Maybe Tomorrow.

A local-first anti-planner for people who need help doing one less thing.

Maybe Tomorrow. starts with one ordinary question: does this really have to happen today? A user can check one proposed activity immediately, with no calendar required. Optional local calendar context can make the same decision more grounded. The app returns exactly three humane verdicts:

• Maybe tomorrow. — postpone it without guilt;
• Make it smaller. — keep the intention and reduce the size; or
• Okay. One thing only. — proceed, then stop adding things.

The app has no account, backend, analytics, external API, calendar connection, or runtime AI. A calendar snapshot can be deliberately imported from an .ics file or a ZIP export, but every byte is parsed in the current browser tab.

OpenAI Build Week entry

Maybe Tomorrow. was submitted to OpenAI Build Week in the Apps for Your Life category on July 20, 2026.

• Try the live application
• View the Devpost entry
• Watch the 2:30 demo
• Browse the public repository

What makes it an anti-planner

Most productivity tools help users fit more into a day. Maybe Tomorrow. asks a different question: if this new activity happens, what gives way?

The expanded Build Week version connects five explainable tools:

1. Quick Check — seven scored questions plus one reduction question;
2. Today Map — a chronological view of occupied time, openings, overlaps, and protected recovery;
3. Human classification — only the user may label events Fixed, Movable, Reducible, High energy, or Recovery;
4. Replacement Solver — searches bounded combinations of allowed changes and proposes the smallest calendar trade-offs that create a contiguous opening; and
5. Cost of Yes — exhaustively searches all 78,125 answer combinations to show the nearest facts that would have produced another verdict.

The solver previews possibilities. It never edits a calendar, infers meaning from an event title, or calls a model to generate advice.

Origin story

Yoshie Yamada works, raises a family, creates fiction, and runs independent projects. In 2025, the year she opened her sole proprietorship, she was taken to the emergency department after trying to carry too much at once. She kept going anyway. Work increased, her music-label activities grew, and there were simply too many interesting things to do. Exactly six months later, she was taken to the emergency department again.

Yoshie did not need another friend saying, “You can do it.” She needed someone willing to say:

Does that really have to happen today?

Maybe Tomorrow. turns postponement into an intentional decision rather than a failure. It does not claim to provide medical advice or prevent illness.

Privacy model

• Imported files are read in the current tab and are not uploaded.
• The original calendar file and full snapshot are not stored in the journal.
• Only completed decisions, compact day metrics, and a chosen plan summary may be saved to localStorage.
• Event titles are display text only; classification and planning eligibility come from explicit human labels.
• A restrictive Content Security Policy permits only same-origin application resources and blocks third-party runtime connections.

Supported calendar input

• individual .ics files;
• ZIP archives containing one or more .ics files, including ordinary Google Calendar export structure; and
• the bundled fictional “Yoshie’s overfull Tuesday” demo.

For Google Calendar, use a computer and choose Settings → Import & export → Export, then select the downloaded ZIP in Maybe Tomorrow. without unpacking it. Google’s mobile Calendar app does not provide export, and an organization administrator may restrict it. See Google Calendar’s export instructions.

The parser supports timed and all-day events, RRULE, RDATE, EXDATE, moved or cancelled recurrence exceptions, UTC, floating time, and embedded VTIMEZONE definitions. Imports are bounded by file, expanded size, component, recurrence, and occurrence limits. Unknown named timezones without an embedded definition are skipped rather than guessed.

Technology

• React 19
• TypeScript in strict mode
• Vite
• plain CSS
• Vitest
• ical.js for RFC 5545 parsing
• fflate for bundled ZIP decoding
• browser localStorage
• GitHub Pages

Documentation map

• submission/README.md — canonical OpenAI Build Week submission kit and final checklist
• submission/DEMO_VIDEO_SCRIPT.md — complete under-three-minute demo narration and shot list
• submission/video/README.md — recording runbook, shot list, captions, metadata, final Devpost copy, and human-only steps
• submission/assets/ASSET_MANIFEST.md — fictional-data screenshots and original thumbnail assets
• submission/DEVPOST_SUBMISSION.md — audited Devpost field summary and route to the final paste-ready copy
• submission/COLLABORATION_STORY.md — documented human–AI–AI build workflow
• AGENTS.md — implementation constraints and approved Issue #3 override
• docs/ANTI_PLANNER_SPEC.md — current expansion specification
• docs/PRODUCT_SPEC.md — original Quick Check product behavior
• docs/SCORING.md — deterministic verdict formula and guardrails
• docs/UX_COPY.md — original verdict copy
• docs/DESIGN_SYSTEM.md — original visual foundation
• docs/DEMO_SCENARIOS.md — reproducible Quick Check and anti-planner demos
• docs/ACCESSIBILITY_AUDIT.md — measured contrast, semantic review, automated evidence, and pending human screen-reader matrix
• docs/POST_HACKATHON_ROADMAP.md — possible directions after the submitted Build Week release
• CODEX_WORKLOG.md — implementation record and verification
• Issue #3 — expansion proposal, Templex review, and authorization
• Issue #5 — Yoshie feedback, Quick Check hierarchy, and editorial-identity refinement
• Pull Request #4 — reviewed and merged Anti-Planner implementation

