Citations

How this portfolio was built, what AI touched, and what stayed fully mine

Every decision on this site has a origin. This page documents where AI was present in the work, where it wasn't, and what that distinction means for how I think about craft, authorship, and transparency in design practice. I'm not interested in hiding the assist or overstating the hand. I'm interested in modeling what honest attribution looks like when the tools are this embedded in the process.

Citation Framework

Four levels of attribution

Four levels of attribution

Every artifact ships with a citation rating assigned against these four levels. The distinction is decision authority, not effort.

Every artifact ships with a citation rating assigned against these four levels. The distinction is decision authority, not effort.

Level
Decision criteria
Example artifacts
Prompt notes
Human-led

The generative logic, structural frame, and primary constraints originated entirely with a person. The designer authored the problem framing, the approach, and the core output without AI initiating any structural decision.

Original strategy decks
Hand-sketched flows
Facilitated workshop outputs
Written briefs and principles
Handcrafted illustrations

No prompt involved. If AI was used only for spell-check or translation and made no structural decisions, this level still applies.

Co-created

The designer defined prompt constraints against explicit success criteria, evaluated competing outputs, and introduced at least one consequential structural edit that changed the artifact's decision surface before it shipped.

Iterated Framer components
Refined case study copy
Workshop decks with human-reframed structure
AI-drafted frameworks edited for org context

Log model, iteration count, and what structural edit was made. Prompt should specify constraints, not just a topic. If you only accepted output without editing logic, this is AI-drafted.

AI-drafted

The model produced the primary output and the human applied surface-level edits — copy polish, color, layout compression — without touching the underlying logic, structure, or decision surface of the artifact.

First-draft documentation
Boilerplate UI copy
Template-based reports
Auto-generated alt text
Routine email drafts

Note the prompt and model. If the human reorganized sections or changed the argument structure — making a judgment call that altered meaning — it's co-created, not drafted.

AI-generated

Output shipped near-unmodified. Human role was limited to prompt authorship and acceptance decision. No structural editing, logic review, or consequential human judgment was applied to the content itself.

Auto-generated summaries
Bulk image alt text
Boilerplate legal copy
Templated notifications
Automated test data

Full prompt, model, and version must be logged. Uncertain cases between drafted and generated should default to generated and be flagged for rubric review.

*Edge cases are logged with documented rationale and fed back into the rubric on the next version cycle. Assignment follows a decision tree, not a gut call.
Human-led
Decision criteria

The generative logic, structural frame, and primary constraints originated entirely with a person. The designer authored the problem framing, the approach, and the core output without AI initiating any structural decision.

Example artifacts
Original strategy decks
Hand-sketched flows
Facilitated workshop outputs
Written briefs and principles
Handcrafted illustrations
Prompt notes

No prompt involved. If AI was used only for spell-check or translation and made no structural decisions, this level still applies.

Co-created
Decision criteria

The designer defined prompt constraints against explicit success criteria, evaluated competing outputs, and introduced at least one consequential structural edit that changed the artifact's decision surface before it shipped.

Example artifacts
Iterated Framer components
Refined case study copy
Workshop decks with human-reframed structure
AI-drafted frameworks edited for org context
Prompt notes

Log model, iteration count, and what structural edit was made. Prompt should specify constraints, not just a topic. If you only accepted output without editing logic, this is AI-drafted.

AI-drafted
Decision criteria

The model produced the primary output and the human applied surface-level edits — copy polish, color, layout compression — without touching the underlying logic, structure, or decision surface of the artifact.

Example artifacts
First-draft documentation
Boilerplate UI copy
Template-based reports
Auto-generated alt text
Routine email drafts
Prompt notes

Note the prompt and model. If the human reorganized sections or changed the argument structure — making a judgment call that altered meaning — it's co-created, not drafted.

AI-generated
Decision criteria

Output shipped near-unmodified. Human role was limited to prompt authorship and acceptance decision. No structural editing, logic review, or consequential human judgment was applied to the content itself.

Example artifacts
Auto-generated summaries
Bulk image alt text
Boilerplate legal copy
Templated notifications
Automated test data
Prompt notes

Full prompt, model, and version must be logged. Uncertain cases between drafted and generated should default to generated and be flagged for rubric review.

