Curated, single-string prompts. Free to copy, free to run on Brainboot (up to a daily cap). Every one is written to be so structurally tight that using it teaches you something — even if you never pay us a cent.
No prompts published yet. Run supabase-prompts-tier.sql and supabase-prompts-additional.sql to seed the catalog.
Every prompt above is a string you copy into an LLM and hope. That's fine for one-off work. But the moment a prompt is load-bearing — when you need it to run the same way every time, catch its own mistakes, or compose with other prompts — it has to stop being a string and start being something closer to code. That's the rest of Brainboot.
Take the Long-Form Draft Engineprompt you just saw. Turn it into a brain and it gets: a Zod schema for input (voice config, thesis, audience), a Zod schema for output (body, word count, self-report), enforced invariants ("no slop phrases in output", "word count within 20 of target"), and a test suite that runs before every deploy. When it fails, it tells you why instead of silently shipping bad output.
Browse brains→One brain drafts a page. Another writes the SEO metadata from the draft. A third runs the draft against a quality rubric and returns pass / fix / block. A Blueprint chains them into a single execution where the output of one brain becomes the input of the next — all transactional. If any brain fails its invariants, the whole blueprint halts and tells you which step broke.
Browse blueprints→The Content Empire Circuit runs five blueprints three times a month: 150 SEO pages, 13 newsletters, 12 authority essays, 3 site audits, 500+ platform adaptations per quarter. You configure your voice once. The rest runs on its own, in your absence, with every run fully audited. A Circuit is what replaces a team — not a tool you use, a system that runs in the background.
Browse circuits→The prompts on this page are free because the single-string format has a ceiling. Past that ceiling is where we sell — not because we're trying to trick you into paying, but because reliable cognition genuinely requires more than a string. If the prompts above were enough for your use case, we're glad. When they're not, you'll know what's next.