Editorial Standards
Last updated: 13 September 2026
This page sets out the rules I hold this site to. It exists because I write about AI governance, and a site on that subject should be able to show its own controls: where claims come from, what the reader is being asked to take on trust, how the machines were used, and what happened when I got something wrong.
What gets published here
Everything on this site is my own analysis, drawn from twenty-five years of delivery work and the published research I cite. There is no sponsored content, no affiliate links, no paid placements and no guest posts. Frameworks are published as versioned references with a change history. Client work appears only as case studies the client has agreed to, with figures the client has confirmed; where a figure has not been confirmed, none is given.
How claims are sourced
Every empirical claim carries a source, listed in APA 7 form at the foot of the piece and linked to the publisher's own page rather than to a secondary write-up. Primary sources are preferred: the report itself, the regulator's text, the standard. Vendor and consultancy research is cited when it is the best available evidence, and is labelled as such.
Posts that rest on statistics carry an Evidence & Methodology box that grades each claim:
- Measured. An observed result from a study, survey or audited dataset, with the sample described.
- Forecast or Estimate. A projection or model output. Cited because it is influential, not because it is proven.
- Adapted. A finding from one context applied to another, with the stretch made explicit.
- My call. A judgement from practice. No external source is claimed for it.
Each box closes with the condition under which the argument would be wrong. A statistic that cannot be traced to a real, findable source is removed, not softened.
How AI is used in producing this site
I use AI tools in research, drafting and editing, under my direction, in the same way I advise clients to: as a fast assistant whose output is checked before it counts. The positions, frameworks, judgements and recommendations are mine. Every statistic, quotation and attribution is verified against its source by a person before publication, a rule tightened in August 2026 after AI-assisted drafts were found to have introduced citations that did not hold up. Those cases are in the corrections log below.
Hero and illustrative images are AI-generated unless captioned as photographs; project photographs are real unless the alt text says the image is representative. Testimonials are quoted verbatim from LinkedIn recommendations and link to the original. No testimonial, quotation, case-study figure or client name on this site was generated or embellished by a model.
Independence
This is a personal site. The views, frameworks and publications here are my own analysis. They do not speak for Orion Five Engineering or any past employer or client, and they do not draw on the confidential information, data or proprietary methods of any of them.
Client reference letters are held for procurement teams on request and are not published, quoted or summarised here, at the clients' preference. Where a project page mentions such a letter, it says only that one exists.
Versions and deposited records
Frameworks, technical reports and reference architectures carry a version number, a first-published date and a last-revised date. Substantive changes produce a new version; the PDF deposited on Zenodo under each DOI is frozen, so a new version means a new deposit. The record on this site and the deposited file are kept byte-identical. Copy edits do not change the version.
Corrections
When a factual error is found, the page is corrected in place. Substantive corrections are noted on the page and recorded in the log below with the date and what changed. Trivial fixes (typos, formatting, broken links) are made without notice. To report an error, use the contact page; you will hear back within one business day, and if you are right the fix and the log entry will follow.
Corrections log
| Date | Correction | Where |
|---|---|---|
| 2026-09-13 | Sultan Haitham City site area corrected to 1,570 hectares, matching the client letter on file. | /projects/, project page and résumé |
| 2026-09-12 | AIGP credential issuer corrected to Singapore Management University. Earlier copy said IAPP, or "SMU via IAPP". The IAPP exam has not been taken. | Credentials, résumé, footer chips, partner portal |
| 2026-09-11 | Five posts corrected during the evidence-grading rollout: a GitClear figure that did not match its source, an unsourced closing-note statistic, a cited "SmartDev" study that could not be found and was removed, composite examples that read as real cases and are now labelled, an unverifiable Gartner pilot-failure rate, and a JPMorgan/Netflix case pairing that could not be substantiated. Three carousel decks carrying the same numbers were regenerated. | Blog |
| 2026-09-02 | Site-wide blog audit across 64 posts: a hallucination-rate figure matched to its real source, a mislabelled APA citation fixed, a PwC figure re-quoted correctly, unverifiable claims softened or removed, and the EU AI Act enforcement date corrected in three posts after confirming the postponement (originally written as August 2026). | Blog |
| 2026-08-21 | Six posts corrected after citation research: GitClear duplicate-code finding is 4x, not 8x; a coding-speed statistic credited to McKinsey belongs to Microsoft, GitHub and MIT; a cognitive-atrophy finding was wrongly credited to Harvard; three figures in the "wisdom gap" post (a Forrester ROI claim, a 23% measurement statistic, an MIT Sloan attribution) could not be verified and were removed; a PwC "creative self-destruction" citation could not be found and was removed; manufacturing ROI figures were re-read from the original report. | Blog |
| 2026-08-19 | Topic images on the speaking page contained AI-generated text with invented statistics. Replaced with images carrying no figures. | /speaking/ |
| 2026-08-07 | Cool N Lite Solar Film engagement end date corrected to 2025. | /about/, résumé PDF |
Contact
Questions about these standards, or about a specific claim, go to Terence Kok via the contact page.
Publishing my own corrections log felt uncomfortable the first time I wrote it out, which is roughly how I know it belongs here. I ask boards to show me their AI incident register before I believe their governance slides. It would be odd to hold a website about that discipline to a lower bar. If you find something wrong, tell me. It goes in the log, and the log is the point.