*Edge cases are logged with documented rationale and fed back into the rubric on the next version cycle. Assignment follows a decision tree, not a gut call.
The most important design work I do isn't on a screen. It's in the leaders people ultimately become.
APPRENTICESHIPThe most important design work I do isn't on a screen. It's in the leaders people ultimately become.
I design teams the way nature designs systems — for resilience, not just pure efficiency.
TEAMINGI design teams the way nature designs systems — for resilience, not just pure efficiency.
I build practice from the inside, shaping direction, close enough to the work to earn trust.
LEADERSHIPI build practice from the inside, shaping direction, close enough to the work to earn trust.
Everything I touch is a system — from the platforms I shape to the people I grow, develop and lead.
SYSTEMSEverything I touch is a system — from the platforms I shape to the people I grow, develop and lead.
Like mycelium, my craft runs beneath the surface — quietly feeding everything I build and grow.
DESIGNLike mycelium, my craft runs beneath the surface — quietly feeding everything I build and grow.
The most important design work I do isn't on a screen. It's in the people I'm turning into leaders.
APPRENTICESHIPThe most important design work I do isn't on a screen. It's in the people I'm turning into leaders.
I design teams the way nature designs systems — for resilience, not just efficiency.
TEAMINGI design teams the way nature designs systems — for resilience, not just efficiency.
I architect practice from the inside — still a designer first, still deep in the heart of the work.
LEADERSHIPI architect practice from the inside — still a designer first, still deep in the heart of the work.
Everything I touch is a system — from the platforms I shape to the people I develop and lead.
SYSTEMSEverything I touch is a system — from the platforms I shape to the people I develop and lead.
Like mycelium, my craft runs beneath the surface — quietly feeding everything I build.
DESIGNLike mycelium, my craft runs beneath the surface — quietly feeding everything I build.

AMA Element

How Ask Me Anything was built

How Ask Me Anything was built

Seeded by hand, shaped through conversation, designed to answer like a person

The idea was simple enough: give people visiting the portfolio a way to ask real questions and get answers that actually sounded like me. The hard part was making it feel that way. It started not with code but with content, feeding the answer engine the kinds of responses I would give in person, calibrating the range of what it could handle versus what should stay in a real conversation and route to email instead.

The build happened in Claude Code. The component went through extensive back-and-forth iteration before it landed, covering dark mode, preloaded question chips, prompt tuning, and the logic for knowing when a question deserved a direct answer and when it deserved a follow-up. Like the carousel it lives as a custom TSX file loaded directly into Framer as a code element.

What took the most time wasn't the code. It was the answer architecture: deciding which questions to anticipate, how to phrase the seeds so responses felt considered rather than generated, and where to draw the line between what the component could hold and what needed a human on the other end of it. That boundary is still the most intentional part of the whole thing.

The idea was simple enough: give people visiting the portfolio a way to ask real questions and get answers that actually sounded like me. The hard part was making it feel that way. It started not with code but with content, feeding the answer engine the kinds of responses I would give in person, calibrating the range of what it could handle versus what should stay in a real conversation and route to email instead.

The build happened in Claude Code. The component went through extensive back-and-forth iteration before it landed, covering dark mode, preloaded question chips, prompt tuning, and the logic for knowing when a question deserved a direct answer and when it deserved a follow-up. Like the carousel it lives as a custom TSX file loaded directly into Framer as a code element.

What took the most time wasn't the code. It was the answer architecture: deciding which questions to anticipate, how to phrase the seeds so responses felt considered rather than generated, and where to draw the line between what the component could hold and what needed a human on the other end of it. That boundary is still the most intentional part of the whole thing.

Live example — AskTyler
AskTyler interactive component
Framer TSX · Custom code element · Local phrase-matcher · tylerhartrich.com
Co-created
Model: claude-sonnet-4Answer Keys: 167Human edits: answer architecture, dark mode, question chips, email routing logicPrompt constraints: respond like a person, route ambiguous questions to follow-up
*Built in Claude Code. The human contribution was the answer architecture: deciding which questions to anticipate, how to seed responses so they felt considered rather than generated, and where to draw the line between what the component holds and what needs a person on the other end.
Ask Me Anything

Ask me a real question.

