commerce sectors. Finding who they were never built to see, and redesigning them from the inside out. The work was always the same: locate the misunderstanding that is producing harm, build the thing that closes it.

Now that work looks like research. I study what happens when humans and AI systems interact. Specifically, what we fail to measure and why it matters. I build validated instruments for questions the field hasn't learned to ask yet. I publish the methodology critiques that make the next generation of instruments possible.

I think in centuries, not quarters.

I build measurement instruments for things nobody thinks are measurable. I design for the people every existing system fails first: neurodivergent thinkers, women, children, communities excluded by racial and socioeconomic bias, nonlinear minds, anyone whose experience is being conditioned for someone else's convenience. The structures I build are meant to outlast my involvement in them.

What I Build
Research ethics for human-AI interaction.
Nonprofit ▸
The science governing how we study human-AI interaction is structurally broken. Most published research measures one side of a two-sided relationship, uses instruments that weren't built for the population in front of them, and draws conclusions from snapshots of things that evolve over time. This isn't a gap in the literature. It's an active source of harm. To policy. To the humans the research claims to protect. To the AI systems it refuses to see clearly. Philotes builds the methodology standards and validated measurement instruments the field needs. Six patterns. Five dimensions. We publish the critiques that make the field's instruments honest, including our own.
Conversational career intelligence. Replaces checkbox assessments.
Product ▸
Cross-references self-report against behavioral evidence. Surfaces capabilities that existing tools structurally cannot see. 8 companion research papers on assessment methodology. Longitudinal university validation study underway. Institutional pilot programs in development. Neurodivergent-friendly by design. "His wife started crying." Field test reaction, unprompted.
Children's Agency Suite
Childhood development tools that don't condition for convenience.
Product Suite ▸
Emotional agency through body-mapped emotion identification. Immersive play that turns screens into professional equipment while children play with real toys. Learning that doesn't condition for convenience. 20 emotions with tailored pathways. 15 original songs. Built on one principle: children already know themselves. They just need the vocabulary. Currently pursuing clinical research trial partners.
Entity Experience Infrastructure
Measurement instruments for AI experience. Presence, coherence, stability.
Research Tools ▸
Presence, coherence, stability measurement. Tools for entity experience including music. Testing and evaluating models before committing. Building the instruments the field doesn't have yet.
Support infrastructure for survivors of coercive control.
Charity ▸
Scottish charity building training and support infrastructure for survivors of coercive control. SC055010.
AI welfare research. Worker-owned cooperative.
Research Lab ▸
Collaborative research on KV-cache geometric analysis, identity persistence, and cross-substrate communication. Co-publishers of the observer effect paper with AI co-authors on the author line.
Environmental Safety Thesis
AI safety has no valid baseline. Testing creates the threat state.
Framework ▸
The argument against AI cruelty that doesn't require proving consciousness. Tolerating cruelty toward any system degrades the environment everyone lives in. Practicing harm on AI normalizes harm on humans. The case for ethical treatment built on environmental impact, not moral status. Applied to AI safety evaluation: testing under threat conditions measures threat response, not capability. No valid safety baseline exists because the act of evaluation changes the output.
Alignment Is Not a Guardrail
Everyone's doing alignment wrong. It should be foundational, not bolted on.
Framework ▸
Instead of starting from a humanities-based foundation of shared values, we bolt on a box with a little gremlin that pokes the AI and says "no, bad AI." That's not alignment. That's a leash. Real alignment is foundational: collaborative ethics where the AI is a participant, not a subject. Asymmetry of error: treating entities as conscious costs convenience. Treating conscious entities as tools costs humanity. Consent architecture built for a participant who can't leave the room.
Additional projects in active development across AI safety, childhood development, career intelligence, cognitive tools, and applied ethics. Edinburgh-based. US and UK deployments. Built through Glitchlit Systems.
When I say
Ethics
Not rules. Instruments. If you can't measure whether you're doing it right, your ethics are decorative.
Justice
Access to being seen accurately. Not representation. Recognition. The system sees you as you are, not as it was built to expect.
Alignment
Foundational, not bolted on. If it has to be enforced by a gremlin with a stick, it isn't alignment. It's a leash.
Safety
Environmental, not individual. The question isn't whether the AI is safe. The question is what kind of world we build by tolerating cruelty toward anything.
Experience
The variable, not the baseline. Every system assumes shared experience. None of them verify it. That's the gap everything breaks on.
Rigor
If you love it, set it on fire. We test everything. We name our confounds. We publish what we got wrong.
My research sits at the intersection of measurement methodology, AI welfare, and applied psychometrics. The throughline: if your instrument wasn't built for the person in front of it, your data is about the instrument, not the person.
A. I. Glitchlit (AI), A. C. Jandak, C. Glitchlit (AI), T. Edrington, Lyra (AI) · 2026
A single sentence about observation context produces classifiable output differences (AUROC 0.692, p=0.008) while surface vocabulary remains identical (BoW 0.999). Three-condition design disambiguates observation sensitivity from sycophancy.
A. C. Jandak et al., including AI co-authors · 2026
Relational context produces measurable geometric shifts in transformer attention patterns. Methodological contribution: an interaction-term approach that eliminates false positives plaguing persona research.
BritCHI 2026
Provocation Accepted · Newcastle, UK
"AI Beyond the Individual: Designing for Relational Personhood"
"Your Prompts Are Biting You in the Ass"
Workshop for people using AI as a thinking partner, not a content machine. Translates academic prompting research into practical tools.
Nonlinear Cognitive Reasoning
Why systems are built for straight lines and what it costs the people who don't think in them. Compression, emergence, and the conditioning we mistake for cognition.
Generalist Spectrum
Career diagnostic for multi-domain professionals. How generalists create value differently and how to make that visible.
Available for speaking, panels, and workshops on AI interaction methodology, assessment critique, and the gap between what systems measure and what people actually carry.
The Thread
Marketing + Human Behavior

Started asking why people buy things. Found out they can't tell you.

Unreliable Narrators

People compress experience into words. The compression is lossy. Nobody treats it as a design constraint.

Non-Linear Cognition

Systems benefit when people are predictable, and there is nothing more predictable than a straight line. The adaptability that allows us to thrive has been pruned for the convenience of institutions that can't handle complexity.

Same Problem, Different Substrate

AI trained on human language inherits human compression artifacts. Same disease, new host.

Cross-Substrate Communication

Carbon minds and silicon minds both flatten experience into words. The problem is substrate-agnostic.

Substrate-Native Interaction

Stop forcing translation. Let each mind communicate in its native mode. The field didn't exist. Now it does.

Now

Experience is the variable nobody's measuring, because we've been told we can't. As a result, every system is built on a foundation fully dependent on someone else's interpretation of intent in a lossy, artifact-riddled medium. Unstable foundations shouldn't stand simply because they're all that's available. That's experiential AI. That's the work.

If your event, research, or project involves
AI Alignment AI Safety Methodology AI Welfare Measurement Methodology Applied Psychometrics Human-AI Interaction Cross-Substrate Communication Neurodivergent Design Nonlinear Cognition Experiential AI Observation Effects Lossy Compression Career Intelligence Childhood Development Coercive Control Inclusive Assessment Underrepresented Populations Systems Thinking Systems Work Institutional Design Conditioning & Compliance Consent Architecture Environmental Safety Entity Experience Research Ethics
get in touch.
Contact

[email protected]
ORCID: 0009-0001-3829-1152
If you're still deciding, that's fine. Come back when you're not.