The timeline

Box of Amazing has been published every Sunday since January 2017. That is five years and ten months before ChatGPT appeared, and it means the questions this research now answers were being asked here weekly while they were still curiosities rather than a category. This page is the dated record: when each idea first surfaced, what form it took, and how it developed. It exists for a specific reason. In a field filling rapidly with people who discovered that humans matter in the age of AI the week ChatGPT launched, a continuous public record is the one dimension nobody can manufacture retrospectively. It is also, deliberately, an honest record: the early issues were curation and commentary rather than original argument, and they are described as such.

2017: the first year

Issue 1 went out in the third week of January 2017, hand-sent, to a list of a few hundred people. Issue 2, "Imagine if there was no Uber", is dated 29 January 2017. The issues were numbered and carried deliberately odd codenames: Chineasy, Cyborgia, Mutancy, Cashish, Bubblicious. Weekly, on Sundays, without a break.

Three issues from that first year matter for what came later. Issue 18, "Employment Armageddon" (21 May 2017), previewed as "Automation Apocalypse", is the earliest dated instance of the automation-and-employment question in this archive. Issue 21, "Curiosity" (11 June 2017) carries, as a title, one of what would become the seven SuperSkills, nine years before the book. And "Reinventing Education" (5 November 2017) opened the education thread that runs through everything since.

In July 2017 the list moved to Mailchimp at Issue 26, titled "Dawn of a New Era", sent to 441 people. The format changed with it: numbered codenames gave way to descriptive subjects.

2018: the prediction habit begins

The annual prediction series starts here, and it has not broken since. "Predictions for 2018" (7 January 2018) went to 496 subscribers. It closes with "50 Themes to watch out for in 2019" (23 December 2018), the load-bearing one, whose seventeenth theme was that soft skills would be the X-factor in the workplace. That was written four years before generative AI made the point fashionable.

The territory sharpens through the year. "Ethical AI, telexistence, robot workers" (10 June). Yuval Noah Harari on being human (19 August). "AI Robot Teachers" (26 August). Soft skills for teenagers (16 September). "Girls in AI and the Future of Work" (23 September). "AI in Education" (14 October). "AI beats lawyers, hologram education and AI lie detection" (4 November). "Unethical Superhumans" (2 December). The list grew from 496 to 997 across the year.

2019: emotional intelligence and early careers

Edition 100 went out on 10 February 2019. Two issues from this year read differently now than they did then. "AIEI: Artificially Intelligent Emotional Intelligence" (6 January 2019) put artificial intelligence and emotional intelligence in the same frame, which is where the empathy argument in this research now sits. And "Minterns taking Minternships?" (25 August 2019) asked what happens to internships, three years before the missing-rungs argument had a name.

"How I curate Box of Amazing" (18 August 2019) is worth noting for a different reason: it is the first issue about the method rather than the material. The list passed 2,000 in December.

2020: the education year

The pandemic pushed the newsletter into education and the future of work, and it stayed there. "All about COVID-19" (1 March), "Remote Everything" (22 March), "What the future looks like" (19 April). Then a run that reads, in retrospect, like a table of contents for what was coming: "AI Grading" (16 August), "Experiments in Education" (30 August), "AI Tutors" (13 September), "Future Proof" (20 September), "Why technology has a will" (13 December) and "A machine will do your job in 2025" (20 December).

AI tutoring was the subject of a Sunday newsletter in September 2020. The first serious field experiment on whether AI tutors help or harm learning was published in 2025, and it is now the anchor of how humans learn with AI. Readership peaked at 3,085 in August 2020.

2021 to 2022: the pause

The 200th edition went out on 7 February 2021. "Winning the Future" followed in October. The fifth anniversary issue was 23 January 2022, and the final send of that era, "An important note from Rahim", was 13 February 2022, to 2,945 subscribers. Roughly 250 issues, weekly, over five years.

Nine months later, ChatGPT was released.

2023 to 2026: from curation to argument

Box of Amazing returns on Substack in 2023, and the change is not just the platform. The earlier era gathered what was happening; this one argues about what it means. The Architecture of Drift sets out the drift versus design matrix. Are You Flying or Are You Being Flown? (1 March 2026) takes the aviation analogy into skill atrophy. The Great Unbundling of Work (25 May 2025) argues that jobs are not disappearing but dissolving into tasks. The Case for Being Bad at Things (18 January 2026) states the missed-reps argument directly. The Thing That Proves You're Human (25 January 2026) names outsourced recognition. The Flake 99 Theory of Being Human (8 March 2026) argues that constraint was the curriculum.

