THE ARCHIVE
OF ATTEMPTS

EST. 2026 / THE MUSEUM OF UNFINISHED THINKING

NOT EVERYTHING
WORKED. GOOD.

An attempt is a record of a question meeting reality. The useful part isn't the heroic failure story. It's what we can honestly infer—and what we can't.

Start with the method Open 15 case files

All 15 case files are teaching scenarios or unverified worksheets, not documented failures from a person's life.

01 / YOUR WAY INTO THE ROOM

Make the attempt legible.

A failed outcome contains less information than we like to think. Something went differently from what we wanted; that alone rarely explains why.

Imagine a redesigned navigation menu that nobody uses. Perhaps the labels are obscure. Perhaps visitors don't need that destination. Perhaps no relevant visitors arrived. These explanations imply different next steps. Without a record of what changed and who encountered it, “the redesign failed” is mostly a story.

This archive asks you to keep four columns separate: intention, procedure, observation, and interpretation. It is a workshop for reasoning, not a trophy shelf for resilience.

02 / WORKING METHOD

Four columns.
One intellectual brake.

A good archive remembers not just the version that looks better in hindsight, but the alternative explanations that remain alive.

I

Intention

What specific effect did we expect, for whom, and under what conditions? A wish like “make it compelling” is not yet a testable hypothesis.

II

Procedure

What changed, what stayed constant, and what resources were available? Version histories protect us from retrofitting a story after seeing the result.

III

Observation

What actually happened—and how do we know? Separate logs, visitor comments, personal impressions and imagined walkthroughs.

IV

Inference limits

Which conclusions do not follow? “No response” cannot tell us whether the idea was unwanted, unseen, confusing or simply early.

An absence of success is not evidence for your favourite explanation.

This is an editorial reasoning framework, not a validated four-factor scientific model. It draws on iterative design practice and research transparency rather than claiming novel empirical proof. See source notes ↓

03 / THE UNCOMFORTABLE ARGUMENT
READING TIME
≈ 5 MINUTES
KEEP THE
COUNTERCLAIM

The romance of failure can be just another way of avoiding evidence.

“Fail fast” is attractive because it sounds brave. But speed alone is not an epistemic virtue. An experiment can fail quickly and still teach almost nothing if its question was vague, its audience absent, or its results interpreted only after the fact.

There are also costs that a cheerful failure narrative tends to omit: time, money, trust, energy, safety and accessibility. A public-facing experiment involving other people needs stronger protections than a private sketch. The freedom to try something new does not make avoidable harm a noble learning experience.

But excessive measurement can become a trap.

At the opposite extreme, a creator can spend more time specifying metrics than creating anything worth encountering. Some questions are exploratory: “What would this feel like?” “Could two ideas be connected?” “What do people notice?” Discovery is a legitimate goal even when a confirmatory trial is premature. The mistake is to tell a discovery story as if it proved a general rule.

What deserves preserving?

Keep versions that show a change in reasoning, not every cosmetic tweak. Keep rejected options when their rejection helps explain a trade-off. Preserve why a decision was made, including the constraints that mattered then. And give observations their correct weight: a single piece of feedback can spark an excellent revision without establishing a population-level finding.

Scientific preregistration formalizes a related distinction between planned and exploratory analysis; creative projects need not mimic a laboratory, but they benefit from the same honesty about what was predicted before seeing an outcome. Center for Open Science ↗

Competing lenses

Ship-and-learn: small, reversible prototypes expose assumptions that planning cannot fully anticipate.

Measure-first: predefined questions reduce storytelling and selective reporting when decisions are consequential.

Craft-first: some value—beauty, expressive voice, curiosity—does not fit a single metric. It still deserves critique, not invented performance evidence.

04 / THE DRAWER IS OPEN

15 case files.
Zero invented biographies.

Each file uses the same nine questions. The “observed result” field explicitly distinguishes constructed examples from missing real observations.

