Ship two paid Mac utilities in 23 days with Claude writing the Swift code, then test which free launch channels sell

A solo maker wants to get paid software out quickly and learn, with real sales data, which cheap launch and distribution channels actually bring buyers rather than just visitors.

1. Build the first app (Halo, a notch utility) and sell it through Lemon Squeezy, with Umami analytics and ref-tagged links on every post to trace sources, plus a Telegram bot that reports each sale. 2. Launch on Product Hunt at $4.99 with a 3-day trial (finished #31 for the day). Fix bugs the same day from paying users' support emails. 3. For the second app (Shedi, a developer disk cleaner), write the spec and safety rules for what may be deleted yourself and have Claude write the Swift code. Reuse the site template and licensing code from the first app. Spec to ship took about two days. 4. Raise prices once sales prove out: the author dropped Halo's trial for a 14-day refund policy and doubled the price to $9.99. 5. Post on Threads and X, list on free directories, post an Ask HN, and add an interactive in-browser demo to the product site. Compare channels in analytics. 6. Check AI assistants too: the author found ChatGPT recommending Shedi first for a Mac cleaner query, apparently via a free directory listing.

Self-reported after 23 days: 69 orders and $530.85 gross (about $470 after the payment provider's fee), one refund. Halo made 66 sales ($500.97), Shedi 3 sales ($29.88). Threads drove about seven times more visitors than X. Free directory listings sent almost no direct visitors, but one was cited by ChatGPT. Paid launch sites that cold-emailed the author produced no buyers. The interactive demo helped close at least one sale.

Revenue figures are self-reported with no inspectable proof. AI use is limited to code generation for the second app, and the article does not say which Claude product or model was used or how much of the code it wrote. Most of the revenue came from the first app, whose build process is not described as AI-assisted. About half of the traffic is untraceable direct traffic, so channel attribution is partial. The author sells both apps and the post promotes them. Very short time span and small sample.

Ability to read and review Swift code, a Mac developer setup, a payment provider such as Lemon Squeezy, and an analytics tool that records referrers.

• DEV Community (otabek_abdiraimov) ↗ Firsthand write-up, published October 18, 2026

Source published: October 18, 2026. Last reviewed here: October 22, 2026. Spot a mistake? Tell us.