Influence
What shaped the maker or the work: a song heard at fourteen, a city, a rival, a book.
Test: Would the work be recognisably different without it? Influence is usually acknowledged, rarely owed anything.
Exhibit 01Opening argument
The cook and the wheat
Evidence documented fact Established framework existing standard or rule Interpretation this page’s reading Hypothesis testable guess Illustration worked example Fiction invented scenario
FictionPicture a cook in a small Delhi kitchen. She rolls a chapati, slaps it onto a hot tawa, flips it once and lets it puff over the flame. She did not grow the wheat. She did not mill the flour, lay the gas line, forge the iron pan or invent the method; her mother taught her that, and someone else taught her mother. And yet “she made lunch” is not a lie.
It is a compression. Every credit line is one. “Directed by,” “Written by,” “Made in India,” “by me” all fold a crowd into a name. Compression is necessary. Nobody wants a forty-page acknowledgement on a lunch plate. The trouble starts when we forget what was folded in, and when the short sentence quietly becomes something it never meant: a property deed, a moral verdict, a pay slip.
InterpretationAuthorship is a claim about choices and accountability. Ownership is a claim about rights. Contribution is a claim about causes. They overlap, and none of them settles the others.
Most arguments about credit get stuck because the people in them are answering different questions. One says, “I did most of the work,” which is about contribution. Another says, “It’s in the contract,” which is about agreed credit. A third says, “You’ll be blamed if it breaks,” which is about responsibility. All three can be true at once. Nobody is lying; they are just using one small word, made, for several different jobs.
New tools add a fresh layer to the problem, but they did not create it. Software that drafts text or images forces the old question into the open, because it can supply material that looks like expression rather than mere capacity. Saying “it’s just a tool” is a reasonable start. It tells you who should answer for the result. It does not, by itself, tell you who owns the output, whether anyone whose work trained the tool was treated fairly, or what you should say in public about how the thing was made.
This page gives you a way to keep those questions apart. You can try it on a fictional creation in the credit map. Then you can write an honest disclosure for your own work, one that stays on your device.
Exhibit 02Framework
Six questions hiding inside “Who made this?”
InterpretationThe six labels below are this page’s working vocabulary, not a legal taxonomy. Each one comes with a test question you can actually ask.
Contribution
What actually went in: labour, materials, judgement, code, words, money, time.
Test: Point to the part of the work that would be missing. Contributions are causes, not titles.
Legal rights
What law grants or protects: copyright, moral rights, property in materials, patents, trade marks.
Test: Which country’s law applies, and what do the facts show? Answers vary by jurisdiction and change over time.
Agreed credit
What people agreed: bylines, ghostwriting, employment terms, licences, contributor agreements.
Test: Is it written down, did everyone understand it, and was it fair when signed?
Responsibility
Who answers if it harms, misleads, infringes or simply fails.
Test: Who chose to publish it, and who controlled the part that went wrong?
Public claim
What you say, in public, about who made it and how.
Test: Would your audience feel misled if they saw the full process?
None of this is entirely new. Research publishing already separates several of these questions. Established frameworkThe CRediT taxonomy (ANSI/NISO Z39.104-2022) lists 14 contributor roles, among them conceptualization, software, validation and writing – review & editing, so contribution can be described separately from the author list. The ICMJE recommendations for medical journals go further. To be listed as an author you must also approve the final version and accept accountability for the work. People who helped but do not meet all the criteria are acknowledged instead.
InterpretationThat second idea is worth borrowing outside medicine. In that tradition a byline is a promise as well as an honour: I will answer for this. If you would not answer for something, your name probably should not sit on it as its author.
Exhibit 03InteractiveFictional scenarios
The credit map
Pick a fictional creation and decide which of the eight inputs it had and what kind each one was. Then look at the result through four lenses. The threads change weight as the answers change. That is the point: the same creation has different credit maps depending on the question you ask.
IllustrationThe map never gives percentages or ownership verdicts. Its labels (clearly, partly, depends, rarely, no) are rough editorial readings of typical situations, written to help you ask better questions. They are not legal conclusions.