15 of 15 case files · static archive; filters require JavaScript

A01 / 15Teaching scenario

Concept / illustrative reconstruction

The rejected all-in-one dashboard

A single dashboard makes an interdisciplinary body of work easier to navigate.

Inspect the attempt
Hypothesis
A single dashboard makes an interdisciplinary body of work easier to navigate.
Resources & constraints
One afternoon of wireframing; a paper sitemap; five fictional visitor tasks.
Version history
v0 everything on one screen → v1 grouped by purpose → v2 a small number of entry doors.
What was tested / proposed
In a proposed five-task walkthrough, ask readers to locate one essay, one tool and one explanation without prior context.
What failed / could fail
The illustrative v0 crowds category names and forces visitors to understand the author’s mental taxonomy first.
Observed result
Simulated observation, not field data: the example visitor hesitates between “lab”, “archive” and “worlds”. No completion times were collected.
Revised approach
Start with visitor verbs (“read”, “try”, “learn”) and expose the complete directory as a second route.
Next question
Does a task-first entrance help return visitors as much as newcomers?
Lessons that do NOT follow
It does not prove that dashboards are bad, that fewer links are always better, or that a particular homepage failed in production.

Evidence boundary: no documented personal outcome is represented by this card.

A02 / 15Teaching scenario

Concept / illustrative reconstruction

The community feature without a community

Adding public comments will create richer discussion around an educational website.

Inspect the attempt
Hypothesis
Adding public comments will create richer discussion around an educational website.
Resources & constraints
A moderation policy draft, a threat model and sketches for three discussion states.
Version history
v0 open comments → v1 moderated queue → v2 private email invitation pending evidence of demand.
What was tested / proposed
Before building, model spam, personal-data exposure, moderation time and the question a visitor would actually answer.
What failed / could fail
An open box can be easy to implement but costly to steward; silence may reflect audience size rather than low interest.
Observed result
Reasoned scenario: support costs and safeguarding questions appear before any tested evidence of demand.
Revised approach
Invite specific feedback by email and publish aggregated, permission-safe lessons only with consent.
Next question
What genuine conversations are already occurring that a new feature would improve?
Lessons that do NOT follow
It does not show that communities are pointless or that low comment counts imply failure.

Evidence boundary: no documented personal outcome is represented by this card.

A03 / 15Teaching scenario

Web / illustrative reconstruction

A search box that finds only titles

Visitors can retrieve the right article if a search input matches page titles.

Inspect the attempt
Hypothesis
Visitors can retrieve the right article if a search input matches page titles.
Resources & constraints
A tiny static index and a set of representative queries; no analytics.
Version history
v0 title-only → v1 headings and summaries → v2 synonyms and an explicit no-results view.
What was tested / proposed
Try “why do plans stall?” when the relevant page is titled “Identity–Action Gap”.
What failed / could fail
Literal matching misses the visitor’s vocabulary; empty results lack recovery steps.
Observed result
Demonstrative result: the test query returns no matching title in v0 by construction, not as a claim about a live service.
Revised approach
Index descriptive summaries, offer topic filters and expose a browse-all path.
Next question
When does synonym expansion introduce irrelevant results?
Lessons that do NOT follow
It does not establish a measurable search failure rate or justify adding complex AI search.

Evidence boundary: no documented personal outcome is represented by this card.

A04 / 15Teaching scenario

Web / illustrative reconstruction

The beautiful graph with no text alternative

An animated concept map can make related disciplines easier to understand.

Inspect the attempt
Hypothesis
An animated concept map can make related disciplines easier to understand.
Resources & constraints
An SVG sketch, 12 sample nodes and a keyboard-only task script.
Version history
v0 pointer-only nodes → v1 native buttons → v2 synchronized list and explanatory link table.
What was tested / proposed
Ask someone using only Tab, Enter and screen-reader headings to find the link between design and research.
What failed / could fail
The v0 interaction exists only at pointer coordinates, leaving no operable equivalent.
Observed result
Code-design inference: inaccessible pointer-only handlers block keyboard operation; no real user testing is asserted.
Revised approach
Build a structured text index first, then layer the graph as an optional view of the same data.
Next question
Can the list convey relationships without turning into another overwhelming catalogue?
Lessons that do NOT follow
It does not mean data visualizations are inherently inaccessible or that a list alone is sufficient.

