CyberAlchemy
An operating language for agentic creation
Turn ambitious ideas into named workflows, governed artifacts, validator-backed evidence, and reusable tools people can understand.
Three pillars
Different layers, different jobs. DomainSpec says what is meaningful. Arcanum says how the work is done. Saturn closes the loop with evidence. Read the full explanation →
A closed vocabulary of 25 meta-types and 29 typed relationships. From a typed model you derive tests, metrics, and contracts by rule, not by guess. A deterministic engine already does this — same docs in, byte-identical output. Read the paper →
Reusable agent capabilities, composed routes, and the development loop from a wild idea to a validated slice. The part you can open and read today. Open Arcanum →
Observe what was built, compare it to the DomainSpec baseline, detect drift, correct, verify, feed back to governance. The intended frame — not yet a shipped public surface. Explore Saturn →
The method
CyberAlchemy treats names, prompts, schemas, validations, and artifacts as working material. A good name is not decoration; it is a handle for a repeatable process. Read the full method →
Start with the messy thing you keep doing. Name it so it can be pointed at, improved, taught, and reused.
Use refine, invoke, guides, x-rays, work-packs, and task-sessions to make the idea inspectable before mutation.
Every serious output should leave receipts: sources, decisions, validators, screenshots, residues, and the next route.
Arcanum loop
Describe the idea, user, promise, constraints, unknowns, and ambitious future.
Extract the core shape, compare routes, expose gaps, and choose the next owner.
Author definition, design, layering, work-pack, and handoff artifacts.
Use decision-gate, dispatch-spec, x-ray, or another refine pass when uncertainty appears.
Run one bounded task-session with validation and synchronized evidence.
Preserve residue, telemetry, follow-up gaps, and the next loop.
Arcanum directly
The public Arcanum system carries the real material: sigils, guides, validators, research traces, work-packs, and examples that can be inspected, reused, and improved.
Evidence-backed imagination
The Whisper example shows how a broad writing idea becomes cores, candidate routes, hard gates, composition parts, validators, and task-session receipts. The same structure can shape a tool, product workflow, frontend validator, or research route.
Start here
The best first prompt is rich, unfinished, and honest about what is exciting, risky, unclear, and worth proving.
[$refine] I want to develop this idea: Describe the product, audience, promise, current workflow, missing pieces, constraints, risks, unresolved decisions, MVP proof, and ambitious future. Please refine this into definition, design, plan, gaps, blockers, and the next executable task.