The interactive map needs JavaScript. The text version below covers the same eight inputs and four lenses for one typical case of each input.
Every change updates the notes, the thread diagram and the reading below it. Your choices are not saved or sent anywhere.
3 · Assemble the inputs
- Clearly
- Partly
- Depends
- Rarely
- No thread: no, or not included
Reading the map
Text version: typical readings of the eight inputs
| Input (typical case) | Contributed? | May claim ownership? | Bears responsibility? | Benefits? |
|---|---|---|---|---|
| Maker Makes the key choices directly | Clearly. Makes the choices that give this creation its particular shape. That is the strongest ordinary basis for being called its author. | Depends. Often the default claimant, but employment terms, commissions and local law can move rights elsewhere. Making something is not the same as owning it. | Clearly. Answers first for what the work does and says, because they chose it. | Partly. Benefits through satisfaction, skill, reputation or payment, but rarely captures all the value the work creates. |
| Collaborators Informal help, nothing agreed | Clearly. Informal help is still a contribution. Friends’ labour is the easiest to forget because nobody wrote it down. | Depends. When nothing was agreed, people’s assumptions fill the gap. Some legal systems have joint-authorship rules, and they turn on the facts. Talk early. | Partly. Shares responsibility for what they did, usually less than the person who published. | Rarely. Often gets the least benefit for the effort, unless someone makes a point of thanking, crediting or paying them. |
| Source materials Bought or commodity materials | Clearly. Raw inputs are causally essential. No wheat, no chapati. | No. Buying materials usually makes them yours. Their producers do not usually claim the finished creation. | Partly. Producers answer for the quality and safety of what they supplied, not for what you made from it. | Partly. Paid at the point of sale. Whether that price was fair is a separate question this map cannot answer. |
| Teachers A tradition with no single author | Partly. Recipes, techniques and forms carried by many people. A real influence with a diffuse origin. | Rarely. Traditions are usually nobody’s property in the copyright sense, though some communities assert forms of custodianship worth respecting. | No. No individual answers for how you use a tradition. You do. | Rarely. Communities that carry a tradition rarely benefit directly when it is sold. Credit and context are the minimum. |
| Tools Conventional tool (pan, scale, editor) | Partly. Tools extend what you can do and limit your options. A contribution, but the tool makes no choices. | No. Toolmakers do not usually claim what you make with a tool. Check software terms anyway. | Partly. Toolmakers answer for defects. You answer for how you used it. | Partly. Benefit at purchase or through a subscription. |
| Infrastructure Shared or public infrastructure | Partly. Roads, water, libraries, standards, the internet: silent contributions that make making possible. | No. No claim over the creation. | Rarely. Answers for the infrastructure working, not for what passes through it. | Rarely. Benefits indirectly, through taxes, fees or a healthier commons. |
| Reviewers Light feedback | Partly. Feedback that changes your mind is a contribution even when it adds no words. | No. Reviewing does not generally create a claim. | Rarely. Reviewers do not usually answer for what you published, unless they formally signed off. | Rarely. Benefits by being thanked, if you remember to. |
| Audience Customers or readers | Partly. Attention, use and money complete the circuit. A meal nobody eats is a different thing. | No. Buying a copy is not buying authorship. | No. Not responsible for the work. They are who the work is responsible to. | Clearly. The work exists partly to benefit them. If it does not, check the other lenses. |
How to use what you see
Three patterns come up again and again. First, contribution is wide and ownership is narrow. Nearly everything contributes; very little can claim. Second, “depends” clusters around agreements. Collaborators, licensed sources, platforms and AI tools are where unwritten expectations do the most damage, so they are where a short conversation, or a written agreement, pays for itself. Third, benefit often runs away from contribution. Try turning a source into “used without permission” and switching to the benefits lens: the thread disappears. A missing thread is often where fairness questions start.
Exhibit 04Five short essays
Five short essays on the many hands
Each essay takes one ordinary object and asks one question about it. They can be read in any order.
