CyberAlchemy

The CyberAlchemy Method

Make the reasoning reusable, not just clever.

The method behind Arcanum treats agent work as more than prompt execution. A useful capability is grown — through intent, research, discovery, tension, evidence, and reflection — into a governed artifact. The goal is not an agent that sounds smart for a moment. It is reasoning that is reusable, inspectable, and improvable.

At a glance

Five things stay visible the whole way through.

CyberAlchemy is for work where the answer should become a reusable artifact, not just a momentary response. These five anchors keep the middle of the work honest.

objective

What are we solving?

The problem the work exists to address.

output artifact

What should exist when this is done?

The concrete thing left behind — a spec, design, plan, decision record.

discovery

What must we learn first?

What has to be found before the artifact can responsibly close.

tension

What could make it brittle?

Where it might be oversized, misleading, premature, or unsafe.

route

Who owns the next step?

The lifecycle authority that takes it from here.

guiding stars

Not chains

Objective and artifact orient the work. When discovery proves a different output would solve the objective better, you rename them — and continue.

The loop

Orient, discover, shape, stabilize, evolve.

The loop is recursive: later discovery, tension, or validation can send the work back to an earlier step. Closure is contextual, not final.

1steps 1–3

Orient

Name the seed, bound the context, and state the current target result.

2step 4

Discover

Find the evidence, gaps, unknowns, and alternative frames that matter.

3steps 5–8

Shape

Choose the route, draft the artifact, test it with tension, revise toward closure.

4steps 9–12

Stabilize

Make it navigable, preserve openness, record the trace, route ownership.

5step 13

Evolve

Reflect after use, so repeated work improves the method itself.

The synthesis pattern

Scattered intent becomes a structure that survives handoff.

Synthesis is not summarization. It is a transformation: from a vague start into an object a future human or agent can pick up and continue.

seed intent
  → bounded context
  → objective and output artifact
  → research and discovery
  → local vocabulary
  → candidate structure
  → tension and critique
  → revised artifact
  → validation evidence
  → lifecycle route

It succeeds when a stranger can see what was asked, what was produced, why this shape was chosen, what was rejected, what is still open, and what happens next.

Who does what

Agents are role-bearing collaborators, not passive executors.

  • Human — provides intent, taste, constraints, approval, and the blocker decisions.
  • Primary agent — owns the run, integrates the evidence, edits the artifacts, keeps the final account.
  • Researcher — finds source evidence, outside patterns, missing context, alternative frames.
  • Proposer — builds candidate structures, options, reductions, plans.
  • Balancer — challenges scope, closure, hidden complexity, brittle minimalism, unsupported assumptions.
  • Observer — records telemetry and notices when reflection is due.
  • Lifecycle authority — owns promotion, validation, implementation, or maintenance for the target.

The human gate stays: agents prepare options and trade-offs, but blocker decisions, scope approval, and contract changes need human validation.

Governing principles

The values the method runs on.

intent before machinery

Don't build before the intent is clear

No system before the intent is clear enough to carry through a lifecycle artifact.

discovery feeds synthesis

Don't just rearrange the context in front of you

Research, inspect, compare, and interview until the synthesis can stand on its own.

artifact over vibes

Leave something reviewable

A good run leaves a spec, design, glossary, decision record, or validation result — not just a clever reply.

ergonomics is governance

Clarity is structural

Work easy to understand, navigate, and resume is how the system avoids misuse, duplicated reasoning, and stalled handoffs.

smallest coherent unit

Small, but still meaningful

Find the smallest unit that still has responsibility, inputs, outputs, and a way to recompose upward.

tension is productive

Critique is not failure

Tension reveals where the structure needs sharpening, routing, or a human decision.

traceability is memory

Remember why

Decisions without a trace become future ambiguity. The trace is part of the work, not residue.

open-ended, not overbuilt

Avoid both extremes

No premature complexity, and no brittle minimalism. Preserve the smallest useful extension boundary.

In one breath

The method, stated plainly.

governed synthesis through artifacts,
guided by objective and output,
fed by research and discovery,
shaped by tension,
made navigable for humans and agents,
bounded by lifecycle ownership,
made durable through trace and observation.

This is the method underneath Craft, the audits, and the deterministic engine. Building with it? Work with me → LinkedIn. Part of the CyberAlchemy system.