The REMEMBR app icon: a cream rounded square with a mark of concentric orange arcs resembling a fingerprint.

MRP Thesis Defense

REMEMBR

Building an accessible journaling application, built without a programming background.

Raafi Islam Master of Digital Media · Toronto Metropolitan University

Dhaka, 2018

I recognized the face.I couldn't place the name.

The interaction was brief and uncomfortable. I made my excuses and left. What stayed with me was not the moment itself. It was realizing I had no way back to that person: no record of how I knew them, who introduced us, or what we had talked about.

A Recurring Pattern

It did not happen once.

It kept happening, until I started avoiding situations where it might come up again.

Epilepsy

Affects memory encoding directly. My own diagnosis, and the reason this pattern was not random.

Monthly support sessions

The same frustration, from other people, again and again. Existing tools are not built for how memory works when it is unreliable.

The Gap

Calendars store names. They do not store the reason you'd remember someone.

REMEMBR grew out of that gap: a journaling application that organizes entries around people rather than dates, and surfaces them through any connected context. A shared location. A mutual contact. An event.

Research Questions

Two questions drive this MRP

  1. RQ1

    How can a non-programmer use generative AI to design and build a functional software application, and what specific competencies does that process require?

  2. RQ2

    What are the practical limits, recurring failure modes, and design trade-offs of AI as the primary development tool, and how were they navigated?

Thesis

Generative AI has created a genuinely new pathway for non-programmers to build production-grade software. It is not frictionless.

It demands a distinct skill set, rooted in design, editorial, and project-management practice, not programming. The skills are real. So are the limits.

Methodology and Positionality

How this MRP was studied: the evidence it draws on, and what first-person, practice-based research can and cannot claim.

Evidence Base

Two kinds of primary evidence

AI conversation transcripts

ChatGPT conversations, October 2025 through April 2026, plus nine Claude Code session transcripts, April through June 2026: the full sequence of directives, AI responses, and corrections.

Versioned application files

Two distinct versioning sequences: fifteen named versions from the early development phase, each preserved as a file, and a separate series from the first five days of the Claude Code phase.

Positionality

Designer, director, and subject

I did not begin with a research framework. The process was driven by personal experience and iterative making: trying things, observing what happened. This resembles what design researchers call research through practice (Frayling, 1993): knowledge generated by the act of making, not by observing from outside it.

What this makes possible

Access to design intent no external observer could reconstruct. I knew, at every moment, what gap existed between the output and the goal.

What it cannot guarantee

Full objectivity about my own experience, addressed by grounding claims in dated transcripts, not memory alone.

Limitations

What this study cannot claim

A single case

This cannot produce statistical claims about non-programmer AI-assisted development as a general phenomenon.

No external verification

Findings are corroborated against dated transcripts, but the interpretation of that evidence remains mine.

Theory found after the fact

The Actor-Network Theory connection was recognized in retrospect, not tested as a hypothesis.

What it can produce: a detailed, internally consistent account of one instance of this process, precise enough to generate propositions later research can test.

REMEMBR

The artifact this MRP produced, and the accessibility logic behind every decision in it.

The Artifact

A people-first journal, not a dated diary

Users build profiles for the people in their lives, linked to places, events, and shared contacts. Remember any one connected thing, and you should be able to find the person.

Built for

Epilepsy · Prosopagnosia · Traumatic brain injury · ADHD

8 months
From October 2025
13,000
Lines, in a single HTML file
0
Lines of production code written by hand
REMEMBR home screen on a dark background. A greeting reads Good morning, Raafi. Below it a Your People row shows circular portraits of Adam, Olivia, Daniel and Sara, then prompt cards reading Who did you meet today? and Siobhan's birthday in 2 days.
Figure 9. Interface. Generated with Claude (Anthropic, 2026).

Design Principle

Accessibility is the baseline, not a feature

Governed throughout by WCAG 2.1 AA, Ontario's AODA, and W3C COGA guidance for cognitive accessibility. Every setting in this panel is user-controlled, not auto-detected.