Essay 01 · Materials and making
The chapati does not belong to the wheat
EvidenceIn 1958 Leonard Read published I, Pencil, a short essay told in the voice of a pencil. Its claim is that no single person knows how to make one. Cedar, graphite, lacquer, brass and rubber pass through thousands of hands that never meet. Read used this to argue that markets coordinate work without central direction.
InterpretationYou can take the observation without taking the whole political conclusion. That a lunch depends on farmers, millers, truckers and gas networks is simply a fact. What you then conclude about prices, states or markets is a separate argument, and people disagree about it in good faith. This page needs only the fact.
The fact raises a puzzle. If the farmer’s contribution is essential (no wheat, no chapati), why does nobody say the farmer made lunch? The usual answer is about choices. The cook decided how thin to roll, how hot to run the pan, when to flip and what to serve it with. Her choices give this particular chapati its shape. The farmer’s choices shaped the wheat. Authorship tends to follow the last meaningful choices that gave a thing its character.
That answer is good but dangerous. A choice-centred idea of authorship can quietly erase people whose work involved fewer visible choices and a great deal of effort: the miller, the cleaner, the assistant who prepared every sample. They may not be the authors, and that is fine. They still contributed, and they may still be owed pay, safety, thanks or a name. Not being the author does not mean you contributed nothing. The cook can say “I made lunch” and also remember the wheat.
Essay 02 · Tools and responsibility
The doctor’s knife
IllustrationA surgeon makes an incision. The scalpel does not share responsibility for where the cut went. The manufacturer answers for whether the blade was sound. The hospital answers for whether it was sterile. The surgeon answers for where and why it cut. InterpretationAs a rough rule, responsibility follows control and foreseeability: you answer for the parts you could steer and could reasonably have seen going wrong. Legal doctrines put this in different ways in different places. The everyday intuition holds up well.
This is why “AI is just a tool” is half right. Established frameworkThe ICMJE recommendations say chatbots cannot be authors because they cannot take responsibility for the work. Humans stay accountable, and they must disclose how such tools were used. If you publish it, you answer for it, whatever helped you make it.
The analogy has limits, though, and they matter. A knife extends a surgeon’s hand; it does not offer three alternative incisions. A generative system can supply expressive material: sentences, images, melodies. Calling it a tool settles who is accountable for the use. It leaves at least three other questions open.
EvidenceOne is whether the output can be owned at all. The US Copyright Office’s 2025 report on copyrightability concluded that, with current technology, prompts alone generally do not give enough control to make the user the author of the output. Human expressive inputs, and creative selection, arrangement or modification, can be protected, but only the human part, and only case by case. The second question is whether the people whose work trained the tool were treated fairly. The third is what the vendor owes its users. These are separate questions, and the knife analogy answers none of them.
Essay 03 · Organisations
The company is an instrument
InterpretationA company, studio or shop is a way of coordinating people and spreading the value they create. It pools money, assigns roles, signs contracts, carries some of the risk and decides who gets paid what. That is useful work. It is also why credit so often flows upward to the name over the door, while the people inside it become “the team.”
Sometimes that is fair. People agreed to it, were paid well and preferred to stay unnamed. In many jurisdictions, rights in work made in the course of employment go to the employer by default. That is a deliberate legal design, not a theft, and the details vary enough that you should check your own contract and local law. Sometimes it is an erasure, where the arrangement was never really discussed and the brand takes the credit along with the money.
IllustrationThink of a hypothetical family shop. The sign says the family’s name. But the supplier who extends goods on trust, the helper who remembers what every regular buys, and the neighbour who sends customers over all help keep it running. None of them is on the sign, and nobody thinks they should be. The question is whether the arrangement treats them well, not whether they are named.
The deeper point is about identity. When an organisation becomes someone’s whole purpose, people stop asking whether it is distributing value fairly, because questioning the instrument feels like questioning the self. Keeping the company in its place as a means makes that question easier to ask. The Dignity essay makes a related argument: politeness at work cannot stand in for fair conditions. Credit works the same way. A warm thank-you note does not replace a fair contract.
