The method

Every thesis in these books comes with the condition under which it would be false.

“These books” are books 2 and 3. Each chapter ends with a thesis that could be proven wrong. The register at the back lists them all: the thesis, how mature it is, what would show it false, and when we look again. At first printing none of them is tested. “Supported” doesn’t mean “true”.

What is new here

• A register you can check. It turns “trust me” into “test me”. The frame is Imre Lakatos: a hard core (the chain: much of our pain is passed on) and a belt of theses that must hold or fail.
• The machine as a contrast agent. Book 3 asks what a machine doesn’t need: “a body that gets tired, a person who leaves and comes back, a face that can be lost, a time that ends.” That shows what a human being needs. Each comparison gets a level (illustration, structural finding, evidence), and the book never concludes from machines to people.
• Stories that declare their fiction. Book 2 tells seven stories across 500 generations. After each, “What Actually Happened Here” says what was invented. No citations in the stories, no invention in the commentary.
• Nobody to blame, and still responsible. “An order that rewards adaptation produces people who raise their children to adapt—not against their love, but with it.” And the gap: a woman understands everything at 8:52 and at 11:10 does what her grandmother did.
• A program as a portable prompt. Six steps, run on your own in the AI app you already use. What you write goes to that app’s provider, not to us. Review edition: ongoing tests, feedback welcome.
• The Seventh Criterion. Four requirements, testable from outside, for AI that talks with children. Saying what it is is already required in the EU (AI Act, Art. 50) and in California (SB 243, for companion chatbots), and California also requires pointing to crisis services. Handing over to a person, sending the child to its own people and not trying to bind the child are what it adds. Submitted to SSRN, under review. Read it on SSRN.

What is not new, and who said it first

• That we are made for a life we no longer lead, and that it went wrong when people settled: said since Jean-Jacques Rousseau (1755); Jared Diamond (1987); Yuval Noah Harari (2014). For diet and disease: S. Boyd Eaton and Melvin Konner (1985).
• That there was one wrong turn: disputed. David Graeber and David Wengrow argue there was no single break (2021). James C. Scott questioned the usual story of farming and the first states in Against the Grain (2017).
• That patterns pass from parents to children: attachment research since John Bowlby; in Germany Sigrid Chamberlain (1997) and Sabine Bode (2009). That childhood adversity shapes adult health: the ACE study (1998).
• Humanity as an operating system that needs an update: a metaphor, not a finding.
• Know thyself: a motto, not a thesis.
• Asking what makes us human by comparison with a machine: Brian Christian, The Most Human Human (2011). Book 3 asks about need, not ability.

What’s missing

Each book names the most important claim it could not make testable, and keeps it anyway.

• Book 2: that the error is a pattern, not an event. Every case that doesn’t fit could be called “a different pattern”.
• Book 3: that what is irreplaceable about humans shows up where the machine doesn’t need it. Any machine that someday needs one of these things could be called “a different case”.

Where it started

Book 1, HumanityOS Update (2025), is where the question started. It was written quickly, with AI research tools, and its references are being corrected for a new edition. It is not the foundation of the theses above.

How these books were made

Books 2 and 3 were written with Claude (Anthropic); each says how in its Appendix C. The author’s conflict of interest is on the About page.

Review the program

• Read the registers: book 2, Appendix A (EPUB).
• Try the prompt: review edition.
• Tell us what’s wrong: comments and feedback, or titscher@itki.gmbh.