CAUM Live Evidence

Three receipts from real CAUM Live runs.

The demo is useful when it produces evidence. These receipts show the CAUM boundary in public: structural telemetry only, hash-chained evidence, no raw content, no semantic truth claim, and no agent blocking.

v1.4 public gateReceipts were generated by CAUM Live production.
Observe-onlyEvery receipt returns allowed_to_block=false.
Zero-semanticProduct boundary returns raw_content_stored=false.
Hash chainEach event carries event_hash, previous_chain_head, and chain_head.

What the three receipts show.

They show CAUM observing structural differences between clean progress and repeated work. They do not prove semantic truth, realized savings, or future failure prediction.

T1 observation

Healthy progress

Varied steps, low loop pressure, no exact-cycle coverage, no live alert.

  • event_typeobservation
  • tokens2,562
  • cost at risk$0.00025951
  • chain headcc1e5de803c4
Open JSON receipt
hard alert

Retry loop

Repeated state transitions created exact-cycle coverage and live structural loop evidence.

  • event_typestructural_loop_continues
  • tokens26,880
  • cost at risk$0.05518568
  • chain headb18c870eccbf
Open JSON receipt
T1 observation

Guarded after pressure

A short retry probe followed by replan progress kept the public signal in observation.

  • event_typeobservation
  • tokens3,498
  • cost at risk$0.00067558
  • chain head2a6bd5cc4e5f
Open JSON receipt

The sale is the recurring pattern.

In this evidence packet, the loop receipt has materially higher reviewable exposure than the guarded receipt. This is a unit contrast from controlled receipts, not a guaranteed savings claim.

retry loop hard alert
$0.0552

Cost at risk inside the loop receipt. Exact-cycle coverage is 1.00 and live_alert=true.

guarded observation
$0.0007

Cost at risk inside the guarded receipt. Exact-cycle coverage is 0.00 and live_alert=false.

CAUM can show reviewable structural exposure. Real customer savings require the customer's baseline, frequency, workflow economics, and operational decision process.

Inspect the public receipts.

These JSON files are intentionally boring in the right way: structural fields, public signal, tier, cost, hash-chain evidence, and product boundary.

Download manifest
Loading receipt...

Claim boundary.

This page is allowed to show attestation because the attestation is structural. It is not a semantic verdict.

AllowedStructural loop evidence, exact cycles, cost/token exposure, hash chain.
Not allowedHallucination detection, truth judgment, failure prediction, blocking claims.
Internal onlyBroad T4/T5 review buckets cannot be published as a waste rate.
Public signalUse public_signal.class, hard alerts, and direct observed exposure.

Now ask CAUM to watch one recurring workflow.

Start with a neutral trace or a recurring agent run. If the same structural pattern repeats, move that workflow to CAUM Live.