I've answered the easy ones below. The interesting ones are yours to ask.

Questions my Design colleagues have asked me.

More than I should probably admit. There's something I genuinely like about the ritual of a week starting together, but only if the meeting has a spine. Rambling check-ins are a different story.

Systems thinking in situations where most people just see chaos. I can hold a lot of complexity and still find the thing that actually matters.

People, always, but people who care deeply about pixels. The craft is how I earn the trust to influence the bigger stuff.

I tell them I make things easier and clearer for people who are confused. They nod politely. I've stopped trying to do better than that.

Sideways, like most people obsessed with it. Came in through interaction design, stayed for the systems. Sixteen years later, I make practices instead of just screens.

Ask Me Anything

Ask me a real question.

I've answered the easy ones below. The interesting ones are yours to ask.

Questions my Design colleagues have asked me.

More than I should probably admit. There's something I genuinely like about the ritual of a week starting together, but only if the meeting has a spine. Rambling check-ins are a different story.

Systems thinking in situations where most people just see chaos. I can hold a lot of complexity and still find the thing that actually matters.

People, always, but people who care deeply about pixels. The craft is how I earn the trust to influence the bigger stuff.

I tell them I make things easier and clearer for people who are confused. They nod politely. I've stopped trying to do better than that.

Sideways, like most people obsessed with it. Came in through interaction design, stayed for the systems. Sixteen years later, I make practices instead of just screens.

Ask Me Anything

Ask me a real question.

I've answered the easy ones below. The interesting ones are yours to ask.

Questions my Design colleagues have asked me.

More than I should probably admit. There's something I genuinely like about the ritual of a week starting together, but only if the meeting has a spine. Rambling check-ins are a different story.

Systems thinking in situations where most people just see chaos. I can hold a lot of complexity and still find the thing that actually matters.

People, always, but people who care deeply about pixels. The craft is how I earn the trust to influence the bigger stuff.

I tell them I make things easier and clearer for people who are confused. They nod politely. I've stopped trying to do better than that.

Sideways, like most people obsessed with it. Came in through interaction design, stayed for the systems. Sixteen years later, I make practices instead of just screens.

I'm always open to meaningful conversations

I’m here for it. Whether you have a question, an idea, or just want to connect, reach out and let’s build a conversation.

Smiling young women and man in a conversation.
A woman and a man working side-by-side at an open workspace together.
A women at a whiteboard ideating with another person from a team.
Close-up of a dark green leaf showing its textured surface and central vein against a muted background.
Close-up of a dark green leaf showing its textured surface and central vein against a muted background.
Three designers working on a whiteboard wall.
A women making a clear statement in a learning environment.
Two learners and an instructor engaging in a conversation.
Two women smiling working side-by-side.

I'm always open to meaningful conversations

I’m here for it. Whether you have a question, an idea, or just want to connect, reach out and let’s build a conversation.

Smiling young women and man in a conversation.
A woman and a man working side-by-side at an open workspace together.
A women at a whiteboard ideating with another person from a team.
Close-up of a dark green leaf showing its textured surface and central vein against a muted background.
Close-up of a dark green leaf showing its textured surface and central vein against a muted background.
Three designers working on a whiteboard wall.
A women making a clear statement in a learning environment.
Two learners and an instructor engaging in a conversation.
Two women smiling working side-by-side.

I'm always open to meaningful conversations

I’m here for it. Whether you have a question, an idea, or just want to connect, reach out and let’s build a conversation.

Smiling young women and man in a conversation.
A woman and a man working side-by-side at an open workspace together.
A women at a whiteboard ideating with another person from a team.
Close-up of a dark green leaf showing its textured surface and central vein against a muted background.
Close-up of a dark green leaf showing its textured surface and central vein against a muted background.
Three designers working on a whiteboard wall.
A women making a clear statement in a learning environment.
Two learners and an instructor engaging in a conversation.
Two women smiling working side-by-side.