Turtle Lab
The public edge of TurtleBlock AI: build notes, learner-approved projects, WorldSpec artifacts, experiments, reflections, and things we learned by making them.
A Turtle conversation is private by default. Turtle Lab is not a raw Discord feed. A learner chooses what becomes public: a project title, a WorldSpec snapshot, selected build outputs, selected dialogue, and reflection.
Field-test sparks
Before public learner publishing is switched on, these are curated examples from the project author's own Turtle tests. They show the kind of meaning Turtle needs to preserve instead of collapsing everything into a few keywords.
“I want to make an abandoned village that has clear structures that are in disarray but were once nice, and then in the corner somewhere of the village a little shining light of hope of rebirth.”
The interesting problem is not merely “village.” Turtle needs to retain history, decay, spatial contrast, memory, hope, symbolism, and the relationship between ruin and renewal.
“Let's make a river that has glowing deep depths and a fishing pier with a hamburger stand.”
A deliberately odd combination is useful: depth, light, water, commerce, fishing, humor, adjacency, atmosphere, and possible ecological consequences all coexist.
What will appear here later
WorldSpec revisions, screenshots or builds, reflection, selected Turtle exchange, and links to remixes or descendants.
The public artifact should show thinking and making without publishing the whole private conversation by default.
Discord Turtle Lab
The preferred experience is for /turtle idea:... to start a persistent Turtle Lab session and, next, a dedicated Discord thread. The learner should then talk normally with Turtle while the same WorldSpec evolves underneath the conversation.
Persistent WorldSpecs
Each session receives a session ID and WorldSpec ID. Turns and WorldSpec revisions are stored separately so the learner's original language, Turtle's interpretations, and subsequent revisions remain distinguishable.
LLM conversation + RAG + security
Turtle's conversational layer uses the Turtle Charter as trusted instruction while learner language, retrieved material, Discord history, Minecraft text, and prior model output remain untrusted project data.
Activate D1 Turtle Lab persistence in production, then add Discord thread continuation so ordinary messages revise the same WorldSpec. After that, wire the opt-in publication projection that feeds this gallery.