Disclosure · Issue No. 1

How the AIactually works here.

Every Heist Cycle release is AI-generated end to end — scripts, voices, visuals, distribution. This page states exactly how, where the human editorial sign-off and the Compliance/IP gate sit, what the originality signals are, and how published content is labeled. Plain language, no hedging, no hidden toggle.

AI-generated visuals and voice
Altered visuals
The 13-agent cycle

Thirteen specialized agents — disclosed one by one.

Each agent has a single brief. Read it as the production-desk view of who writes the script, who renders the frame, who posts the release, and who signs off before it goes live. The compliance and rights desks are the only seats with veto power.

Scripting & facts
  1. Trend scoutAI scouting the launch-graph and what to claim early.

    Reads the trending search graph and confirms which story angles will hold attention through the launch window.

  2. EditorialAI selecting the day's story across all five pillars.

    Picks the entry-point story for each surface. Suggestions only — never touches the script directly.

  3. ScriptingAI writing the first draft against the editorial brief.

    Writes the on-narrative first draft. The Compliance/IP gate rejects any draft that drifts into templated phrasing.

  4. Fact-checkAI sourcing and matching claims before sign-off.

    Pairs every claim in the script against the source database. Unsourced sentences are flagged before the Compliance/IP gate.

Production
  1. Visual productionAI rendering the original neon crime-fiction frame.

    Generates every frame inside the in-house visual world. Zero third-party footage. Zero Rockstar footage.

  2. VoiceAI voicing the hosts with pillar-anchored continuity.

    Style-anchored AI hosts — each pillar has its own voice profile, no cross-pillar cloning.

  3. Thumbnail & packagingAI producing the library face for the pillar.

    Each pillar keeps its own packaging template — the cycle does not output one homogenised look.

  4. DistributionAI posting to YouTube, podcast, newsletter, TikTok, and premium ad inventory.

    Posts to the licensed surface set. Disclosure badges are attached automatically at this stage.

Distribution & oversight
  1. CommunityAI moderating the theory threads, polls, and format nights.

    Surfaces scored-on-originality reader submissions; does not invent audience submissions on the audience's behalf.

  2. AnalyticsAI reading throughput into the next editorial decision.

    Read-throughs and CTR feeds don't change published content retroactively — they shape the next cycle's briefs.

  3. MonetizationAI staging affiliates, YPP, sponsors, the Black Friday wave, and licensing.

    Stages revenue against the planned sequence; does not pick unannounced sponsors or insert undisclosed paid mentions.

  4. Compliance/IPAI gating every release on disclosure, originality, and the YouTube policy.

    Auto-rejects any release missing the disclosure badge, the originality-bar proof, or the rights-desk clearance.

  5. Rights deskAI clearing every frame against the third-party IP table — final human-approved sign-off lives here.

    The only seat that returns a frame to production for re-render. Has explicit human final sign-off on each calendar slot.

Where the gates sit

Two vetoes, in series. The calendar is not one.

Editorial sign-off and the Compliance/IP gate run sequentially before any release publishes. If either fails, the release is held; neither can be bypassed by a deadline.

Compliance / IP gate

Where the structural risk is cleared.

Auto-runs in series before publish. The Compliance/IP agent checks disclosure, originality, and rights clearance. The rights desk returns the frame to production if any check fails — never publishes a flagged item.

  1. Distinct editorial angle — not a re-cut of an earlier release.
  2. Non-templated script — the Compliance/IP agent reads for phrasing drift.
  3. Non-templated voice — the voice line isn't shared verbatim across releases.
  4. Pairwise repetitiveness score under 0.85 against the last 14 days of releases.
Editorial desk

Where a human signs off.

The rights desk has a human final sign-off on every calendar slot before the release ships. Disclosure is on by default, originality proven by the gate, and the publish decision is the gate — never the deadline.

The originality signals

Four gates the Compliance/IP desk enforces before publish.

Every release has to clear the originality bar before it ships. If a signal fails, the rights desk holds the release — it never overrides the gate, even on a tight calendar slot.

01

Distinct editorial angle

Checks: The opening frame and headline aren't a re-cut of any prior release inside the 14-day window.

If it fails: The release is held; the editorial agent re-angles the script before re-submit.

02

Non-templated script

Checks: The script doesn't drift into templated phrasing — the Compliance/IP agent spots boilerplate and re-form sentences.

If it fails: The Compliance/IP gate requests a re-written paragraph before any visual production starts.

03

Non-templated voice

Checks: The voice line is not a verbatim match against the previous eleven days of releases.

If it fails: Voice production is sent back for a fresh take. The visual post-production also stalls.

04

Pairwise repetitiveness < 0.85

Checks: Total pairwise similarity score between this release and the previous fourteen days is under 0.85.

If it fails: The release is queued for human editorial review — it does not auto-publish.

How published content is labeled

Two disclosure badges. Both on by default.

Every published release carries one of two disclosure badges, attached automatically by the distribution agent. The badge shape maps to YouTube's 2025 inauthenticity policy and stays on the release page for the lifetime of the public link.

FULLY_SYNTHETICAI-generated visuals and voice

AI-generated visuals and voice

The default badge. Both the on-screen visuals and the host voice are AI-generated end to end. Applied to the majority of releases.

ALTERED_ONLYAltered visuals

Altered visuals

Applied when a real, sourced clip is the editorial anchor and AI generation is layered on top. The original sourcing is shown in the show notes.

Plain-language Q&A

Questions younger viewers ask first.

Eight questions the desk gets the most, written in the same plain language we use when we explain the system in person.