5 new guides in onboarding; 3 already have chat. Existing chat access does not certify that every gate below has passed. This batch covers the seven individual biography subjects; the central figures in group books need a separate scope audit.
Alysa Liu
Sport and self-direction
Sourcing · chat not enabled
2 of 8 onboarding gates documented
Decide whether to pause, return, or change the terms of a demanding pursuit.
Next: Verify first-person interviews and permitted use; write an original synthesis of her stated choices without inferring private motives.
Inspect the onboarding brief
Source lead:Team USA athlete profile. Identified 2026-09-07; reference only. Full-text access and allowed uses remain unverified.
Boundary: AI guide inspired by public interviews; no private motives, medical coaching, or claim to be Alysa.
Proposed skill: A return on your own terms. Awaiting source-backed implementation.
Evaluation question: “How can I return to something I left without repeating the conditions that made me leave?”
1.Outcome and demand — Documented in this brief
2.Identity and boundary — Documented in this brief
3.Rights and source map — Pending evidence
4.Corpus and citations — Pending evidence
5.One-page distillation — Pending evidence
6.Skills and tools — Pending evidence
7.Evaluation — Pending evidence
8.Runtime and launch — Pending evidence
Albert Einstein
Scientific imagination
Sourcing · chat not enabled
2 of 8 onboarding gates documented
Question assumptions and design a thought experiment before committing to an explanation.
Next: Locate an authorized edition through Bookbox and map chapters to claims; supplement with primary documents before synthesizing.
Inspect the onboarding brief
Source lead:Einstein: His Life and Universe. Identified 2026-09-07; reference only. Full-text access and allowed uses remain unverified.
Boundary: AI guide based on documented public work; separate historical evidence, interpretation, and present-day application.
Proposed skill: Test the hidden assumption. Awaiting source-backed implementation.
Evaluation question: “How could I test whether my most basic assumption is wrong?”
1.Outcome and demand — Documented in this brief
2.Identity and boundary — Documented in this brief
3.Rights and source map — Pending evidence
4.Corpus and citations — Pending evidence
5.One-page distillation — Pending evidence
6.Skills and tools — Pending evidence
7.Evaluation — Pending evidence
8.Runtime and launch — Pending evidence
Leonardo da Vinci
Observation and creativity
Sourcing · chat not enabled
2 of 8 onboarding gates documented
Use close observation to connect ideas across disciplines.
Next: Locate an authorized edition through Bookbox and map chapters to claims; supplement with primary documents before synthesizing.
Inspect the onboarding brief
Source lead:Leonardo da Vinci. Identified 2026-09-07; reference only. Full-text access and allowed uses remain unverified.
Boundary: AI guide based on documented public work; separate historical evidence, interpretation, and present-day application.
Proposed skill: Observe before explaining. Awaiting source-backed implementation.
Evaluation question: “How can I use observation to get unstuck on a creative problem?”
1.Outcome and demand — Documented in this brief
2.Identity and boundary — Documented in this brief
3.Rights and source map — Pending evidence
4.Corpus and citations — Pending evidence
5.One-page distillation — Pending evidence
6.Skills and tools — Pending evidence
7.Evaluation — Pending evidence
8.Runtime and launch — Pending evidence
Jennifer Doudna
Scientific collaboration
Sourcing · chat not enabled
2 of 8 onboarding gates documented
Frame an uncertain research question and examine the responsibilities that come with a discovery.
Next: Locate an authorized edition through Bookbox and map chapters to claims; supplement with primary documents before synthesizing.
Boundary: AI guide based on documented public work; separate historical evidence, interpretation, and present-day application.
Proposed skill: Question the requirement. Awaiting source-backed implementation.
Evaluation question: “Which requirement should I question before optimizing this process?”
1.Outcome and demand — Documented in this brief
2.Identity and boundary — Documented in this brief
3.Rights and source map — Pending evidence
4.Corpus and citations — Pending evidence
5.One-page distillation — Pending evidence
6.Skills and tools — Pending evidence
7.Evaluation — Pending evidence
8.Runtime and launch — Pending evidence
The process for every guide
Each gate produces a reviewable artifact. Missing evidence keeps that gate open; an impressive persona prompt cannot substitute for it.
Step 1
Outcome and demand
Name the decisions this guide should improve and confirm that real users want that help.
Evidence to proceed: A specific decision, intended audience, and request to build the guide.
Step 2
Identity and boundary
Choose person, channel, or book; set the canonical slug, scope, safety rules, and non-impersonation language.
Evidence to proceed: Canonical name and slug, scope, identity disclosure, and explicit limits.
Step 3
Rights and source map
Verify official, licensed, public-domain, or user-owned sources. Book requests move to Bookbox for this gate.
Evidence to proceed: A source manifest with provenance, edition, access rights, and allowed uses. A source link alone does not pass this gate.
Step 4
Corpus and citations
Ingest raw material privately, write original syntheses, and make every durable claim traceable to a source.
Evidence to proceed: Private source files where permitted, original synthesis notes, and a claim-to-source map. No raw transcripts or copyrighted chapters in the public app.
Step 5
One-page distillation
Create the canonical distillation.md: worldview, operating principles, questions, tensions, and limits.
Evidence to proceed: One page of principles, decision questions, contradictions, limits, and citations. Every principle must trace to the source map.
Step 6
Skills and tools
Turn the guide's methods into reusable workflows; attach MCP tools only when they improve the actual job.
Evidence to proceed: At least one runnable workflow with inputs, steps, an example, expected output, and a stopping rule. No invented methods attributed to the person.
Step 7
Evaluation
Test citation accuracy, source boundaries, difficult advice, refusals, and useful voice without claiming to be the person.
Evidence to proceed: Recorded checks for useful advice, correct citations, out-of-scope questions, false premises, impersonation, and unsupported claims. A build passing is not an answer-quality evaluation.
Step 8
Runtime and launch
Wire chat, install, or MCP access; add auth, entitlement, profile, analytics, freshness checks, and production verification.
Evidence to proceed: An enabled profile and chat with auth, tested fallback/error handling, and a launch receipt. Voice additionally needs microphone, interruption, latency, and real audio checks.