Essay 04 · Kinds of value
Value beyond the price tag
InterpretationMoney is one ledger, and a well-kept one. It is precise, portable and comparable. Most made things also write to other ledgers that are kept far less carefully:
| Ledger | What gets created | Example |
|---|---|---|
| Material | Something usable exists | Four people are fed |
| Educational | Someone can now do something new | A nephew learns to knead |
| Relational | Trust or closeness grows | A shared meal settles an argument |
| Aesthetic | Something is pleasing or meaningful to experience | A well-set page, a balanced scent |
| Institutional | Reputation, standards or reliable routines | A shop known for honest weights |
| Monetary | Someone is paid | The invoice, the salary, the sale |
The ledgers do not always move together. A teacher whose method shapes a student’s work for decades is not paid each time it is used. A volunteer maintainer of a small open-source library may see their code run inside thousands of businesses. The monetary ledger records almost none of this. Because money is the easiest ledger to read, it tends to become the only one we check.
The opposite mistake is just as real. “You’ll get exposure” has been used to pay people in non-monetary credit when money was owed. Remembering the other ledgers is meant to help us see more of what someone contributed. It is never a reason to pay them less. A fair arrangement usually needs entries on more than one ledger, and money is often one of them.
Essay 05 · The other end of the thread
Audience-first work
InterpretationThe audience is an odd kind of contributor. It brings attention, use, feedback and meaning. A recipe is not finished until someone eats the result; a website is not finished until someone finds the opening hours on it. That does not usually make the audience co-authors, because they did not make the expressive choices. In community projects with explicit contribution terms, they sometimes are.
“Audience-first” is a useful discipline. It means designing for the person on the other end: clear writing, honest limits, things that work. The maker’s reputation comes second. It also changes how you think about credit. A disclosure about how something was made is not a confession. It is information the audience might need in order to trust the work, or to judge it.
The sibling sites offer small examples. The Mind Mixtape says its remix examples are handcrafted rather than AI-generated, and that its studio runs locally without collecting data. Sahil’s web studio says AI-assisted tools may help with exploring, prototyping and coding, but that the finished work must be reviewed and working, and it calls its project images conceptual art rather than screenshots. These are the sites’ own statements, not independent audits. They show provenance written for a reader.
There is a failure mode too. Audience-first can slide into audience-pleasing, where you say whatever the crowd rewards. One test helps: would this reader thank you a year from now, once they know everything you knew?
Exhibit 05Worked cases
Three worked cases
Each case runs through the six questions. Look for the places where the answers come apart. Those are the places where disputes start.
FictionThe ghostwritten memoir
A retired cricketer’s memoir is written by a professional ghostwriter after forty hours of interviews. The cover carries only the cricketer’s name.
- Influence
- The cricketer’s life, obviously; also the ghostwriter’s reading of sports memoirs.
- Contribution
- The cricketer contributes the stories, memories and approvals. The ghostwriter contributes nearly every sentence and the structure.
- Legal rights
- Set mostly by the contract and governing law. Copyright is often assigned to the named author or publisher. Check the actual terms.
- Agreed credit
- The cover credit was agreed and paid for. The ghostwriter may have chosen to stay invisible.
- Responsibility
- The cricketer answers for the truth of the claims, having approved them. The ghostwriter answers for faithful rendering.
- Public claim
- “By [Cricketer]” is a convention many readers of the genre understand. It would become misleading if the cricketer claimed in interviews to have written every word.
InterpretationThe answers diverge, and that can be fine, as long as the arrangement was agreed and fairly paid, and the public claim does not mislead in a way the audience would care about.
IllustrationThe open-source dependency
A small business website relies on a free code library written by a volunteer.
- Contribution
- Large. The site might not work without it.
- Legal rights
- The volunteer keeps copyright in the library. The licence grants others permission to use it. Many permissive licences require keeping the copyright notice; read the specific one.
- Agreed credit
- Whatever the licence requires, which is often a notice in the source rather than a public byline.
- Responsibility
- Open-source licences commonly disclaim warranties (“as is”). The business answers to its own customers for how the site behaves.
- Benefit
- Flows mostly to users. The maintainer may get reputation, learning or nothing.
