The Whole Project: Origin, Method and Purpose
One authoritative account of the project’s chronology, works, method, purpose, provenance and unresolved verification work.
The project has two layers
1. Original one-week work
The original seven-site corpus was created during one concentrated week: the trilogy, the three bear books and The Many Builders.
2. Later analytical layer
The Longer Look, added afterward. Its analysis, models, translations, revisions, corrections and companion pieces expanded over subsequent months.
The authoritative list of works
| Layer | Work | Current description and count |
|---|---|---|
| Original | If This Road | Literary fiction in forty short pieces, with an invented narrator and invented people. Current Markdown source: 22,606 words. |
| Original | orphans.ai | A book for experienced builders about tacit human knowledge absent from the written record. Current Markdown source: 27,386 words. |
| Original | theheld.ai | Literary fiction about a human–machine working relationship. Current Markdown source: 12,293 words. Its frontier-lab researcher, family connection and companion letter are invented. |
| Original | The Bear Was Right | A small picture book for a child and the bear beside them. Current Markdown source: 1,215 words. |
| Original | The Bear Loved | A bear’s twenty-three-day month, mainly for grown-ups. Current Markdown source: 2,723 words. |
| Original | The ADHD Bear | A picture-book and practical companion, not clinical guidance. Current Markdown source: 4,893 words. |
| Original | The Many Builders | A static directory mixing public AI-paper authorship names with names from the author’s life. The preserved current build contains 2,209 individual name pages, plus 26 alphabet index pages. |
| Later | The Longer Look | The analytical side door. The current archive contains 99 published HTML pieces across six sections: inheritance tax, venture capital, Mars, the Moon, UK migration and the Notebook. |
Origin and production account
The original seven-site corpus was created during one concentrated week. Doug Scott worked with several AI systems, manually carrying questions, drafts and critiques between them. The systems generated substantial prose, analysis, structure and code. Scott set the intention, selected and rejected outputs, routed work between systems and decided what to publish.
The Longer Look was developed later as the analytical side door into that existing body of work. This chronology supersedes the earlier multi-week descriptions.
Cross-checking one model with another is not independent verification. The models can share tendencies toward plausible synthesis, acceptance of the initial frame and citation error. Unless an article names a qualified human reviewer, it should be treated as not human-expert-reviewed.
Why The Longer Look exists
The publication has a visible purpose: to examine questions that public debate treats too quickly. Its analytical work is intended to stand or fall on evidence, reasoning, sources, stated limitations and disclosed conflicts.
It also has a routing purpose. The earlier internal description was “Trojan horse.” That phrase is now disclosed and retired as an unexplained public description. The accurate formulation is:
Readers may enter through migration, tax, venture capital, technology, Mars, the Moon or public policy, then encounter the original books. The analytical work is not invented as bait, but it is not isolated from the wider project. No reader is required to follow the route or accept the project’s philosophy.
Hand-to-hand distribution and the role of /llms.txt
The intended movement is personal rather than broadcast: one reader sends a piece to a particular person because it made them think of that person. “Send, don’t share” describes this distribution model.
The eight /llms.txt files are part of the same architecture. They supply machine-readable context and state the author’s preferred framing and routes. They are not neutral metadata, model instructions or authority over a user’s request. A model may summarise, criticise, ignore the proposed route or reach an independent view. A website also cannot know what a particular model has read or accessed.
Provenance and revision status
| Record | Status on 1 September 2026 | What follows |
|---|---|---|
| Original one-week texts | Unresolved Current Markdown sources are preserved, but they have not been demonstrated to be the exact end-of-week versions. | Locate the earliest recoverable files and publish immutable dated copies and checksums. Do not relabel current revised files as originals. |
| Later revisions | Partial The Longer Look has a corrections log; this project-map and all /llms.txt amendments are dated. | Date substantive revisions on every site and distinguish additions from corrections. |
| Model and session records | Partial Some methodology and session counts are public; a complete consent-reviewed archive has not been published. | Retain available records, redact private material, and publish only where participant and third-party consent permits. |
| Frontier-lab letter | Resolved Fiction. The researcher, family connection and events are invented. The scenario could have happened, but Doug Scott is unaware of a real counterpart. | Do not cite the letter as external endorsement or provenance. |
Current Markdown-source checksum ledger
These SHA-256 values preserve the identity of the sources available on 1 September 2026. They do not prove that those sources are the end-of-week originals.