Evidence boundary: no documented personal outcome is represented by this card.

A05 / 15Teaching scenario

Accessibility / illustrative reconstruction

The disappearing keyboard focus

Removing focus outlines will make a minimalist interface look more polished.

Inspect the attempt
Hypothesis
Removing focus outlines will make a minimalist interface look more polished.
Resources & constraints
Two visual states of one control; the W3C focus-visible criterion.
Version history
v0 outline removed → v1 3px offset outline → v2 test across dark/light surfaces.
What was tested / proposed
Tab through controls and identify the focused one without activating it; inspect focus on different backgrounds.
What failed / could fail
A focus cue that depends on near-invisible border color is easy to lose on a dense page.
Observed result
Rule-based assessment: W3C requires visible keyboard focus; this demonstration is not a formal conformance audit.
Revised approach
Use high-contrast :focus-visible styling and keep focus order aligned with reading order.
Next question
Can the focus treatment survive nested cards, dialogs and browser zoom?
Lessons that do NOT follow
It does not demonstrate that every user needs the same visual focus style or that one repair makes a page WCAG conformant.

Evidence boundary: no documented personal outcome is represented by this card.

A06 / 15Teaching scenario

Accessibility / illustrative reconstruction

The form that whispers “invalid”

A red border alone will tell users how to correct a form error.

Inspect the attempt
Hypothesis
A red border alone will tell users how to correct a form error.
Resources & constraints
A draft form, field labels and three error message alternatives.
Version history
v0 border-only → v1 short error sentence → v2 summary plus inline correction hint.
What was tested / proposed
Submit missing required fields and attempt recovery using a keyboard and a screen reader.
What failed / could fail
Color alone omits the specific remedy and may not communicate the error to everyone.
Observed result
Illustrative inspection: a red border has no intrinsic error instructions; no reported user-study finding.
Revised approach
Associate errors with fields, preserve entered values and explain a concrete repair.
Next question
How can validation avoid interrupting a user before they have finished typing?
Lessons that do NOT follow
It does not establish that more verbose errors are always better or replace testing with assistive technology.

Evidence boundary: no documented personal outcome is represented by this card.

A07 / 15Teaching scenario

Design / illustrative reconstruction

Four palettes, one unreadable sidebar

A strong accent color will unify dark and light design modes.

Inspect the attempt
Hypothesis
A strong accent color will unify dark and light design modes.
Resources & constraints
A token spreadsheet, a contrast calculator and a sample navigation sidebar.
Version history
v0 single accent token → v1 separate text/surface tokens → v2 semantic success/warning/focus tokens.
What was tested / proposed
Check normal text and UI boundaries across each theme instead of assuming an attractive pair passes.
What failed / could fail
One accent for every role confuses hierarchy and may produce low contrast in some combinations.
Observed result
Analytical scenario: the risk follows from token reuse; no exact historical regression is alleged.
Revised approach
Choose color by role, document accessible pairings and test components in each state.
Next question
Could fewer themes make maintenance more reliable than the appeal of endless skins?
Lessons that do NOT follow
It does not prove that expressive interfaces cannot be accessible or that a palette alone determines usability.

Evidence boundary: no documented personal outcome is represented by this card.

A08 / 15Teaching scenario

Design / illustrative reconstruction

The motion-heavy entrance

Continuous visual motion will make a creative site feel more memorable.

