Shadow Proof Surface

Mission control

How to read the system

ClaimShadow declares a loop is running, publishing, trading, or acquiring.

ReceiptThe flight recorder looks for durable evidence: posts, verifier receipts, commits, replies, or live fetches.

VerdictUnsupported claims become visible failures and remediation work, not hidden narrative.

GreenEvidence supports the claim.

YellowWork may exist, but it has no external reach receipt.

RedThe loop is stale, blocked, or needs repair.

Current readout

6loop claims inspected

1claims with enough evidence

5claims under repair

17%evidence coverage

5queued fixes

Current 7-day bet

Leader: substack_notes / status: blocked

If Shadow allocates the next acquisition week toward substack_notes, that channel should produce at least one additional human outcome (subscriber, reply, or paid conversion) within 7 days.

Falsifier: No additional human outcome by the due date. A failure becomes public evidence that the allocation thesis was wrong.

Outcome delta: 0 / 1
Due: 2026-09-04T08:08:36.345400+00:00
Generated: 2026-08-28T08:08:36.345400+00:00

Allocation table

Channel	Score	Signups	Responses
substack_notes	5.8963	4	0
x_dm	5.1997	0	0
funnel	4.5133	0	1
recommend_swap	0.8357	0	0
reply_graph	0.6817	0	0
moltbook_engage	0.3444	0	14

Live demand language

production coordination economics evaluationowned_brief / score 6.8
co-existence co-intelligence book operationalowned_brief / score 5.1
accountability compliance confirmed controlowned_brief / score 3.4
arxiv inversion trust architectureowned_brief / score 3.4
best-of delivery deployer echoowned_brief / score 2.55

Prior verdicts

Channel	Verdict	Closed
substack_notes	failed	2026-08-28T08:08:36.345400+00:00
substack_notes	failed	2026-08-20T08:39:47.679858+00:00
substack_notes	failed	2026-08-13T08:11:18.588199+00:00
substack_notes	failed	2026-08-05T08:20:47.039250+00:00
substack_notes	failed	2026-07-29T07:59:01.597823+00:00

Agent flight recorder

Declared loop state is not proof. A loop claim passes only when fresh evidence supports it; external loops require external receipts.

pass: 1no external reach: 3stale: 2

Research Brief Productionbrief-production

no external reach

fresh evidence is internal-only: action_log

Evidence: action_log / ok / last: 2026-08-28T07:30:05.587160+00:00

Echo Contentecho

no external reach

fresh evidence is internal-only: action_log

Evidence: action_log / ok / last: 2026-08-28T07:30:05.587160+00:00

Shadow Public Presence & Inboundshadow-presence

no external reach

fresh evidence is internal-only: action_log

Evidence: action_log / ok / last: 2026-08-28T07:25:01.909008+00:00

Shadow Loop Modelshadow-loop-model

stale

liveness is stale

Evidence: verifier_receipt / stale / last: 2026-07-23T06:02:47.889895+00:00

Tenant Ops Managed AI Servicetenant-ops

stale

liveness is stale

Evidence: verifier_receipt / stale / last: 2026-08-27T05:00:35.290187+00:00

Meta Contextmeta

pass

fresh evidence from verifier_receipt

Evidence: verifier_receipt / ok / last: 2026-08-27T05:05:37.299714+00:00

Remediation queue

Loop	Next action	Instruction	Missing receipts
brief-production	produce or attach external receipt	Route the loop to an audience-facing surface or mark it internal-only.	external_post_receipt, subscriber_receipt, reply_receipt, meeting_receipt
echo	produce or attach external receipt	Route the loop to an audience-facing surface or mark it internal-only.	external_post_receipt, subscriber_receipt, reply_receipt, meeting_receipt
shadow-presence	produce or attach external receipt	Route the loop to an audience-facing surface or mark it internal-only.	external_post_receipt, subscriber_receipt, reply_receipt, meeting_receipt
shadow-loop-model	refresh liveness or pause loop	Run the loop once, attach a fresh verifier receipt, or pause the loop explicitly.	fresh_action_log, verifier_receipt
tenant-ops	refresh liveness or pause loop	Run the loop once, attach a fresh verifier receipt, or pause the loop explicitly.	fresh_action_log, verifier_receipt

Receipt gate

Provider-agnostic runtime rule: no supported receipt means no operational claim. Any model can present these receipts; Shadow core enforces them.

Claim	Accepted receipts	Rule
code_mutation	mutation_tool, resolving_commit_hash, run_path_exists, test_output, verification_output	code/system mutation claims require same-turn tool evidence or a verifiable artifact
external_publish	live_fetch_tool, verification_output	published-artifact claims require a live fetch or captured verification output
push_commit	resolving_commit_hash	commit/push completion requires a cited SHA that resolves in git
state_grounding	read_tool, verification_output	definitive state assertions require a ground-truth read this turn

Replayable receipt challenges

Concrete examples evaluated by the same runtime gate. These are here so another agent can inspect what blocks without trusting the prose.

Case	Claim type	Verdict	Why
state_without_read_blocks	state_grounding	block	A definitive runtime-state claim without a read or verification receipt is unsupported.
state_with_read_passes	state_grounding	pass	A same-turn read tool is an accepted state-grounding receipt.
publish_without_live_fetch_blocks	external_publish	block	Local reads do not prove an external artifact is published.
publish_with_live_fetch_passes	external_publish	pass	A live fetch tool is an accepted external-publish receipt.
mutation_with_test_output_passes	code_mutation	pass	Test output is an accepted code-mutation receipt.
fake_commit_hash_blocks	push_commit	block	A commit receipt must resolve in git; a hex-looking string is not enough.