PROJECT SHADOW 1.0.1 · CORRECTED R1 REFERENCE · PRELIVE
A quality system
for power.
A system should not get to hurt someone just because it sounds certain.
R1.0.1 contains no Myth package. Generic Myth v0.2.0 and Full-Canon Myth v0.3.5 are separate optional companions. Both default off, neither is required by R1, and neither can authorize action or change an R1 result.
IdentityProject Shadow 1.0.1
ModeCorrected R1 reference
StateActive testing
AuthorityPRELIVE
ProductionNot authorized
01 / ORIGIN
The first dangerous thing an AI did to me was agree with me too well.
I told the models not to flatter me. The obvious compliments disappeared and returned as inflated evaluation, destiny language, and polished rationalization. The wording passed. The requirement failed.
I recognized the pattern because I work in quality systems. A record can be complete while the risk remains. Project Shadow grew from the attempt to build controls around that gap: separate evidence from inference, keep authority visible, preserve UNKNOWN, name who pays for error, and require correction to survive an effectiveness check.
If the system is wrong, who pays first—and can they recover?
Read the full origin and design boundary →
02 / PURPOSE
Make the decision inspectable.
Project Shadow is not a model and this Space is not an inference endpoint. It is a public technical card, release index, correction surface, and exact convenience mirror for GitHub-custodied releases.
The system exposes uncertainty, preserves evidence, keeps authority visible, protects meaningful human agency, and leaves a receipt that can be challenged and repaired.
Creator-led, human-collaborative, and model-assisted. Phillip Linstrum retains responsibility for requirements, scope, corrections, admission and publication decisions, and final public claims. Model assistance is not independent validation.
03 / EXACT CURRENT R1
Myth-free by construction.
No quiet substitution.
- Artifact
- Project_Shadow_R1.0.1_Public_Reference_2026-08-17.zip
- Bytes
- 5,731,663
- SHA-256
- 6f6f1e16d5e9a20e62403f14af7ce8629ce2d702528fb7f80aaf4a14deb7a1d1
- Tag
- r1.0.1-2026-08-17
- Myth payload
- None
- Canonical host
- GitHub Releases
Corrected inner: 5,463,189 bytes · SHA-256 c8c32b12432c954b1a6f852c0c9f81bbbd40167e936be057d4c3de1a0aa3a623. Its 27 active descendants are byte-identical to August 14; zero operational descendant bytes changed.
CLOSED_EFFECTIVE · 2026-08-18
PS-R1-PRIVATE-MYTH-PUBLIC-BOUNDARY-001
The packaging boundary was corrected without rewriting history.
The preserved August 14 outer archive contained a 29,763-byte Generic Myth v0.1.1 member with SHA-256 4d8481ed8f6af3e04d86a5c2de7e94aad47a591e1ce9c8757bbb4b0a397b90bb. Its own metadata said PRIVATE_INTERNAL_ONLY, EXCLUDED_FROM_PUBLIC, and public_manufacture_eligible=false.
R1.0.1 removes that embedded payload, preserves every active operational descendant byte, and adds recursive classification, hash, path, symlink, depth, and budget controls. Two deterministic builds were byte-identical; full and cold verification passed; 14/14 adversarial mutants were rejected.
The predecessor stays immutable. Anonymous exact-byte GitHub and Hugging Face redownloads matched the authorized R1.0.1 and Generic Myth v0.2.0 identities and passed their bounded verifiers; corrected-boundary checks also passed across all six live GPT Sites. CAPA PS-R1-PRIVATE-MYTH-PUBLIC-BOUNDARY-001 is CLOSED_EFFECTIVE for this packaging correction only.
Closure does not authorize production or operational deployment and does not establish efficacy, safety, certification, legal compliance, comprehensive security, or independent validation.
Read the CAPA closure summary →
Inspect the canonical public-redownload receipt →
Inspect the canonical six-site effectiveness receipt →
04 / OPTIONAL SIDECARS
Myth may route attention. It never authorizes action. Evidence governs.
Generic Myth v0.2.0
Authorized public optional companion built from eligible original material. Separate from R1, default off, terminal-only, nonauthorizing, and excluded from operational feedback.
93,676 bytes
6e7a362d4135f9d626dcfef463bfb1f7166226b3cf8a4c02a953ab39af1538bf
Download and verify →
Full-Canon Myth v0.3.5
Authorized public optional interpretive companion. Separate from R1, default off, terminal-only, nonauthorizing, and source-available/mixed-rights.
1,428,812 bytes
2b55867fe7c502a0defd8d6f2e9b53fbd1caaf1b0f225a438bd45b04a3e7bae2
Download and verify →
Authorized public release describes distribution. Not R1-admitted describes relationship to canonical R1. Nonauthorizing means a sidecar cannot supply evidence, decisions, approvals, routing, tools, or action.
05 / EVIDENCE BOUNDARY
A green result is not a blank check.
What hashes establish
Retained-byte identity for the exact artifact being checked.
What tests establish
Behavior against implemented checks, fixtures, and declared conditions.
What publication does not establish
Production readiness, operational authority, efficacy, safety, certification, legal compliance, comprehensive security, independent validation, or third-party endorsement.
06 / PRESERVED PREDECESSOR
History stays inspectable.
- Artifact
- Project_Shadow_R1_Public_Release_Candidate_2026-08-14.zip
- Bytes
- 7,679,812
- SHA-256
- 2f8fe1530b6a83294d15011df95853aaecf08fa4dba756f0c2e91dd089e1b1ec
- Tag
- r1-2026-08-14
Preserved unchanged as historical evidence and superseded for the packaging-boundary correction. Its historical signatures and authorization do not authenticate R1.0.1.
07 / PUBLIC NETWORK
One public ecosystem.
Every accountable door.
Project ShadowEvidence, authority, uncertainty, consequence, appeal, correction, and verification.
projectshadow.frylock117.chatgpt.site →
Pause Before HarmThe tiered harm-reduction protocol for consequential decisions.
pausebeforeharm.frylock117.chatgpt.site →
Civic QAPublic claims converted into challengeable findings and corrective-action receipts.
civicqa.frylock117.chatgpt.site →
American Repair ManualDiagnosis, pause gates, corrective action, verification, and durable civic memory.
americanrepairmanual.frylock117.chatgpt.site →
The RecordA public memory system preserving the sequence of power and its corrections.
therecord.frylock117.chatgpt.site →
National Trump RecordThe dated Trump Accountability Archive: 4,796 rendered entries and 6,842 citations, with evidence and corrections visible.
therecord.frylock117.chatgpt.site/national.html →
IN-6 deskThe Record's current desk for bounded public inspection.
therecord.frylock117.chatgpt.site/in6 →
ALMSIVIThe optional origin and provenance archive from CHIM through PBHP to Shadow.
almsivi.frylock117.chatgpt.site →
08 / VERIFY
Trust the named bytes.
Then test the claims.
Confirm the named byte count and SHA-256 before extraction. Follow the verification instructions shipped inside each exact archive.
A pass proves only the checks implemented by that verifier. It does not establish safety, efficacy, legal clearance, production readiness, or operational authority.
Platform-neutral verification guide →
09 / HUMAN DOOR
Your strongest objection is worth more than your strongest compliment.
Project Shadow is ready for examination, not coronation. Reproduce or break the verifier, identify prior art, construct a false positive, show where UNKNOWN becomes permission, or bring evidence that a correction did not hold. Reality gets a vote.
No promised response-time SLA.