InterpretationThis is credit with the most structure: rights, permissions and obligations are all written down. Even so, the benefits ledger is lopsided. A written licence settles the rights and credit questions and leaves the fairness question open.
FictionThe AI-assisted poster
A designer writes prompts, generates forty images, picks one, repaints a corner by hand, then sets the type and layout for an event poster.
- Contribution
- The designer: concept, prompts, selection, repainting, typography, layout. The image system supplied much of the visual material. Its builders and training data sit behind that.
- Legal rights
- EvidenceIn the US, the Copyright Office’s 2025 guidance suggests the prompts alone would not make the designer the author of the generated image. The hand repainting, selection and arrangement, and the typographic layout may be protectable as human contributions. The D.C. Circuit’s 2025 decision in Thaler v. Perlmutter held that a work listing an AI system as its sole author could not be registered, and the Supreme Court declined to hear the case in March 2026. The UK has a specific provision for “computer-generated” works (CDPA 1988, s.9(3)), and how it applies to modern generative AI is debated. Other countries differ again.
- Responsibility
- The designer and the event organiser answer for what the poster says and shows, including if it closely resembles someone else’s work.
- Public claim
- “Design: [name], with an AI image generator for the base illustration” is honest and costs little.
InterpretationWhether the designer can own the result and whether the designer is responsible for it are separate questions. They are not the same, and they can have different answers.
Exhibit 06Objection
The serious objection: this map is too polite
The objection, stated as strongly as I can
A neat vocabulary of six questions treats credit as a matter of describing things clearly. But credit is also about bargaining power. Suppose a tool trained on the work of many illustrators lets a client skip hiring any of them. The “it’s a tool” framing then hides a transfer of value from many contributors to the tool’s owner and its users, and naming the categories carefully does nothing to reverse it.
Worse, the new tools may make attribution structurally harder. A credit map has eight boxes. A model may draw on millions of works, and no output can list them. Asking makers to “disclose honestly” might put the whole burden on individuals while the system that blurred the credit stays out of view.
What this page can say back
InterpretationMost of the objection is right, and this page does not solve it. A vocabulary cannot fix a power imbalance. Contracts, collective bargaining, licensing markets, regulation and courts might, and every one of those is contested. The US Copyright Office has published a separate pre-publication report on generative AI training, which shows how unsettled that ground still is.
What the six-question split does is stop the debate from collapsing. “Is this output copyrightable?”, “Were the people whose work trained it treated fairly?” and “Who benefits?” are separate questions with separate remedies. Once they are kept apart, a satisfying answer to one cannot be used to dismiss the others.
HypothesisSome contributions will never be creditable one by one, at the level of a single output. If that is right, fairness for diffuse contributors would have to come from structures such as collective licensing, funds or standards, not from longer credit lines. That is a guess worth testing, not a conclusion.
Exhibit 07Misconceptions, limits, counterexamples
Misconceptions and limits
- “I paid for it, so I made it.”Payment can buy materials, labour or rights. It does not buy the history of who made the choices. You may own the result without having authored it.
- “It’s my tool, so I own whatever it produces.”Ownership depends on law, contract and the tool’s terms of service. For AI output, at least in the US, it also depends on how much human expression went in.
- “The person who worked hardest is the author.”Effort and authorship are different things. Effort deserves pay, acknowledgement and decent conditions. Authorship tracks choices and accountability.
- “Naming more contributors makes me look smaller.”Usually it makes you more believable. One exception is real: in fields where author positions are scarce and ranked, credit can genuinely be zero-sum, and that is exactly why written criteria matter there.
- “Disclosing AI use is admitting I cheated.”A disclosure is information. Whether the use was appropriate depends on the context, the rules and what was promised. Hiding it is what turns an ordinary tool into a breach of trust.
- “Provenance proves the content is true.”Standards like C2PA Content Credentials can record where media came from and how it was edited. A record of origin does not check whether the content is accurate. You still have to judge that yourself.