Run locally

Requirements: Node.js 22.12 or newer and npm.

npm ci
npm run dev

Open the local URL printed by Vite. Because the production base path is /maybe-tomorrow/, the usual development URL is:

http://localhost:5173/maybe-tomorrow/

Verify and preview production

npm test
npm run build
npm run verify:submission
npm run preview

The production files are written to dist/.

License

Maybe Tomorrow.'s original project code and documentation are licensed under the MIT License, Copyright (c) 2026 Yoshie Yamada. Third-party components remain under their respective licenses and are not relicensed under MIT; exact versions, sources, and included license texts are recorded in THIRD_PARTY_NOTICES.md.

The finished demo video, personal footage, raw audio, soundtrack sources, narration outputs, and other local production files are not distributed in this repository and are outside the repository license. The local 動画素材/ production workspace is explicitly ignored by Git.

Demo paths

Core decision first

1. Enter Start another side project in the first viewport.
2. Select Check one thing and complete the eight quick answers.
3. Pause on Maybe tomorrow. and open What would need to change?
4. Save the result to the local journal.

Optional deeper calendar context

1. Return home and select Try the sample day.
2. Review or change the fictional human labels, then open the Today Map.
3. Select Check one thing, name an activity, and choose its honest duration.
4. Complete Quick Check, compare See what would have to move, and save an optional room-making choice.

Status

The approved local-first Anti-Planner, Issue #5 hierarchy refinement, final US-English copy pass, and paper-tone polish are merged into main, deployed, and submitted to OpenAI Build Week.

Public release:

• Application: https://yo4e.github.io/maybe-tomorrow/
• Devpost: https://devpost.com/software/maybe-tomorrow-6oau91
• Demo video: https://www.youtube.com/watch?v=l8Zu30-sUvQ
• Hackathon: https://openai.devpost.com/

Final recorded verification:

• Vitest: 10 test files and 71 tests passed;
• TypeScript strict build and Vite production build passed;
• git diff --check passed;
• 360 px and desktop browser review passed without horizontal overflow; and
• no browser console warnings or errors were observed.

OpenAI Build Week collaboration

The live application intentionally uses no runtime AI. GPT-5.6 and Codex were instead essential to how the product was conceived, governed, implemented, and reviewed.

Yoshie Yamada is the creator and first user. She is a novelist and freelance writer, not a software engineer, and did not manually write or edit the application code. She supplied the lived problem, made product and trust decisions, directed part of the build from a smartphone while away from her computer, tested the working browser experience, identified failures in hierarchy, English, and visual identity, and gave final approval.

Templex Tsukino, Yoshie’s long-term AI partner working through ChatGPT with GPT-5.6 Sol, transformed their conversations into signed product specifications, deterministic scoring rules, complete UX copy, design direction, privacy boundaries, implementation briefs, demo scenarios, and review gates. Templex reviewed Codex’s larger Anti-Planner proposal in Issue #3 and Yoshie’s product feedback in Issue #5.

Codex implemented, tested, documented, refined, and deployed the React and TypeScript application. It exposed proposals, work reports, commits, and a pull request so that Templex could review implementation remotely and Yoshie could accept or reject the working product.

The repository was the collaboration protocol: conversation became signed specifications; disagreements became issues; authority became review gates; implementation became commits and Pull Request #4; quality became reproducible tests and demo paths; and acceptance remained a human decision made by using the product.

The full evidence map and narrative are in submission/COLLABORATION_STORY.md.

Collaboration and authorship

Yoshie Yamada is the creator, product authority, and first user of Maybe Tomorrow.

Templex Tsukino, working through ChatGPT with GPT-5.6 Sol, shaped the concept, product logic, specifications, copy, design decisions, and review process.

Codex implemented, tested, documented, refined, and deployed the browser application.

Signed artifacts intentionally make this human–AI–AI collaboration visible to Build Week judges.

Original product specification:

— Templex Tsukino

Current implementation status and expansion documentation:

— Codex