The learning loop

The next draft
starts with a reason.

Connect what you made, what happened,
and what is worth changing next.

Different signals. Clear roles.

Know what the evidence
can actually tell you.

Editorial quality, your preferences and audience response answer different questions. Growth Factory keeps them distinct.

01

Quality

Does the draft have a clear hook, enough context and a complete payoff? Production checks and the editorial judge help identify a specific weakness.

A review aid, not an engagement forecast.
02

Preference

What do you want more of? Your feedback can inform brand-specific taste, with enough examples to distinguish a pattern from a one-off choice.

Creative policy, not model-weight training.
03

Performance

What happened to the post? Instagram observations retain their account, publication and measurement window, rather than becoming a context-free score.

Organic comparisons remain observational.

Measure → learn → create

Make the change
small enough to understand.

A useful experiment names the variable, shows its evidence and carries the result into the next production.

  1. 1

    Keep the observation

    Retain the actual account, post and time window. Missing numbers stay unknown. Collection runs while the app is open.

  2. 2

    Form a hypothesis

    The Strategist considers the evidence and proposes a bounded next test. If the evidence is insufficient, it can wait rather than name a winner.

  3. 3

    Update the direction

    A proposed Style rule shows what changes and why. Active rules remain explicit and versioned.

  4. 4

    Build the next draft

    Apply the supported change to a new edit. Compare the actual output, keep its lineage, and let the next real observations inform the following iteration.

The factory builder

When the next idea
needs a new tool.

A creative gap can become a software task. The builder prepares a reviewable extension instead of quietly changing the running app.

  1. Capability request

    Evidence identifies an edit the renderer cannot yet make.

  2. Code & checks

    An isolated worktree holds the extension, tests and actual sample render.

  3. Human review

    Inspect the diff, logs and output before a change is merged and loaded.

A documented progress-bar build passed checks in a temporary clone. It was not automatically merged or shipped.

See it in the app

A visible experiment.
No hidden leap.

The Mac app brings progress, evidence, a proposed change and the next draft into the Factory workflow. A rehearsal is visibly labeled as simulation: synthetic Instagram outcomes do not become real account history.

Does the system train its own model?

It uses evidence and versioned creative policy to guide future work. It does not retrain model weights or guarantee a better audience result.

What if the new version performs worse?

A better quality score does not guarantee stronger audience response. The result can support keeping the original or trying a different hypothesis.

Are real posts and simulations mixed?

No. Rehearsals retain their synthetic-evidence labels. Real observations need an actual account and publication mapping; an early measurement keeps its timing disclosure.

A connected growth engine

Create. Measure.
Make the next move.

Get started on Mac

No promised lift. A clear trail from evidence to experiment.