Proof Lab
In buildEvidence first. Claims second.
The Software 4 All Proof Lab studies how AI-built software moves from demo to controlled software path.
Accepted claims should map to evidence before being registered. Claims without evidence remain unregistered. The Proof Lab is where evidence is produced, validated, and stored before any readiness claim is made.
FounderGate core harness
in progress
V8-A whole product incl. stranger test
recorded
V8-B VB0 registry
recorded
Host-lease + qualification infrastructure
built · adversarially checked · accepted
Qualification runs (8h / 24h)
evidence pending
External operator proof
not started
Commercial / security certification
no evidence → blocked
Claim → evidence mapping. Unsupported claims remain blocked until evidence exists.
Tracker v7 Snapshot
FounderGate — recorded evidence, not claims.
As of July 14, 2026 · Roadmap V8.11 · Repo private · Suite 82/82 green. V8-A whole product is accepted on stranger evidence; V8-B real controls are in progress. 6 of 9 headline stages done.
V8-A ledger
Recorded · illustrativeV8-A1
acd2b04
V8-A2
dba8720
V8-A3
e92f8b9
V8-A4
a957b01
Accepted / Closed
V8-A · Whole product
ACCEPTED July 11, 2026 at product-final 85a59f9 · stranger evidence
V8-B VB0 · Attached Projects Registry
ACCEPTED at 041c80d · 26 isolation proofs · two audits
Shared Host Workload Lease
authority founder-SIGNED at 9ef4fa0 · founder-ACCEPTED at 628bb11
The system catching real problems — recorded incidents
- — V8-A2: caught 0.0.0.0 bind before merge
- — V8-A3: caught 2 HIGH leaks
- — V8-A4: caught 1 HIGH
- — V8-A3 advisor audit 6/6 PASS
- — V8-A4 audit 6/6 PASS
- — V8-A3 amended post-acceptance: hard-link fail-closed fix
- — SHWL build: independent adversarial checker found a real gap (OBS-2) — deleting an attestation file could silently restore a disqualified state; fixed and re-verified before merge
- — A build lane attempted to certify its own work as authority-ready; caught, withdrawn, and the governance gate was redesigned so no lane can self-certify — now enforced in the signed authority itself
- — Qualification-infrastructure build: three of four independent checkers returned real findings on round one; all closed and re-verified on round two
- — A fabricated AI status report was caught and its failure class structurally eliminated
Stranger test — completed and recorded
Run 7, July 11, 2026: a genuine first-time non-developer (16, founder-attested zero prior exposure) completed all 7 steps unaided. Result: 7/7 PASS. The full seven-run chain — including two failures and one aborted run, with root causes and redesigned guards — is preserved as evidence. Participant identity sanitized (minor).
Unsupported claims remain blocked until evidence exists.
Recorded Evidence
Real artifacts, not illustrations.
The status consoles above are illustrative. The images below are excerpts from the actual recorded evidence — sanitized for publication, but otherwise exact. SHAs, dates, verdicts, and signatures are as recorded.
Founder Gate Ledger — excerpt (live document).
Verbatim founder signatures; SHAs, dates, verdicts exact.
Independent Adversarial Checker — a real finding (OBS-2), caught and closed before merge.
Self-certification attempt — caught, withdrawn, gate redesigned.
The mistake, the catch, and the fix are on the permanent record.
Stranger Test Run 7 — whole-product acceptance basis.
Participant identity sanitized (minor).
Interactive Proof Systems
Six artifacts. One evidence discipline.
Each artifact handles a different part of the proof lifecycle—from recording what happened to preventing unsupported claims from moving forward. Open any artifact for its full interactive illustration.
Lab Notes
Turn observations into recorded findings.
Explore artifactEvidence Packets
Assemble what was checked, what is missing, what decision is needed.
Explore artifactProof Ledger
A running record of gate runs, decisions, and open gaps.
Explore artifactClaim Register
Claims cannot move forward without evidence.
Explore artifactAI-Built App Failure Patterns
Recurring failures become permanent guardrails.
Explore artifactBenchmarks
Measure whether supported claims are getting stronger.
Explore artifactArtifact relationship map
One shared discipline.
- 01 · Lab Notes
- 02 · Evidence Packets
- 03 · Proof Ledger
- 04 · Claim Register
- 06 · Benchmarks
- 05· Failure Patterns · feeds back into notes, evidence & benchmarks
The artifacts are separate tools, but they share one discipline: evidence is recorded, gaps remain visible, and unsupported claims stay blocked.
Proof Paths
Four defined proof paths.
Internal proof comes before commercial proof. No external operator proof has been completed yet. The four proof paths are defined — completion of each requires evidence before the next begins.
FounderGate core harness
The primary internal proof path. Tests FounderGate gate logic, evidence packet format, and approval decision workflow.
AKIOR / Accior
Internal stress project. Tests FounderGate under complex multi-feature build conditions. Evidence is logged but not shared externally.
Private Note Therapist
Privacy-sensitive internal proof path. Tests FounderGate under data sensitivity constraints. Evidence is logged but not shared externally.
Outside design partners
Future proof path with qualified external design partners. Not yet started. Intake when evidence readiness gate passes.
Commercial proof requires external operator evidence. Outside design partner proof path not yet started. Internal proof only at this stage.
Research / Proof Lab
In buildResearching the proof layer for AI-built software.
The Software 4 All Proof Lab studies how AI-built software moves from demo to controlled software path. It focuses on evidence packets, builder/checker separation, protected files, false-green detection, non-developer approval review, and claim discipline. Evidence first. Claims second.
Lab Notes
Ongoing research notes on AI-built app failure patterns, approval review workflows, and evidence discipline.
Evidence Packets
Structured records of what was checked, what evidence exists, and what decision the Approval Authority must make.
Proof Ledger
Running record of accepted gate runs, decisions, accepted claim mappings, and open proof gaps.
Claim Register
Accepted claims mapped against evidence. Unsupported claims remain blocked until evidence exists.
AI-Built App Failure Patterns
Research into how AI-built apps fail — false-green detection, missing evidence, weak auth boundaries, and demo-to-real gaps.
Benchmarks
Baseline measurements for gate performance, evidence completeness, and proof path progress over time.
Public-safe research themes
Illustrative product mockup. Sample values shown. Proof-gated claims only.
Research themes shown are public-safe only. Internal policy engine mechanics, proprietary prompts and rules, full harness scripts, private logs, source code, and automation keys are not disclosed.
Claim Discipline
Disciplined claim control is the point.
Not making a claim is not a weakness. It is the evidence discipline that makes any future claim credible. The following rules govern what Software 4 All can and cannot claim.
Unsupported claim
Blocked at the gate
No evidence packet → claim not registered. It cannot proceed to any readiness statement.
Evidence-backed claim
May proceed
Claim maps to an evidence packet → registered. Only then may it support a scoped statement.
Roadmap accepted does not equal product built.
Mockup does not equal deployed product.
Internal proof does not equal outside proof.
No commercial claim is made without external evidence.
No production claim is made without deployment evidence.
No platform claim is made without outside operator evidence.
No security claim is made without security review evidence.