Example: Launching Fernleaf, an AI field notebook

Example content. This case study is a placeholder to show how work entries look. Replace it with your own project.

The seed

Fernleaf began as a question asked on a long walk: what if a notebook could listen the way a forest does — patiently, without judging, and remembering only what matters? The team had a working prototype that summarised voice memos, but no clear reason for anyone to care.

What we learned in beta

We invited 212 people into a private beta over nine weeks. Three findings shaped the launch:

1. People did not want summaries, they wanted threads. Linking an idea from Tuesday to one from three weeks earlier drove 63.4% of repeat sessions.
2. Trust was earned in the first minute. Showing exactly which sentence an insight came from cut “I don’t believe this” feedback by more than half.
3. Quiet beats clever. Removing proactive notifications raised week-four retention from 31% to 44.7%.

Shaping the launch

We rewrote the product story around one idea — your thinking, connected — and cut the feature list on the landing page from eleven items to three. The launch sequence ran over two weeks:

• A short, honest founder note explaining what the model does and does not do
• A public changelog, so early users could see their feedback turn into shipped work
• A small, invite-based waitlist that let us scale inference costs gradually

Outcome

Fernleaf reached 4,870 sign-ups in the first fortnight, with a model-cost per active user that stayed 22% below forecast. More importantly, the support inbox was full of people describing moments rather than bugs.

What I would do differently

Start the evaluation harness earlier. We spent the last week before launch hand-checking outputs that a small, well-labelled test set would have caught in minutes.