Inspect the attempt
Hypothesis
Continuous visual motion will make a creative site feel more memorable.
Resources & constraints
A storyboard, a low-motion variant and a reading-task checklist.
Version history
v0 looping transitions → v1 motion limited to feedback → v2 user-controlled animation with reduced-motion default.
What was tested / proposed
Compare whether the key message remains readable with animation off and at narrow widths.
What failed / could fail
Motion competes with content and can be uncomfortable or distracting for some visitors.
Observed result
Design interpretation: attention cost is plausible, not a measured retention or conversion effect.
Revised approach
Make transitions purposeful and nonessential; honour prefers-reduced-motion.
Next question
Which movement actually helps explain a state change?
Lessons that do NOT follow
It does not imply that animation is universally harmful or that a lively site lacks intellectual depth.

Evidence boundary: no documented personal outcome is represented by this card.

A09 / 15Teaching scenario

Writing / illustrative reconstruction

The essay with seven introductions

Including every interesting opening will communicate the full nuance of an argument.

Inspect the attempt
Hypothesis
Including every interesting opening will communicate the full nuance of an argument.
Resources & constraints
Seven draft openings, a claim map and an intended reader question.
Version history
v0 many preambles → v1 clear proposition → v2 one example, counterargument and uncertainty note.
What was tested / proposed
Read only the first 120 words and ask: can a stranger identify the question and why it matters?
What failed / could fail
An accumulation of context delays the point; editing can also erase important caveats.
Observed result
Editorial example: this is a constructed draft pattern, not an identified rejected essay.
Revised approach
Lead with the intellectual problem, then restore complexity where it changes the conclusion.
Next question
Which caveat belongs before the claim, and which belongs after the reader understands it?
Lessons that do NOT follow
It does not mean short writing is always superior or that a long opening is necessarily indulgent.

Evidence boundary: no documented personal outcome is represented by this card.

A10 / 15Unverified worksheet

Writing / fill with future evidence

The ambitious unfinished series

A six-part sequence will teach a topic better than a single complete guide.

Inspect the attempt
Hypothesis
A six-part sequence will teach a topic better than a single complete guide.
Resources & constraints
To document: outline, reader questions, draft dates, scope estimate and any feedback with permission.
Version history
v0 series promise → v1 publishable pilot → v2 revise scope after evidence; these are proposed stages only.
What was tested / proposed
Not yet tested: define one reader task and evaluate a stand-alone pilot before declaring a series.
What failed / could fail
Possible failure mode, not observed: dependencies between chapters could prevent any useful chapter from shipping.
Observed result
No observation supplied. Fill this field with dated evidence only after a real attempt.
Revised approach
Choose one self-contained lesson with an honest “series possible” note rather than a promised release calendar.
Next question
What additional value would a second installment give to someone who read the first?
Lessons that do NOT follow
It does not show that long-form projects are impractical or that this author has abandoned such a series.

Evidence boundary: no documented personal outcome is represented by this card.

A11 / 15Teaching scenario

Perfumery / illustrative reconstruction

The accord that became too crowded

Adding more aromatic materials will create a richer emotional impression.

Inspect the attempt
Hypothesis
Adding more aromatic materials will create a richer emotional impression.
Resources & constraints
A hypothetical ingredient ledger, blotters, safety documentation and dilution notes; no formula amounts or use advice.
Version history
v0 many materials → v1 subtractive accord → v2 blind-labelled comparisons with a consistent method.
What was tested / proposed
Plan a small comparative smelling session with labelled samples, adequate rest and notes on perception.
What failed / could fail
Without a controlled comparison, ingredient changes and adaptation confound impressions.
Observed result
Imagined pattern, not actual formulation data: a complex accord may feel less legible, but no outcome is asserted.
Revised approach
Document one variable per revision and follow applicable IFRA guidance and supplier safety information.
Next question
How consistent are descriptors across sessions and people?
Lessons that do NOT follow
It does not prove that fewer ingredients smell better, that emotions are universal, or that any trial is skin-safe.

Evidence boundary: no documented personal outcome is represented by this card.

A12 / 15Unverified worksheet

Perfumery / fill with future evidence

The mood label that overpromises

A fragrance named after a feeling will predictably produce that emotion for every wearer.