Where this framework fits badly
Collective traditions. For a folk melody or a family recipe, individual credit is the wrong grain, and the better question is how to respect a community’s custodianship. Chosen anonymity. Some contributors want no name, for privacy or safety, and a credit regime should not override that. Danger. Naming a source can expose them to harm. Cultural fit. The six questions lean on ideas from copyright and research ethics, and many cultures have their own systems of attribution and obligation that these categories do not capture. Speed. Not everything needs a credit audit. A birthday card does not need a provenance statement.
Exhibit 08ToolRuns locally
Write a provenance disclosure
Fill in what is true about a piece of your work, and the generator assembles a plain, honest disclosure you can paste under it. It only uses what you type. It will not invent checks you did not do or describe AI use you did not report. Blank sections are left out.
The generator needs JavaScript. Without it, you can use this template: Made by [names/roles]. Human choices: [what people decided]. AI assistance: [none / which tools, for what]. Sources: [materials, licences]. Edits: [what changed after drafts]. Checks: [what was verified]. Uncertainties: [what might still be wrong].
Your disclosure
Updates as you type. Read it once before you use it; you know the facts, the template does not.
Worth a second look (not added to the text)
Privacy: everything you type stays in this browser tab. Nothing is sent, stored, cookied or logged. Closing or reloading the page erases it, and “Clear everything” erases it now.
Exhibit 09Unresolved
An open question to take with you
Credit, at its best, is an honest story about causes, told briefly enough to be useful. The cook can name her mother’s method. The designer can name the library. A researcher can list fourteen kinds of contribution.
But what would it take for that story to include the people we can no longer name, whose work is spread across a tradition, a dataset or a supply chain, without either pretending they are not there or burying every work in footnotes?
This page does not have the answer. If you have a good partial answer, or a counterexample that breaks the framework, write to Sahil. Related essays live in his Medium archive.
ArchiveSources, checked 24 Oct 2026
Sources, and what each one does and does not show
- US Copyright Office, Copyright and Artificial Intelligence, Part 2: Copyrightability (January 2025). The Office’s view that prompts alone generally do not give authorship of output with current technology, that human expressive inputs and creative selection, arrangement or modification can be protected, and that this is decided case by case. What courts in other countries will decide, or what US courts will decide in every future case.
- Thaler v. Perlmutter, No. 23-5233 (D.C. Cir., 18 March 2025); certiorari denied 2 March 2026 (SCOTUSblog docket summary). Under the US Copyright Act, a work must be authored in the first instance by a human, so a work listing an AI system as sole author cannot be registered. How much human involvement is enough when people use AI. The court expressly left that open.
- UK Copyright, Designs and Patents Act 1988, section 9(3). For computer-generated works, the UK treats as author the person who made the arrangements necessary for the work’s creation. How that provision applies to modern generative AI, which is debated.
- NISO, CRediT: Contributor Role Taxonomy, ANSI/NISO Z39.104-2022. An established standard of 14 roles for describing contributions to research outputs, separately from the author list. Who should be an author, or how to weigh one role against another.
- ICMJE, Defining the Role of Authors and Contributors. Medical-journal authorship requires accountability and final approval as well as contribution; AI tools cannot be authors and their use must be disclosed. Rules for fields outside medical publishing, which have their own norms.
- Leonard E. Read, I, Pencil: My Family Tree as told to Leonard E. Read (Foundation for Economic Education, 1958; Library of Economics and Liberty edition). A classic, widely cited illustration of how many dispersed people contribute to a simple object. Its policy conclusions about markets, which are argued rather than demonstrated, and which this page does not adopt or reject.
- Coalition for Content Provenance and Authenticity (C2PA), Content Credentials. An open technical standard for recording the origin and edit history of digital media. That content carrying credentials is accurate or true.
- Creative Commons, Best practices for attribution. The practical TASL pattern (Title, Author, Source, Licence) for attributing reused CC-licensed works. Requirements for works under other licences or none.
The sibling sites (Mind Mixtape, Web Studio, Dignity) were visited on the same date. They are quoted for what they say about themselves, not as independent evidence. Full notes are in RESEARCH_AND_CREDITS.md.