What I Built
SlumberBloom is a garden that only grows when you leave your phone alone.
You tap “Tuck In,” put the phone down and go to sleep. When you wake up, the app sees how long you were away. If it was at least 30 minutes, you get a Night Block: a SHA-256 hash chained to your previous night. Claim it as Morning Dew, settle it at Dawn and it becomes Bloom, which feeds your Baby Moonflower.
It rewards you for not using it. There’s no tracking, no microphone, no tokens and no wallets. Your hours of patience are verifiable, and a “Light the Lantern” button re-checks any block with a Merkle proof, right in your browser.
This is a v1 that proves the core idea. The big vision (eras, a world simulation, outdoor photo quests) is still being built, and what’s here is the working foundation.
Code
Armansiddiqui9
/
slumberbloom
Proof-of-patience garden - your sleep becomes verifiable Night Blocks with SHA-256 + Merkle
SlumberBloom
Proof of patience. Turn sleep away-time into a verifiable garden — SHA-256 Night Blocks, RFC-6962 Merkle proofs, Charter-supply math. No tokens, no wallets. Just your time, made beautiful.
What it is
A local-first garden game. You tuck your phone in → time passes (≥30 min, ≤10h/day) → wake → your Night Block mints (SHA-256 hash) → claim → Dawn settles → Lantern verifies → water → bloom.
Every block is proof. The rules are public math, not magic numbers. Your garden is yours — no server trust needed to verify it.
AI inside (on-device, never touches consensus)
-
llama3.2:3b→ quest text + garden journal -
moondream→ photo judging (strict JSON, graceful fallbacks) - All local — photos never leave your device.
What you see
- Garden — living specimen, growth stage, sunlight, bloom
- Bedtime / Tuck In — start the night, see your window
- Morning Dew — landing + protocol intro
…
How I Built It
The code is in three layers: a pure TypeScript engine (hashing, Merkle proofs, block rules), a local adapter that runs the game in the browser, and a React PWA on top. The UI never touches the rules directly.
Done in v1: the Tuck In → Dew → Claim → Dawn → Lantern loop, the rules (under 30 minutes is dust, 10-hour cap per block), a growing plant, and golden tests for the core math.
Next: Charter and Census math, the Sky Draw, wilt and revive, and a World Sim.
Open-source AI (in progress): small local models through Ollama. llama3.2:3b would write quests and journal lines, and moondream would judge a photo quest, which is the real “touch grass” part. They stay out of the core rules, with template fallbacks if the model is off.
Why Does Open Innovation Matter?
The pitch is "You don’t have to trust me; you can check.” That only works if the rules are open, so anyone can re-run the hash and verify a proof without trusting a server.
It also matters for the AI. A photo from your day is personal. Open-weight models running locally mean the image goes to localhost and nowhere else, which a closed API couldn’t promise.
Top comments (1)
tr.ee/dev-to