AI Made Me Doubt Everything About Programming

Felienne Hermans — professor of computer science in Amsterdam, high-school CS teacher, and creator of the Hedy teaching language — opens her DDD Europe 2026 keynote with a confession: "Recently I totally fell out of love with computer science." The talk is her explanation. Two decades of evidence that the field values difficulty over accessibility: her spreadsheet research was dismissed as "not real programming" even after she built a Turing machine in spreadsheet formulas, and Hedy — a free, Python-like language localized into 71 languages, including Arabic with Eastern Arabic numerals, which mainstream languages reject as an "invalid character" — was met at a language-design conference with "why don't they just learn English?" and, more tellingly, "why was that hard?", as if ease of implementation subtracted worth.

Her diagnostic tool is feminist epistemology — asking who creates knowledge and why. A 2016 paper on glaciers showed researchers oversampling heroic mountain glaciers while neglecting accessible rural ones; she maps this directly onto a field that venerates Haskell and C ("close to the metal") and dismisses spreadsheets, PHP, and JavaScript ("easy"). The same lens reads DDD as "talking to customers" wrapped in "requirements engineering," "event storming," and "ubiquitous language," because plain human practice doesn't sell tickets. She then widens it to history: von Neumann calculating detonation heights for maximum civilian casualties, IBM's Thomas J. Watson receiving a medal from Hitler for selling 1.5 billion punch cards a year to Nazi Germany, and IEEE still awarding a Von Neumann Medal under a mission of "advancing technology for the benefit of humanity" — a field, she argues, that never reckoned with its origins and is now repeating them with AI in warfare.

The AI argument runs through chess. Historian Nathan Ensmenger calls chess the "Drosophila of AI": the field's obsession with a binary, win/lose domain established a benchmarking methodology that researchers carried into language and art, where true/false doesn't apply. But chess also proves society can reject a solved problem — engines have beaten world champions since 1997 and competitive chess still bans them. Herbert Simon saw this in 1956: "In 10 years a computer will be the world champion in chess — unless it is barred from competition." Hermans wants programming to follow the same model: adopting LLMs is a choice, not an inevitability. Drawing on Peter Naur's 1984 distinction between intelligent activity and intellectual activity (having a theory you can explain, defend, and answer for), she argues LLMs produce output but not theories: "We have artificial intelligence, but certainly we don't have artificial intellectual activity."

Her indictment lands on the field itself: a 2017 study found the biggest predictor for not studying computer science is caring about social change, so the people who remain build complex things for their own satisfaction — "programming is to make programmers happy and programming is to make programmers proud." Against that she sets Ada Lovelace's 1843 line — "its province is to assist us in making available what we are already acquainted with" — as the best summary of a machine that reads the internet and rehashes it back, and closes on David Graeber: "the world is something that we make. We are allowing this to happen, but we can also not do that."

The gist pairs a cleaned-up summary with the full auto-generated transcript; the transcript garbles names (Hedy as "Haddy," Ensmenger as "Ensmanger," Naur as "Nauer"), so quote from the summary.