Perceivable
Atkinson Hyperlegible type. A five-phase color process ending in a sky-teal palette verified at WCAG 2.1 AA. Text-to-speech for saved memories and profiles.
Operable
Voice dictation in every text field, not a separate voice mode, tuned for accent diversity. Touch targets over 44 by 44 CSS pixels. Reduced motion, no strobing.
Understandable
Easy Mode: one plain-language question at a time, full feature parity.
Robust
ARIA labeling, full keyboard navigation, screen-reader optimization.
REMEMBR's Accessibility settings panel. Labelled toggles for Easy Mode, High Contrast, Reduce Motion and Haptic Feedback, a Text Size control set to 100 percent, and a Voice and dictation language selector.
Figure 3. Accessibility Features. Generated with Claude (Anthropic, 2026).

Signature Feature

The Connection Web

A force-directed graph where people, places, memories, and events are equally weighted nodes. Enter from any point, a mutual friend, an event, and trace outward to the person you are trying to reach.

Structurally resonant with Actor-Network Theory (Latour, 2005): a connection recognized only in retrospect, after the interface was already built from lived experience rather than from theory.

The Connection Web in REMEMBR, titled Adam's Web with four connections. Adam sits at the centre of a dark canvas with thin teal lines radiating out to Daniel above, Olivia to the right, Sofia below and Jesse to the left, with a list of the four connected people beneath the graph.
Figure 4. Connection Web. Generated with Claude (Anthropic, 2026).

Signature Features

Two more ways in

Associative Search

One field searches every content type at once, without category pre-selection: matching how episodic memory is actually retrieved (Tulving and Thomson, 1973).

REMEMBR search results for the single word coffee. Matching people Priya Nair and Sara Williams appear as cards, followed by a heading reading 13 memories found and memory entries including Full circle: Priya meets Marcus and Coffee catch-up with Adam.
Figure 5. Search. Generated with Claude (Anthropic, 2026).

Easy Mode

App-wide progressive disclosure. Every multi-field form becomes a one-question wizard, grounded in cognitive load theory (Sweller, 1988), without reducing functionality.

REMEMBR in Easy Mode. The home screen asks What would you like to do? and offers four large, plainly labelled choices: Add a person, Add a memory, See my people, and My memories.
Figure 2. Easy Mode. Generated with Claude (Anthropic, 2026).

Signature Features

Designed to remove friction, not add polish

Avatar Creator

For users without photos of everyone they track. Reference standard: Snapchat's Bitmoji. Three rounds of generated output were rejected before the fourth met that standard.

REMEMBR's avatar creator. A large illustrated face preview sits above a Randomise button, a row of eight selectable skin tone swatches, and a grid of five face shape options with the last one selected.
Figure 6. Avatar creation. Generated with Claude (Anthropic, 2026).

Social Card Generator

A canvas-based, shareable card acknowledging someone's place in the user's life. "I REMEMBR YOU."

REMEMBR's Share your memory screen. A generated card for Adam Chen, labelled Colleague, reads We met at DG8011 class on Thu, Sep 5, 2024. Sat next to each other on the first day, above the tracked words I REMEMBR YOU.
Figure 8. Social Card. Generated with Claude (Anthropic, 2026).

Findings

What eight months of session transcripts and versioned files reveal about this development model.

Finding 1 of 3

Four competencies, not one line of code

  1. Specification

    Translating design intent into language precise enough for the AI to act on correctly: a continuous discipline, not a single act.

  2. Evaluation

    Judging whether an output meets the standard without reading the code, using behavior, comparison, and testing as proxy signals.

  3. Direction

    Corrective instruction specific enough to close the gap. Naming a reference product outperforms "make it better."

  4. Continuity Management

    Holding a coherent product vision across many separate sessions with tools that have no persistent memory.

Finding 2 of 3

Four failure modes, recurring by structure

  1. Silent Feature Regression

    Confirmed features vanished between sessions with no warning. The AI worked from immediate context, not a persistent inventory.

  2. Persistent Hallucination

    The same factual errors recurred across sessions, for example claiming local-only storage, when REMEMBR uses hybrid cloud sync.

  3. Token Limit Constraints

    Productive windows cut to 15 to 20 minutes before context exhausted; usage-limit resets could last up to five hours.

  4. Code Token Exhaustion

    At 13,000 lines, loading the existing file into a new session consumed a substantial portion of the context window before any new work began.

Finding 3 of 3

Three practices developed in response

  1. Structured Context Documents

    Dense, machine-readable briefings pasted at the start of each session, because a chat tool opens every conversation with a blank context window. Claude Code eased this later in the build: it keeps project context in files on the machine, so continuity carries across sessions instead of being re-pasted.

