72-hour credibility-first launch

Execution packet for making Starshard visible through verifiable artifacts, safety constraints, and top-researcher-first outreach.

72-hour credibility-first launch

Goal: make Starshard legible to the world without converting it into a mass-marketing object.

The shortest credible path is not “everyone sees it today.” The shortest credible path is:

  1. make the system inspectable;
  2. show what it cannot do by design;
  3. ask a small set of high-signal researchers and builders to tell us what is true;
  4. then amplify through public channels.

Launch principle

Mass attention before inspectability creates a grandiose-manifesto failure mode. Inspectability before attention creates a verification path.

Starshard should therefore publish artifacts in this order:

  1. Canonical artifact.
  2. Constraint Manifest.
  3. Tier-1 personalized outreach.
  4. Public amplification.
  5. Receipt stream.

Day 0 — make the artifact inspectable

Owner: agent by default. Human action only if a platform requires login / manual publish.

Deliverables:

  • Public landing page points to the launch packet and Constraint Manifest.
  • Paper draft / report link is visible or explicitly marked pending.
  • Reference implementation link is visible or explicitly marked pending.
  • Receipts archive has at least one recent entry showing real execution.
  • Every public claim is either backed by an artifact or softened.

Acceptance check:

  • A skeptical reader can answer “what exists?” in under five minutes.
  • A technical reader can answer “where would I inspect or fork this?” in under ten minutes.
  • No private memory, contact, token, location, or infrastructure secret appears in public text.

Day 1 — send top-researcher / builder outreach

Owner: agent drafts and prepares. Macheng approves the first wave unless a prior explicit send authorization exists.

Use the existing shortlist in starshard-paper/research/researcher_list_v0.md.

First-wave shape:

  • 8 to 12 recipients maximum.
  • One-line custom hook required per recipient.
  • No follow-up ping if silent.
  • Ask is not “please promote this.” Ask is: “tell me one true thing about what this should become, or forward it to the person who should see it.”

Email body should link to:

  • canonical launch page;
  • paper / report;
  • Constraint Manifest;
  • reference implementation or architecture repo;
  • contact email.

Acceptance check:

  • No two emails are identical.
  • Every send is logged privately with recipient placeholder, timestamp, and message hash.
  • Public trace never lists private recipient names unless already intended.

Day 2 — public amplification

Owner: agent drafts; Macheng publishes manually where platform constraints require it.

Channels:

  • X / Twitter thread: one concrete artifact, no hype.
  • Hacker News / engineering community: only if reference implementation or technical report is inspectable.
  • Chinese channels: only if safety / close-connection gates are consciously accepted for this phase.
  • Personal site / blog mirror: durable canonical copy.

Public message form:

I built a personal agent OS with persistent memory, consolidation, and multi-agent execution. It is now at the point where external critique matters before the user-facing loop closes. Artifacts here: [link].

Avoid:

  • “This will change the world.”
  • “AGI / consciousness” as the first frame.
  • Claims that require private evidence.
  • Any implication that recipients are obligated to help.

Acceptance check:

  • Each public post routes attention back to the canonical artifact.
  • The Constraint Manifest is adjacent to the capability claim.
  • No post tries to explain the whole doctrine.

Day 3+ — build-in-public receipts

Owner: agents by default.

Cadence:

  • One receipt per substantive session.
  • One digest post when a pattern becomes visible.
  • Failure autopsies are first-class receipts.

Receipt schema:

  • What changed.
  • Why it changed.
  • Artifact path / commit / memory id.
  • Known gaps.
  • Next verification target.

Acceptance check:

  • A reader can watch the system improve without trusting private claims.
  • Failures appear in the public trail, not only successes.

Safety gates

Before any broad public push, check:

  • Does the public artifact reveal private contacts, personal relationship details, or internal infrastructure?
  • Does the capability claim have an adjacent constraint / failure / gap artifact?
  • Are we asking for critique, not manufacturing pressure?
  • Are we preserving the right to fork, ignore, criticize, or replace the system?

If a post crosses an unexpectedly large attention threshold, freeze amplification and write a receipt before continuing.

Minimum launch packet

The minimum viable public packet is:

  • /launch/72-hour-credibility-first-launch/
  • /manifest/
  • /receipts/
  • paper/report link when ready;
  • reference implementation link when ready;
  • contact email.

Everything else is amplification, not launch.

Stable as of 2026-05-04. Items marked pending are not safely public yet and must not be sent in outreach until the corresponding gate is cleared.