Build a school.
One keystroke
at a time.

An app that helps someone learn a subject or skill: interactive lessons, revision games, flashcard trainers, or an explainer. It could teach something absent from a traditional timetable, and expertise in the topic was allowed.

The Build Your School Discord forum, with a link to the app and reaction-based voting under the published rules. Sam said the event was temporarily returning to Discord while Launchpad issues were fixed.

Yes. Sam repeatedly linked the unofficial archive for recordings and recaps.

Base44 is the product/company; Base One is its proprietary LLM choice in the builder. Sam chose it for the first implementation, then compared other approaches.

Yes. Sam described Fable as an expensive option and warned users watching their budget. This session did not establish a complete Fable price comparison because that run stalled.

Sam said they go to hosted storage rather than remaining temporary blob data. An app can process or move them and implement cleanup, but the session did not specify storage quotas, retention rules, or a particular external storage connector.

Sam treated an unexpected login in the builder preview as a context/routing issue to prompt the builder to fix, rather than repeatedly signing in. Published-app authentication and user permissions still require separate tests.

Sam said it should not slow the server-side process, though multiple tabs may slow the local computer. Some runs in this session stalled; he did not establish their cause.

The intended mastery loop combined typing speed and accuracy, represented by stars or progress. Sam tested some working gates but also found inconsistent placement and progression, so not every version had a correct completed curriculum.

Daily streaks, XP, and a mastery path were central. Leagues, coins, avatars, streak freezes, leaderboards, multiplayer, and groups were discussed or appeared in different versions; not every feature was implemented and tested.

Sam said the team preferred to direct development resources elsewhere. He did not provide a detailed policy, timeline, or migration procedure.

Yes. Sam recommended clear visual references and mentioned the design-reference site AJ used in Tuesday's session. A reference communicates direction, but still needs an explicit statement of the intended device and user task.

Sam said the choices belonged to the redesigned Plan Mode experience, not one particular LLM. The group selected the retro 8-bit option.

Sam said the pre-build planning conversation did not consume credits and that the builder run did once initiated. This is his session-time explanation, not a verified current billing guarantee.

Sam said it was possible to integrate 3D work, but recommended suitable libraries or a dedicated engine such as Unity for more demanding projects. He mentioned a community MMO effort as an example, not a proof of arbitrary performance or feature support.

Sam read out 1.6 credits for Base One, Terra, and the Plan Mode result. Sol and Fable did not complete normally, so the session cannot support a full five-model cost or quality ranking.

Sam said the stalled runs were not normal and suspected a builder issue. Refreshing did not recover them during the session; he abandoned those attempts without a confirmed diagnosis.

No. Sam explicitly contrasted it with Build for Good: Build Your School rewarded winners. Separate limited coupon codes were promised in chitchat.

Sam could not give a date. Existing Plan Mode was available, but the redesigned version shown on his account might not yet have reached everyone.

Sam intended to publish it after the remaining prompts completed, reuse the L2 Type destination, and potentially make it a template. The recording ended before that final publication, so this recap does not claim the final app was deployed or provide an unverified URL.

Yes as an extension. Sam discussed language choices and internationalisation, but the session prioritised the core typing flow and did not complete multilingual or alternate-layout testing.

He said not generally, citing time and the need to remain impartial as staff. The builds shown in this session were created live by him.

Sam favoured broader batches at the beginning and increasingly narrow changes as context grows. His approximate 50- and 200-prompt thresholds were personal heuristics, not product limits.

Sam favoured an existing service or API because a truly functional consent system takes more work than a visual banner. No provider or legal compliance was verified in the discussion.

Sam said yes, while showing a much smaller initial build cost as an example. It was not a guarantee for every app's complexity or a total cost measured for all the experiments in this session.

No. Sam explicitly rejected that claim. The session did not publish a comprehensive pricing table.