All experiments

Experiment 01 / Magimica / 2026

How fast can a serious product actually be built?

Magimica turns one goal into a day-by-day plan — and helps you actually get it done.

Product strategyUXFull-stack engineeringAI-assisted delivery

01 / Hypothesis

Speed and seriousness are not opposites.

A small architecture and clear direction could take a useful product from foundations to beta in days. I would remain responsible for its coherence.

AI suggests. You decide.
You ownYour objectiveEditable at any time.
AI may helpSuggested missionsIdeas, never obligations.
You approveAccept · Edit · DeclineChoose what enters the plan.
Approved workYour focused planTrack actions and progress.

BenefitLess planning. More focus.

Risk controlAI changes nothing without your approval.

02 / Bet

Build the smallest complete loop.

Objective → missions → dependencies → schedule → completion. AI suggests tasks. A TypeScript engine calculates points, order, capacity and conflicts.

  • One codebase, separate modules
  • Migrations record schema changes
  • Validate AI output before saving
  • Deploy while learning

03 / Division of labour

Capable agents still needed direction.

AI leverage / broad

Help across the application

Agents supported research, architecture, task planning, code, tests, review, debugging and copy.

Human judgment / accountable

Three decisions stayed mine

The concept; the design and generated-image feature; the final QA and level of finish.

04 / Evidence

Five days, including the fixes.

Model → selective AI → working beta → production fixes → hardening. The sequence shows how the build changed, not whether the market wanted it.

Magimica planning screen showing a selected direction and actions to refine the plan or generate missions.
Production UI / direction before generation
Magimica review screen showing generated missions, dependencies and scheduling controls.
Production UI / review before scheduling
~5build days — owner account
07database changes in the build window
03test layers
02product languages

Where five-day delivery comes from

UX, architecture, implementation, QA and deployment advance together, pause for validation and converge into production. Five build days, reconstructed from history.

Objective; UX; Architecture; Code; QA; Deploy; Validation; PRODUCTION; 5 days.

  1. One goal launches five parallel workstreams
  2. Each stream pauses at its validation gate
  3. Validated work converges
  4. One product ships — in five days
  1. 01Aug 19

    Foundations

    The schema connects users, objectives, missions, dependencies and events.

  2. 02Aug 20

    Selective AI

    Generated illustrations arrive without changing the deterministic planning engine.

  3. 03Aug 21

    A working beta

    The full loop works. Database tests, secret scanning and authenticated browser checks follow.

  4. 04Aug 22

    Production answers back

    Localization and PWA support arrive alongside CORS, metadata, session and mobile fixes.

  5. 05Aug 23

    Hardening

    Security, forms, deployment, mobile navigation and Safari take the final day.

Source / repository history, dated database changes and first-person account.

Reality check / what speed exposed

Compressed implementation moved the bottleneck to verification. In a later, heavier agentic-QA experiment, systematic orchestration cost roughly twice the time and five times the tokens without consistently better acceptance results. My current rule is proportion: gate rigorously where failure is consequential; elsewhere, ship, observe and repair fast.

Visit Magimica