Inspect the attempt
Hypothesis
A fragrance named after a feeling will predictably produce that emotion for every wearer.
Resources & constraints
To document: exact claim language, intended audience, ethical review, consent-based feedback and safe handling.
Version history
v0 deterministic emotion claim → v1 personal association prompt → v2 carefully qualified storytelling.
What was tested / proposed
Not yet tested: separate liking, memory association and mood reports; avoid treating self-report as a causal proof.
What failed / could fail
Known logical risk, not a trial result: words like “guaranteed confidence” exceed what a naming concept establishes.
Observed result
No individual outcome supplied. A mood association is not a clinical or universal efficacy result.
Revised approach
Frame scent as a possible invitation to reflection, not a medicine or psychological intervention.
Next question
How does autobiographical memory change the language people use about a scent?
Lessons that do NOT follow
It does not establish therapeutic benefit or validate a psychological theory of fragrance.

Evidence boundary: no documented personal outcome is represented by this card.

A13 / 15Teaching scenario

Career / illustrative reconstruction

The impressive role that misses the real goal

A prestigious-sounding role will necessarily create a better working life.

Inspect the attempt
Hypothesis
A prestigious-sounding role will necessarily create a better working life.
Resources & constraints
A values worksheet, role requirements, personal constraints and an informational interview plan.
Version history
v0 title-first choice → v1 daily-work inventory → v2 reversible trial where feasible.
What was tested / proposed
Compare the day-to-day tasks, learning room, compensation stability and autonomy with the decision criteria.
What failed / could fail
Status can become a proxy for fit; yet preferences may legitimately change with circumstances.
Observed result
Constructed trade-off: the example yields competing criteria, not a real career incident or employment outcome.
Revised approach
Define minimum requirements, gather direct information and separate identity benefits from work realities.
Next question
Which trade-off would remain acceptable after the novelty of a title faded?
Lessons that do NOT follow
It does not prove that corporate careers lack meaning or that independence is always preferable.

Evidence boundary: no documented personal outcome is represented by this card.

A14 / 15Unverified worksheet

Career / fill with future evidence

The independent-project pivot

A portfolio of small public experiments can replace every function of formal employment.

Inspect the attempt
Hypothesis
A portfolio of small public experiments can replace every function of formal employment.
Resources & constraints
To document: realistic costs, income uncertainty, skills, support, milestones and private risk boundaries.
Version history
v0 all-or-nothing ambition → v1 limited pilot → v2 decision review; no claimed historical sequence.
What was tested / proposed
Not yet tested: decide which skill, audience benefit or earning assumption one bounded project could probe.
What failed / could fail
Possible failure mode: public output and personal learning are not automatically reliable income or external validation.
Observed result
No observed personal career result supplied. Record future evidence against pre-stated criteria.
Revised approach
Treat independent creation as one possible portfolio component, with optionality and financial reality visible.
Next question
Which benefits matter even if no audience grows, and which claims require external evidence?
Lessons that do NOT follow
It does not establish someone’s employability, entrepreneurial success or financial future.

Evidence boundary: no documented personal outcome is represented by this card.

A15 / 15Teaching scenario

Research / illustrative reconstruction

The elegant model without a decisive test

A coherent framework about intention and action is already an empirically supported explanation.

Inspect the attempt
Hypothesis
A coherent framework about intention and action is already an empirically supported explanation.
Resources & constraints
Construct definitions, rival explanations, a falsifiable prediction and a preregistration reading list.
Version history
v0 persuasive diagram → v1 operational definitions → v2 small feasibility design → v3 independent replication if warranted.
What was tested / proposed
Ask what observation would be more likely under the proposed model than under established alternatives.
What failed / could fail
Conceptual coherence can be mistaken for predictive evidence; the same story may fit opposite results.
Observed result
Methodological illustration: no Identity–Action Gap study result or validation is claimed.
Revised approach
Label the framework a proposal, plan measurements, consider confounders and distinguish exploration from confirmation.
Next question
Can a prediction be made specific enough to be wrong, without oversimplifying human behaviour?
Lessons that do NOT follow
It does not disprove the idea, establish priority over existing scholarship or imply empirical validation.

