How to Use AI as a Lost Innovator

A practical, step-by-step method for doing credible intellectual work with AI — and earning a hearing for it — without an institutional credential. Written for "Lost Innovators" and useful to any author, it walks through eight steps: set a base model and add independent specialist models; draft; have separate models adversarially attack the work; verify every source against primary materials; integrate honestly while following disconfirming evidence; keep producible transcripts; disclose AI use via the AIast notation; and publish openly for a dated DOI. It documents the verification failure modes that matter most — real sources misrepresented, fabricated "example" citations, tasks blurring mid-conversation — and frames the discipline as a way to avoid fooling oneself as much as to earn a reader's trust. This guide is the practical companion to the AIast disclosure standard ("Earned Trust") and to "The Lost Innovators Hypothesis." Changes in version 1.2: the cover and byline are brought into conformance with the author's other published work and the Earned Trust standard v1.6.3 (byline: William Stafford, ADN, LI-AIast3; version, deposit DOI, and licence stated on the cover); Step 1 adds model-selection guidance (architectural diversity across providers; using the strongest reasoning tier available for critique and verification; recording service, tier, and version); Step 7 is aligned with v1.6.3 (the byline numeral counts user-facing AI services in substantive roles; the scoped negative declaration AIast0; version-recording guidance); the disclosure is updated accordingly. No changes to the method's eight steps.