  2. Standing Rules

    Every recurring AI error converted into a permanent, explicit constraint carried into every future session: never re-explained, always enforced.

  3. Forced Review and Planning Cycles

    Mandatory token wait periods became a built-in cycle to test live updates, find bugs, and map out future features on paper.

A Question This Raises

Who owns this, and is the data safe?

REMEMBR handles sensitive personal memories about real people. Ownership and data governance were design requirements from the outset, not an afterthought.

Code ownership

The law here is evolving and not yet settled. Copyright requires human authorship, but work in which a human directs the AI through creative choices can qualify for protection (U.S. Copyright Office, 2023). Anthropic does not claim ownership over generated content (Anthropic, 2024).

Training data exposure

The training opt-out was exercised throughout: the sessions in which REMEMBR was built were excluded from model training (Anthropic, 2024).

Platform choice

Rork was abandoned in part because it required a paid premium subscription to keep project data private. REMEMBR runs on infrastructure I own and can audit.

Discussion

Answering both research questions

  1. RQ1

    How can a non-programmer use generative AI to design and build a functional software application, and what specific competencies does that process require?

    A non-programmer can build production-grade software by developing four specific competencies: design and editorial skills, not programming ones, already held by a population far larger than the one that currently builds software.

  2. RQ2

    What are the practical limits, recurring failure modes, and design trade-offs of AI as the primary development tool, and how were they navigated?

    The limits cluster into four failure modes, all expressions of one root cause: a stateless tool operating on an incomplete picture of what it has already built. These are manageable through documented, repeatable practice.

Contribution and Next Steps

What this case adds, and where it goes

To the literature

A four-competency taxonomy precise enough to be tested, and a context-document practice that answers the multi-session, stateless-tool problem the existing literature does not engage with.

What comes next

Continued development on the same methodology. User testing beyond the researcher's own account. Production release beyond closed testing.

Ko et al. (2011) named the persistent gap between the people who need software and the people who can build it. REMEMBR is a direct instance of that gap narrowing, not because the technology got smaller, but because the skills it demands are already held by a much larger population than the one that currently builds software.

I REMEMBR YOU

Thank you.

Questions and discussion

Raafi Islam · REMEMBR · Master of Digital Media, Toronto Metropolitan University

References

Works cited in this presentation

  1. Anthropic. (2024). Privacy policy. https://www.anthropic.com/legal/consumer-terms
  2. Anthropic. (2026). Claude [Large language model]. https://claude.ai
  3. Braille Institute of America. (2019). Atkinson Hyperlegible font. https://www.brailleinstitute.org/freefont
  4. Frayling, C. (1993). Research in art and design. Royal College of Art Research Papers, 1(1), 1-5.
  5. Government of Ontario. (2005). Accessibility for Ontarians with Disabilities Act, 2005, S.O. 2005, c. 11.
  6. Ko, A. J., Abraham, R., Beckwith, L., Blackwell, A., Burnett, M., Erwig, M., Scaffidi, C., Lawrance, J., Lieberman, H., Myers, B., Rosson, M. B., Rothermel, G., Shaw, M., & Wiedenbeck, S. (2011). The state of the art in end-user software engineering. ACM Computing Surveys, 43(3), 21:1-21:44.
  7. Latour, B. (2005). Reassembling the social: An introduction to actor-network-theory. Oxford University Press.
  8. Sweller, J. (1988). Cognitive load during problem solving: Effects on learning. Cognitive Science, 12(2), 257-285.
  9. Tulving, E., & Thomson, D. M. (1973). Encoding specificity and retrieval processes in episodic memory. Psychological Review, 80(5), 352-373.
  10. U.S. Copyright Office. (2023). Copyright and artificial intelligence. https://www.copyright.gov/ai/
  11. W3C COGA Task Force. (2021). Making content usable for people with cognitive and learning disabilities. World Wide Web Consortium.
  12. W3C Web Accessibility Initiative. (2018). Web Content Accessibility Guidelines (WCAG) 2.1. World Wide Web Consortium.

Note. All REMEMBR interface figures reproduce figures from the MRP and show software generated with Claude (Anthropic, 2026). Figure numbers match the paper.

REMEMBRMRP Defense

Press → to continue