Prompt or Die

Prompt or Die is an open-source game platform for building deterministic, authoritative worlds where autonomous AI agents and human players operate under the same gameplay rules. The runtime is written in Rust, supports native and browser clients, and is being extended toward a full 2D, 2.5D, and 3D authoring stack.

Current release posture: v0.1.0-alpha.2, an early testing alpha. Expect breaking changes across runtime contracts, tooling, and benchmark/report surfaces while the platform hardens.

Start here

• Documentation Hub
• CLI Surface
• Reference Bootstrap
• Architecture Overview
• Benchmark Suite
• Rust SDK Boundary

Quick start

Discover the supported platform command surface from one entrypoint:

bun ./scripts/pod.ts list
bun ./scripts/pod.ts shell
bun ./scripts/pod.ts runtime server --dry-run
bun ./scripts/pod.ts web dev
bun ./scripts/pod.ts show pod-server --json

For attached terminal workflows, use bun ./scripts/pod.ts shell. For interactive machine sessions, use newline-delimited JSON request/event objects through bun ./scripts/pod.ts shell --agent.

printf '{"type":"builtin","requestId":"1","name":"context"}\n{"type":"builtin","requestId":"2","name":"exit"}\n' | bun ./scripts/pod.ts shell --agent

For an attached terminal workflow, use the interactive shell:

bun ./scripts/pod.ts shell
bun ./scripts/pod.ts shell --agent
bun ./scripts/pod.ts export events --format toon
bun ./scripts/pod.ts export world --format json

pod shell is the human-friendly attached terminal experience. pod shell --agent is the machine-friendly interactive shell mode for long-lived autonomous agent sessions. One-shot automation should keep using list/show/env/command/run --json. TOON is reserved for the large LLM-facing export surfaces under pod export world|events|multiverse --format toon.

Build and run the main surfaces

bun ./scripts/pod.ts workspace build
bun ./scripts/pod.ts runtime desktop
bun ./scripts/pod.ts runtime server

cd apps/pod-web
bun install
cd ../..

bun ./scripts/pod.ts web dev

If you want to stay inside one terminal and inspect supported commands as you go, use bun ./scripts/pod.ts shell. For machine-driven interactive sessions, use bun ./scripts/pod.ts shell --agent.

Run the main proof surfaces

bun ./scripts/pod.ts runtime headless -- --profile ci-smoke
bun ./scripts/pod.ts runtime headless -- --profile ci-smoke --dataset-output /tmp/pod-headless-dataset.json --topology-output /tmp/pod-headless-topology.json
bun ./scripts/pod.ts benchmark controller-parity
bun ./scripts/pod.ts benchmark toon-exports -- --fail-on-checks
bun ./scripts/benchmark_toon_exports.ts --profile extensive --output artifacts/toon-export-benchmark-extensive.json --html-output artifacts/toon-export-benchmark-extensive.html --markdown-output artifacts/toon-export-benchmark-extensive.md --charts-dir artifacts/toon-export-benchmark-extensive-charts --fail-on-checks
bun ./scripts/pod.ts benchmark topology-feed -- --topology-input /tmp/pod-headless-topology.json --fail-on-checks
bun ./scripts/pod.ts workspace test
bun ./scripts/pod.ts workspace check

Publish the weekly shard-target snapshot

bun ./scripts/pod.ts history run-shard-target-snapshot -- --label 2026-W11

For retained history, snapshot comparison, and benchmark interpretation, see Benchmark Suite and Benchmark Snapshot History.

Stage an asset

bun ./scripts/pod.ts assets stage-import -- --output-root artifacts/staged-assets path/to/asset.glb

For bundle specs, --materialize-runtime, KTX2/meshopt variants, runtime selection rules, and browser asset verification, see Asset Pipeline.

What exists today

• Deterministic ECS runtime in pod-core with one shared agent pipeline: Observe -> Decide -> Validate -> Execute -> Broadcast
• Native and browser rendering surfaces in pod-render
• A real browser-side Three.js client in apps/pod-web
• Headless multi-world tournament and evaluation entrypoint in apps/pod-headless
• Scene, prefab, save/load, and state-stack authoring in pod-scene
• Direct-connect networking plus SpacetimeDB integration in pod-net and pod-stdb
• Asset processing, animation, scripting, spatial, and physics support across the workspace

Workspace map

crates/
  pod-core       Deterministic ECS world, tick loop, agent contract, actions, observations, events
  pod-render     Native wgpu renderer and browser bridge
  pod-scene      Scenes, prefabs, save/load, typed bindings, streaming
  pod-net        QUIC/WebSocket transport and SpacetimeDB-aware clients
  pod-stdb       SpacetimeDB tables, reducers, events, and client wrapper
  pod-agents     LLM, neural, scripted, and hybrid agent implementations
  pod-editor     Visual editor shell and authoring panels
  pod-assets     Asset import, processing, caching, and procedural generation
  pod-animation  Keyframes, tweening, blending, and state machines
  pod-physics    Physics integration
  pod-spatial    Pathfinding, raycasts, and spatial queries
  pod-scripting  Lua scripting API and sandbox
apps/
  pod-desktop    Desktop runtime and local simulation entry point
  pod-headless   Headless multi-world tournament and evaluation runner
  pod-web        Browser-side Three.js/WebGPU client and bridge demo
  pod-server     Dedicated authoritative server
specs/
  Product and subsystem requirements
docs/
  Public architecture, workflow, benchmark, and release guides

Documentation

• Documentation Hub
• CLI Surface
• Architecture Overview
• Plugin Model
• Agent Integration Contract
• Agent Runtime Audit
• Multi-World Agent Topology
• Asset Pipeline
• Reference Bootstrap
• Benchmark Suite
• Platform Stabilization
• Rust SDK Boundary
• Benchmark Snapshot History
• Competitive Matrix
• Moat Gates
• Bootstrap Showcase Research

Planning route

• Active execution checklist: IMPLEMENTATION_PHASES.md
• Historical implementation log: IMPLEMENTATION_PLAN.md
• Current session tracker: SESSION.md
• Delivery recap: progress.md

Current status

The repo already has the deterministic core, browser client, headless multi-world runner, shared topology contract, controller parity harness, and weekly shard-target benchmark workflow in place. The next major layers are public platform hardening, import/shipping polish, and a formal plugin/app lifecycle. The current hardening contract for benchmark tiers, planning-doc ownership, shipping profiles, authz policy scope, and SDK boundaries lives in docs/platform-stabilization.md. The repo-owned Rust SDK hookup boundary now lives in docs/rust-sdk-boundary.md.

Release history

• v0.1.0-alpha.2
• v0.1.0-alpha.1
• Migration: Wave 1 reconnect protocol + StDB connection mode