1. Start With Human-Readable Source
A Sley project begins with source that a person can review. The public point is simple: the readable projection matters because humans remain accountable for intent, approval, and audit.
2. Ask The Compiler For Structure
An agent should not begin by rewriting arbitrary text. It should ask the compiler for bounded structural context: where the relevant unit is, what it depends on, what calls it, which authority it declares, and what the current checker believes.
3. Read Diagnostics Before Editing
Compiler feedback is part of the language design. Sley is being built so diagnostics can carry stable identifiers, useful locations, and repairable intent without letting an agent bypass authority.
4. Plan A Narrow Change
The preferred path is not "patch the file and hope." The preferred path is a narrow planned change generated against the structural surface. That planned change can be previewed and rejected before source is touched.
5. Verify Authority
External authority stays explicit. The v0 runtime uses deterministic gate seeds for sensitive paths so examples can prove the control flow without touching a live database, network, shell, model provider, secret store, deployment target, or spend account.
6. Accept With Evidence
When a change is accepted, the private toolchain can leave provenance, verification, package, seal, and handoff artifacts. The durable idea is that accepted agent work should leave more than an unexplained text diff.
7. What The Open Tutorial Will Add Later
| After Source Release | Why It Is Held Back Now |
|---|---|
| Runnable starter programs | They reveal grammar and idioms before the language is ready to publish. |
| Command transcripts | They expose the private tool contract and release cadence. |
| Structural edit payloads | They reveal the operation layer that differentiates Sley from text-first tools. |
| Schema examples | They make it easier to reconstruct internal output contracts. |
| Fixture corpora | They disclose the conformance strategy and edge-case map. |
8. Safe Mental Model
Sley is a language for a world where agents write more code, but humans still own judgment. It tries to turn program change into a compiler-mediated negotiation: inspect structure, propose a legal move, prove authority, verify behavior, and leave evidence.