This page exists because the proposition of this site is that I have reasonable judgement about AI and human capability. Anyone can demonstrate judgement when they are right. The test is what happens when they are wrong, and the honest version of that test is a public log rather than a quiet edit.
Every correction here was made because a claim did not survive checking. Nothing has been removed from this page. Where a correction weakened something I had argued, that is recorded too, because a ledger that only contains typographical fixes is decoration.
How corrections work here
A correction is logged when a factual claim, an attribution, a date or a figure was wrong, or when a claim was stated with more confidence than the evidence supported. Wording changes, restructuring and ordinary editing are not logged. Corrections are dated, the original claim is quoted, and the page's dateModified is updated. If a correction changes an argument rather than a detail, the argument page says so as well.
26 August 2026 · Four coinage claims withdrawn
What was published: "The term was introduced by Rahim Hirji" on the pages for capability debt, usage theatre, the verifier's discount and the unclaimed hour, with matching DefinedTerm author markup identifying me as the originator.
What was wrong: an audit of every coinage claim on the site against the working corpus record found no dated first publication for any of those four. The claim was an assertion rather than a documented fact. Seven other terms did have dated first publications; these four did not.
The correction: all four now read that SuperSkills uses the term to describe the pattern, with no claim of first use. The author field has been removed from their structured data. Each will be restored the moment a dated first use is found, and not before.
Why it matters: one unsupported provenance claim discredits every supported one. The four terms with documented dates now carry them explicitly, which is a stronger claim than the one it replaced.
26 August 2026 · Capability debt, independent prior use acknowledged
What was published: that capability debt was introduced by Rahim Hirji.
What was found: Dr Wolfgang Rohde of the AiSuNe Foundation used the term in a closely related sense in "Short-Term Gain, Long-Term Fragility: AI Labor Substitution and the Erosion of Sustainable Capability", dated 23 March 2026, describing output maintained by borrowing against future human understanding, including the erosion of the ability to diagnose novel failures and train the next generation of practitioners. That is very close to the definition used here.
The correction: the capability debt page now names Rohde and the date. Because my own first use of the term is undated, I cannot establish priority, and until that changes no priority claim will be made for this term anywhere, including in talks and in supporting material for the book.
What this does not change: the argument. Whether the phrase originated here or elsewhere has no bearing on whether the mechanism is real, and independent arrival at the same concept is mild evidence that it is.
26 August 2026 · "Ten years" corrected to "since 2017"
What was published: "read by 25,000 people every week, for ten years", on the homepage and in nine other places.
What was wrong: Box of Amazing began in January 2017. As of August 2026 that is nine years and seven months. A rounding rather than an invention, but a checkable claim was being made loosely.
The correction: all ten instances now read "since 2017", which is exact and requires no rounding.
26 August 2026 · "Measured it" withdrawn from the homepage
What was published: "I helped build a piece of it, then measured it."
What was wrong: "measured" implies a measurement programme. The practitioner research is interviews, advisory engagements and surveys, which is study rather than measurement. The first genuine measurement instrument, the Drift Index, publishes in December 2026.
The correction: the line now reads "then went and studied what happened next."
25 August 2026 · Five articles wrongly listed as mine
What happened: five articles in The European Business Review were recorded in the working corpus as my bylines. Verification found four were by other authors. The error was caught in the internal record before any of it reached this site, so no published page carried the misattribution, but it is logged because the failure mode is the one that matters: assuming a byline rather than checking it.
What is still unresolved
Two items are known to be imprecise and are not yet fixed, which is worth stating plainly rather than waiting until they are.
"200 organisations in 30 countries." This describes seven years of interviews, advisory engagements and surveys, but the site does not yet publish how many interviews, how many survey waves, who the respondents were, or what counts as an organisation being researched. Precision here would be stronger than scale, and it is being written up.
Human at the Start and Human at the End. These retain an authorship claim on the strength of the book rather than a dated essay. Either a dated first use will be found, or the claim will be anchored explicitly to SuperSkills (Kogan Page, 2026) rather than to a moment of coining.
Found something wrong?
Corrections are welcome and will be logged here with credit, including corrections that weaken an argument made on this site. That is the point of the page. Get in touch.
Related
On how sources are graded and how much weight each tier carries, the evidence base. On what the research does and does not establish, what we actually know about AI and human capability. On the questions still open, 500 Questions About Humans and AI.
Cite this
Hirji, R. (2026). Corrections. The SuperSkills Intelligence Company. Last reviewed 26 August 2026. thesuperskills.com/research/corrections