Project graveyard

Not everything survived

Ideas that were genuinely started — sometimes months of work — then shelved or cancelled. Knowing when to stop is as much a part of the craft as knowing how to ship.

  • Wendcodebase

    An IDE designed to work entirely with AI, conceived before tools like Claude Code, Codex or Antigravity existed.

    Why: Dropped: established players shipped exactly this kind of tool, with far more resources.

  • Retouche photo par sélection IA

    An advanced photo-editing tool: select a precise element in an image (e.g. five yellow cars) and have AI modify it — change their colour — without touching a single pixel outside the selection.

    Why: Dropped: established photo-editing players, far better equipped, moved into the same space.

  • Séquenceur visuel → code

    An n8n-style tool that generates real, performant Rust code from visually assembled nodes — giving non-technical users access to high-performance code.

  • PromptHub

    A library of roughly 1,200 finely tuned prompts for precise tasks (securing an application, building a module…).

    Why: Cancelled: no defensible business model (prompts are trivially copied), interest shifted to skills and agents, and nothing stops anyone from asking an AI to generate — or iteratively refine — a high-quality prompt directly. Over time the most useful prompts get absorbed into the platforms themselves anyway (like skills), the way Stack Overflow's answers ended up baked into the models.

  • Bookshare

    An online library for exchanging books for free between individuals: see nearby available books, contact their owner, track their availability.

    Why: Cancelled: lending books at scale without publisher agreements or librarian status isn't legal.

  • Coworkhome

    A platform for remote work at someone else's place — a neighbour offering coffee, a sofa, a spot to sit — so remote workers could meet up without travelling far, and get to know their neighbours.

  • Sideproject

    A platform for connecting project partners, with a contract module to formalise working methods and decision-making upfront — asking the awkward questions before they come up, not after.

  • Starcly

    Let landlords and tenants rate each other after a long-term rental.

    Why: Cancelled: the idea evolved into a much more ambitious project.

  • Glainet

    A chat app in the style of a proximity-based mini social network: small communities anyone could easily join, to meet people, share interests, organise events.

    Why: Cancelled after WhatsApp moved in a direction very close to this idea — nearly a year of Flutter development set aside.

  • TVA

    A platform to publicly flag individuals whose dishonesty was backed by solid evidence (filmed theft, an acknowledged but never repaid debt) — recovering debts and letting people be wary of someone already shown to be untrustworthy, as opposed to an unproven accusation.

    Why: Cancelled: publicly displaying someone's name is illegal in France regardless of evidence — they can demand its removal at any time, which would defeat the platform's entire purpose.

  • Procard

    A single developer profile, exportable as a skills dossier for any staffing agency without rewriting it each time — and eventually, a database of experienced profiles to ease recruiting.

  • Hexagone

    A turn-based strategy video game, built with a team of three.

    Why: Dropped: all three developers had to leave due to availability, and the technologies learned for it served no other purpose. A POC could be revived with AI's help.

© 2020–2026 Davy Nana — All rights reserved

--:--