The bylines follow: CEOWORLD on drift versus design (July 2026), the Observer on the next AI power class (July 2026), Entrepreneur UK on AI exposing decision quality (July 2026), the European Business Review on accountability gaps (August 2026). SuperSkills: The Seven Human Skills for the Age of AI is published by Kogan Page in 2026.

When each idea was first published

The Substack archive from 2023 onward has now been catalogued, and it dates the vocabulary. These are earliest confirmed publications rather than moments of coining, since earlier private or spoken use cannot be ruled out. But they are checkable, which is the point.

• The missing rungs, 21 September 2025, in The Missing Rungs: What Nobody Will Tell You About AI and Your Job. Anticipated in Critical thinking a year earlier, on 10 November 2024.
• Drift versus design, 16 November 2025, in Drift vs Design, developed into the matrix in The Architecture of Drift on 15 March 2026.
• Synthetic seniority, 22 May 2026, in Synthetic Seniority, subtitled "Why AI Output Is Masking a Corporate Capability Crisis", with the remedy in Solving Synthetic Seniority on 5 July 2026.
• The missed reps, stated directly in The Case for Being Bad at Things, 18 January 2026.
• Outsourced recognition, 25 January 2026, in The Thing That Proves You're Human.
• The Flake 99 Theory, 8 March 2026. The Great Unbundling of Work, 25 May 2025. The Half-Life of Skills, 8 June 2025.

Several further coinages sit in the archive without a page here yet, among them The New Imposter Syndrome (17 May 2026), subtitled "Not knowing the origin of your own thoughts", The Reverse Singularity (10 August 2025), Symbient Intelligence (2 March 2025) and Zombie Work (1 June 2025). They are listed rather than claimed, because a term is not a framework until it has been argued properly.

Nine years of annual predictions, across two platforms

The annual practice is now documented without a gap from January 2018 to January 2026. In the email era: Predictions for 2018 (7 January 2018), 50 Themes for 2019 (23 December 2018), then editions for 2020, 2021 and 2022. On Substack: 50 Emerging Trends published on 1 January in 2023, 2024, 2025 and 2026, plus 20 Questions for 2025 (8 December 2024) and 20 Questions for 2026 (7 December 2025).

Nine consecutive years of dated public forecasting, kept with the misses intact, is the part of this record hardest to replicate. The assessed version is at the predictions record.

What this record does and does not claim

It does not claim that the SuperSkills framework existed in 2017. It did not. The early issues are curation with commentary: a weekly reading of what was emerging in technology, with a point of view attached. Nothing in 2017 or 2018 advanced an original thesis about human capability, and describing it otherwise would be the kind of retrospective tidying this page exists to avoid.

What it does establish is attention, sustained and dated. The automation and employment question was a Sunday newsletter subject in May 2017. Soft skills were flagged as the workplace X-factor in December 2018. AI tutoring and AI grading were covered in 2020. Roughly 250 weekly issues, and something in the region of five thousand pieces of curation, across five years before the technology arrived that made the subject urgent. The argument came later, and it came from somewhere.

A note on sources

The 2017 to 2022 issues predate Substack and have no public URL. They are cited from the author's own archive, with dates and subject lines verified against Mailchimp send records and, for the earliest issues, the original sent mail. Where a subscriber figure is given it is the number of recipients on that campaign. The Substack essays from 2023 onward are linked directly. Nothing here is reconstructed from memory, and no date on this page is an estimate except the launch of Issue 1, which is placed in the third week of January 2017 from the cadence of Issue 2 and confirmed by the fifth-anniversary issue of 23 January 2022.

Related SuperSkills research

The annual prediction record, with the hits and the misses kept, is at the predictions record. The vocabulary this developed into is at the term canon. The evidence behind the current arguments is in the evidence base, and the wider field map is in the Canon. The claims themselves, banded by evidence strength and including what remains unknown, are in what we actually know about AI and human capability.

About this record

Compiled by Rahim Hirji, author of SuperSkills: The Seven Human Skills for the Age of AI (Kogan Page, 2026) and founder of The SuperSkills Intelligence Company. Dates and subject lines were recovered from his own newsletter send records in August 2026. The record is extended as earlier material is confirmed, and corrected if anything here proves inaccurate. Box of Amazing continues weekly at boxofamazing.com.