Step Back, transcript
An old green-screen terminal switches on. Someone types, in capitals, and deletes each line: AI will change everything. It's just fancy autocomplete. AI will end us all. It's a bubb... Sorry. Let's start again.
Hello. If all that has left you anxious, annoyed, both, or something else, let's step back to early 2023.
January 2023, ChatGPT running GPT-3.5. Prompt: Give me three sources on AI in libraries. It returns three confident, properly formatted references. None of these exist. One study found 55% of its citations were made up (Walters and Wilder, Scientific Reports, 2023).
Then it sped up. March 2023, GPT-4: convincing essays, still inventing sources. May 2024, GPT-4o: sees, hears and talks back. September 2024, o1-preview: stops to think before answering. February 2025, deep research: reads hundreds of sources for you. August 2025, GPT-5: can build an app from one prompt. Then GPT-5.1, 5.2, 5.3, 5.4, 5.5 and 5.6 between November 2025 and June 2026. September 2026: Claude Fable 5.1 on the 1st, GPT-6 Astra on the 3rd, Claude Opus 5.5 on the 22nd, and GPT-6 Sol and Luna about an hour later.
September 2026, same question. It returns three real references with links: Cox, Pinfield and Rutter (2019), The intelligent library, Library Hi Tech; Lund and Wang (2023), Chatting about ChatGPT, Library Hi Tech News; Walters and Wilder (2023), Scientific Reports. All real. It also read 212 more, drafted a literature review, and flagged two claims to check. I asked for three.
Today it can also: draft a research paper, evidenced and properly referenced; build a working app from a description; turn a spreadsheet into a report with charts; read a 500-page report and brief you in minutes; hold a spoken conversation in dozens of languages; translate a conversation as it happens; tutor a student, step by step, at 2am; fill in forms and organise your calendar; work on a project for hours, across many apps; make slides in your template and style; analyse ten thousand survey responses; make a short video, with sound, from a sentence; describe a photo to someone who can't see it; critique your argument before a reviewer does; help mathematicians with open problems; summarise a meeting and draft the minutes; compose a song; turn a voice note into this animation.
The screen clears. Pause.
If that felt too fast, it was. Most of us are still catching up. We'll need to help each other. This isn't simply about productivity, or the end of the world. It's about who we are, and who we become, in a world with tools like these.
Next time a massive opinion piece tells you what to think: take a deep breath, read it closely, talk it over with a friend, and use your critical judgement. That's enough from me.
Sample outputs are illustrative. Carlo Iacono, Hybrid Horizons.