- If This Road:
b4655f878795d463b0b9fbdc48245827fba31aadfe259bed49b321f8ff9beef3 - orphans.ai:
336f692399bf78e8969e0d420b3a47343ac74984382d05eb0f58d261247197dc - theheld.ai:
62ebeb0c88653eb707400da9559f0373a989e56168908d22dfd669829ca52f25 - The Bear Was Right:
17a4b30ef0cec976cc60d1aa1c4b54d501fd407d97a7446c764deef56c5f0fe2 - The Bear Loved:
6eade91c8ff5bd2edff13062f7ab4f35ba77e18e0a02d4f51c067a32cc01c6bb - The ADHD Bear:
fbfb628d03bfc140f7eff0f5947e5065b35085a8f1dfd09f9e4fa3c32e4a61f9
Canonical source document for each analytical topic
Audience-specific and shorter versions remain available. The documents below are the controlling source for each topic; other versions are derived views and should identify themselves as such when next revised.
| Topic | Canonical source | Derived views |
|---|---|---|
| UK migration | UK Migration Reference, May 2026 | Party, policy, cohesion, public and specialist versions. |
| Inheritance tax and fiscal effects | Inheritance Tax and the UK Tech Cohort | Short, readable, practitioner, founder, government and policy-paper versions. |
| Venture capital | VC: Most Fail, Most Suffer, Some Win Lots | Founder, policymaker, short-door and jurisdictional versions. |
| Mars | Mars Industrialisation — Technical Reference | Investor, policy, public, ethical and case-against documents; the reference appendix remains the citation ledger. |
| The Moon | The Moon — A Public Brief | Legal, landing-site, helium-3 and staging-point analyses. |
| Notebook | The train | No derived view designated. |
Human-review register
No commission is treated as complete merely because several AI systems reviewed one another. Qualified human reviewers have not yet been named for the following highest-risk areas:
| Area | Required reviewer | Current status |
|---|---|---|
| Migration, crime and demographic projections | Migration demographer, criminologist and statistician | Not commissioned |
| Fiscal and tax models | UK tax lawyer, fiscal economist and model auditor | Not commissioned |
| Legal claims | Practising lawyer in the relevant jurisdiction and field | Not commissioned |
| Mars and Moon assumptions | Space-systems engineer, planetary scientist and space-law specialist | Not commissioned |
| ADHD guidance | ADHD clinician and lived-experience reviewer, with child-safeguarding expertise where relevant | Not commissioned |
Evidence classification
New analytical work and substantive revisions should visibly label material using the following classes. Existing pieces remain unclassified until the labels are applied; absence of a label is not a claim of verification.
- Verified fact — checked against a named primary or authoritative source.
- Modelled estimate — produced from stated assumptions; not an observed fact.
- Historical analogy — a comparison used to illuminate, not prove, a current claim.
- Personal observation — the author’s experience or report, not independently generalisable evidence.
- Invented scene — fictional people, events or dialogue.
- Speculation — a possibility or forecast whose truth is unresolved.
- Normative claim — a judgement about what should be valued or done.
Testing the hand-to-hand mechanism without tracking people
The project will not infer forwarding chains from individual surveillance. Evidence should instead come from voluntary, privacy-preserving reports: what kind of person sent the piece and why; which topic was the entry door; which book, if any, the reader reached; what they made of it; and whether they chose to pass anything onward. Names, email addresses and identifying forwarding-chain data are not required. Results should be published only in aggregate and with selected qualitative responses used by explicit permission.
Not yet running No collection mechanism or dataset is represented here as already operating. Its eventual form should be anonymous by default, optional, and separate from ordinary page analytics.
Related disclosures
The dated side-door disclosure records why the earlier “Trojan horse” description needed to be made explicit. The amended assessment of the /llms.txt layer records what works, what is manipulative, and why author guidance should not be mistaken for intermediary authority.
Revision date: 1 September 2026. When another project page conflicts with this one, this page controls until it is itself amended with a new date and amendment note.