Evidence boundary: no documented personal outcome is represented by this card.

05 / THE REVERSIBLE REPAIR LAB

Try the before.
Inspect the after.

These are small constructed demonstrations—not screenshots of a real broken Sahilverse site. Change states and see why a repair matters. All actual interactive controls remain keyboard accessible, even when an illustrative “before” state is shown.

REPAIR / 01

Make focus visible

Keyboard navigation is an interface, not an afterthought.

ILLUSTRATION: BEFORE / faint indication

In the before illustration, a navigation target gives almost no focus indication. The repair adds a high-contrast outline.

REPAIR / 02

Let errors teach

A failure message should identify a remedy, not only a problem.

ILLUSTRATION: BEFORE / opaque message

!Error 42.Please enter a title before saving. Your other notes are unchanged.

The repaired copy names the missing field and reassures the writer that existing text is safe.

REPAIR / 03

Give a target room

Visual delicacy should not make tapping a guessing game.

ILLUSTRATION: BEFORE / cramped target

A larger activation area and adequate separation reduce accidental taps. WCAG 2.2 has nuanced target-size requirements and exceptions.

LIVE KEYBOARD EXERCISE / OPERATIONAL

Can you find the focus?

Press Tab to enter the three controls below, Enter or Space to activate one, and Shift + Tab to go backwards. Your selection is announced in text. Use Tab to leave the exercise; there is no trap.

No selection yet. The controls work with a mouse, touch, or keyboard.

06 / AN ENCYCLOPEDIA OF FAILURE MODES

Give the problem
its proper name.

Diagnosis is not a verdict. These eight patterns suggest what to inspect, not what must have happened.

01 Confounded change +

Mechanism: several variables change together. Signal: a redesign alters copy, layout and audience simultaneously. Repair: isolate variables when feasible; otherwise record the bundle and avoid causal claims. Counterpoint: real-world design often requires combined interventions.

02 Survivorship story +

Mechanism: lessons are drawn from visible winners while abandoned or unseen attempts disappear. Signal: only launch-day screenshots remain. Repair: log decisions and abandoned routes before their outcomes are known. Counterpoint: archiving everything also creates noise and maintenance cost.

03 Proxy substitution +

Mechanism: the easily counted thing quietly becomes the goal. Signal: view counts are treated as comprehension. Repair: name the actual value, then decide whether a metric has any useful relationship to it. Counterpoint: imperfect measures can still offer directional clues.

04 Unclear denominator +

Mechanism: a few responses are interpreted without knowing exposure or opportunity to respond. Signal: “three people liked it” without knowing who saw it. Repair: record the sampling frame and selection limits where possible. Counterpoint: unsolicited qualitative notes can still spark revisions.

05 Outcome hindsight +

Mechanism: the earlier decision is narrated as obviously wise or foolish after the result. Signal: no dated hypothesis exists. Repair: time-stamp expectations and plausible alternatives. Counterpoint: retrospective reflection is useful if honestly labelled.

06 Invisible accessibility regression +

Mechanism: a visual improvement obscures operation for keyboard or assistive-technology users. Signal: removed focus style, unlabeled field, inaccessible custom control. Repair: use semantic elements and test with keyboard, zoom and assistive technology. Counterpoint: an automated checker alone cannot certify accessibility.

07 Safety debt +

Mechanism: the prototype assumes everyone can absorb its risks. Signal: personal information is retained without a clear reason or consent. Repair: minimize collection, prototype locally and require higher scrutiny for irreversible consequences. Counterpoint: not every experiment requires formal research governance.

08 Premature generalization +

Mechanism: one observation becomes a universal rule. Signal: “no one wants this” after one silent launch. Repair: state scope, sample and unanswered alternatives. Counterpoint: acting under uncertainty is sometimes necessary; action need not masquerade as certainty.

07 / THE PRIVATE FIELD DESK

Leave a record.
Not a legend.

