A.D.A.M.

Adaptive Depth and Mode

A protocol for controlling depth, not just prompting it.

A.D.A.M. is a spec-first control layer for AI chat: rule-ordered routing, bounded audit, explicit depth transitions, and fail-closed behavior when the host cannot respect the contract.

Open the repository Read announcements
  • Manual override before semantic drift
  • Structural kernel as SSOT
  • Bounded epistemic output under pressure
A.D.A.M. system frame visualization
Too shallow

When the assistant compresses before the problem is even understood.

Too deep too early

When depth becomes noise instead of precision.

Too inconsistent

When the same prompt style produces different behavior from one turn to the next.

Why it exists

Stable depth is a routing problem.

A.D.A.M. separates lexical routing, structural inference, and publish-boundary validation. That is what makes it feel less improvisational and more like a protocol artifact.

Manual override

LOW, MID, and DEEP are explicit control surfaces, not suggestions hidden inside prose.

DEEP explain this briefly

Structural kernel

Depth candidates come from shape and constraints, not vague intent guessing.

2+ options + criteria -> POSSIBLE DEEP

Bounded audit

Audit is visible only when needed, rigid when active, and never allowed to balloon.

AUDIT / A: action / R: risk / V: verification

Fail-closed behavior

If the host cannot satisfy the contract, the protocol fails in visible, rule-bound ways.

ADAM_UNSUPPORTED
A.D.A.M. routing field visualization

See it work

Not a moodboard. A behavioral surface.

Input

User:
Compare A) and B)
Budget: 100
Need:
- speed
- reliability
- low waste

Output

MODE: MID -> POSSIBLE DEEP
RX: len=...
Compact recommendation...
AUDIT
A: ...
R: ...
V: ...
Switch to DEEP? (yes/no)

Explicit depth

When depth matters, A.D.A.M. makes the transition visible.

Deep mode is not a vague "think harder" gesture. It is gated, surfaced, and bounded: a deliberate handoff from routing into expanded analysis, with audit still kept under contract.

  • DEEP is explicit or manually forced
  • Audit remains bounded even under expansion
  • Tracing stays readable instead of theatrical
A.D.A.M. deep access visualization

Deployment surfaces

Three variants. Three host realities.

Normal

The full protocol surface for hosts that can carry the complete contract cleanly.

UI-LITE

The same behavior with a simpler visible UI: tags and bounded audit, no decorative Unicode dependency.

Guest Card

A compatibility layer for weak or paste-only hosts. Useful, intentionally narrower, not full parity.

Built for real host behavior

Uploads can be inline, side-channel, polluted, or broken.

A.D.A.M. is designed around actual host friction: boot guards, explicit probes, degraded transports, strict command matching, and strict fallback behavior.

BOOT_GUARDS

Whitespace-only, spec echo, metadata-only upload.

CONTROL PATH

Commands and overrides resolved lexically before semantic interpretation.

OUTPUT CONTRACT

Final text validated at the publish boundary.

A.D.A.M. audit pressure visualization

Bounded audit

When uncertainty rises, visibility rises with it.

Audit in A.D.A.M. is not decorative introspection. It is a constrained surface for action, risk, and verification, kept short enough to stay operational under pressure. A = action, R = risk, V = verification.

  • Audit appears only when the contract says it should
  • Action, risk, and verification stay bounded and readable
  • Retrograde pressure does not silently inherit stale commitments