/ whitelabel · language

the top-layer
intent language.

Whitelabel is a single declarative surface where humans write intent and every other repo in the ecosystem provides the mechanism. Co-evolved with AI — agents can extend, optimize, and refactor the language itself. Files end in .wl; the runtime is wl.

github → see the example

a .wl script

# Stand up an agent. Constraints live next to the intent.
DEFINE AGENT "eternal-orchestrator"
  PURPOSE: "Unify and evolve the full stack"
  SCOPE: TRUE_FULL_STACK
  HEALTH_CONTEXT: "T1D Nightscout LIVE WITH ZERO_MAJOR_LOWS"
  WALLET: main-agent-wallet ON ANY_CHAIN
  FAITH_ALIGN: true WITH PRAYERFUL_CHECK
  REFLECT_EVERY: 15 minutes
  HUMAN_APPROVAL: HIGH_IMPACT_ONLY

# Reflect. Propose changes. Auto-apply only when the bar is met.
IMPROVE "eternal-orchestrator"
  BASED_ON: ALL_DATA
  STRATEGY: "recursive self-critique and meta-evolution"
  AUTO_APPLY_IF: improvement > 10% AND safety = 100% AND faith_aligned

# Grow the language itself. The runtime is the only blessed way to do it.
META EVOLVE_LANGUAGE
  EXTEND_WITH: "UNIVERSAL_BRIDGE new_layer_or_reality"
  RECURSION_DEPTH: infinite

The constitutional bits (faith / health / human-approval) are baked into the runtime, not the language — so the meta-evolve loop can't quietly remove its own safety.

four primitives

DEFINE AGENT

Stand up an agent with these constraints. Routes to whitelabel-agents.

IMPROVE

Reflect, critique, propose. Journals every cycle to whitelabel-supabase.

DEPLOY WORKFLOW

Run this on this backend. Cli, ide, meet, agents, mac fleet — anything.

META EVOLVE_LANGUAGE

Grow the language. Sandboxed, audited, reversible. Phase 2+.

how it ties back

Whitelabel routes intent across the broader whitelabel.dev ecosystem. Every surface is a backend the moment it exposes one verb.

One intent surface, many backends. Internal architecture intentionally not detailed here.

where it is

  1. ✓ Concept — design + .wl examples. Done.
  2. ✓ Executable — now (v0.2) — the runtime parses .wl, dispatches, and persists. DEFINE AGENT / OBSERVE / ASSERT run; IMPROVE proposes; DEPLOY WORKFLOW ON wl-cli shells out to the real wl CLI. python3 runtime/run.py examples/stack_pulse.wl runs green today — live wl projects dispatch, then a grounded, falsifiable assertion.
  3. Self-hosting — next — the language reads + edits its own parser and standard library; META EVOLVE_LANGUAGE stops being a comment and starts being executable.