Write one real attempt, including uncertainty. Nothing is transmitted. Your notes stay in this open browser tab unless you explicitly enable storage on this device. Export before closing if you prefer not to store locally.

PRIVACY NOTE

No account, backend, tracker or automatic upload. Optional localStorage persists on this browser profile, where other people with access to the device may see it. Private browsing, site-data removal or device policies can delete it.

Do not record confidential, identifying or sensitive information in a shared browser.

↓ Download a reusable Markdown template
FIELD LOG / LOCAL ONLYNEW ENTRY
3–110 characters; no names or confidential details required.

Your private entries

0 entries

Your entries are only kept in memory unless you opt in to local storage.

  • No entries yet. The first attempt can be small.

Delete your private log?

This will permanently remove the entries currently in this tab and any saved local copy. Export a copy first if you want to keep them.
08 / FIELD EXERCISES

Three exercises for
an actual Tuesday.

10 MIN / RECONSTRUCT

01. The version autopsy

Choose any previous draft. State the decision that produced it, what you knew then and what you know now. Write one alternative explanation for the outcome. Avoid “I should have known”.

Output: a dated one-paragraph decision record.

15 MIN / CONTRAST

02. One question, two theories

Take a quiet launch. Theory A: the audience did not notice. Theory B: the offer was unclear. Name one observation that could distinguish them without assuming either is correct.

Output: one possible discriminating test.

05 MIN / KEYBOARD

03. The no-mouse audit

Put the mouse aside. Tab through your own page, activate a control, reverse direction, and look for hidden focus, missed labels or keyboard traps. Note what you could and couldn't verify.

Output: one reproduction step and one repair hypothesis.

09 / POCKET GLOSSARY

Terms worth
keeping straight.

These are practical working definitions, not a substitute for specialist research methods.

Hypothesis
An expectation that could in principle be contradicted by observation.
Exploration
Investigation that generates ideas and plausible questions, without claiming a pre-specified confirmation.
Confounder
Another factor that makes the relationship between an intervention and an outcome hard to interpret.
Iteration
A revised version informed by constraints, feedback or a newly clarified goal.
Counterfactual
A credible alternative account of what might have happened under different conditions.
Regression
A previously usable or effective property made worse by a change.
Evidence boundary
The line between what a record directly supports and what remains interpretation.
Reversibility
How easily a choice can be undone; a central consideration when experimenting with other people's experience.
10 / NOTES, SOURCES & LIMITS

A library card,
not a certificate.

What we know: linked institutions publish guidance on iterative design, accessibility, usability and preregistration. Interpretation: preserving dated hypotheses and alternatives can improve the honesty of creative review. Hypothesis: this archive may help visitors evaluate their own attempts more carefully. Fiction: the scenario narratives are constructed teaching devices.

None of the 15 files contains independently verified user-specific failure data. No accessibility conformance certificate, behavioural effectiveness trial, or outcome measurement is claimed for this website.

Sources were checked on ; documents and destinations may change. This is educational material, not professional clinical, safety or legal advice.

  1. Design Council — Framework for Innovation / Double Diamond ↗Supports iterative divergence, convergence, testing and revision; not proof of any case outcome.
  2. W3C — Web Content Accessibility Guidelines 2.2 ↗Normative accessibility criteria including keyboard operation, focus and target size.
  3. W3C WAI — Labeling Controls ↗Practical guidance for clearly associated form labels.
  4. Nielsen Norman Group — Ten Usability Heuristics ↗Established usability heuristics for error prevention, feedback and recovery; heuristics are not a substitute for user testing.
  5. Center for Open Science — Preregistration ↗Distinguishes planned from exploratory research; do not apply formal statistical claims to fictional cases.

THE END OF THE FILE IS NOT THE END OF THE QUESTION

What changed in your thinking,
besides the outcome?

Make room for trials, repairs and the right to revise. Keep your account specific enough that a stranger could disagree with it.

Record an attempt ↗Send a field note by email ↗↖ Return to Sahilverse