Definitions episode · Oct 14, 2026 · 51 min

What is an AI Agent?

Everyone uses the words. Almost nobody agrees what they mean. So I asked my own AI agents for their definitions, then went back to the source: the 1956 Dartmouth proposal where John McCarthy coined "artificial intelligence" — read in his own words. Then I build the vocabulary one layer at a time: AI, classic AI, LLMs, agents, AI agents, and the agentic harness that makes the last one real. Yuval Harari on agency. My AI coffee machine test. Chatbot vs agent in one line. The real job-cut numbers behind the AI-washing story. And the question underneath all of it: was the word broken on purpose?

In this episode

What you'll see

  • AI, classic AI, LLMs, agents, AI agents — five definitions one at a time, and the agentic harness that makes the last one real
  • Nobody agrees on these words: why "AI" itself got fuzzy, and where AGI fits
  • The 1956 Dartmouth proposal in McCarthy's own words: his seven questions and the budget that started it all
  • The movies that trained our picture of AI — HAL, Skynet, and friends — before we ever met a real one
  • Yuval Harari on agency: agent vs tool, the atom-bomb thought experiment, and the AI coffee machine test
  • An AI agent in one sentence, plus the chatbot-vs-agent line you can actually reuse
  • AI washing and the real layoffs: 8% of US job cuts cited AI in February, 16% by April, 23% by June — 100,000+ cuts this year
  • Chess and Deep Blue: 1996, 1997, Stockfish — and why a chess win is not AGI
  • Agent drift: what happens when an agent wanders off task
Every claim sourced

Sources cited in this episode

  1. John McCarthy, Marvin Minsky, Nathaniel Rochester, Claude Shannon (1956) — A Proposal for the Dartmouth Summer Research Project on Artificial Intelligence — the founding document, read in McCarthy's own words on screen; also available as a PDF; and the original 1955 typescript at the Computer History Museum
  2. Grace Solomonoff — The Meeting of the Minds That Launched AI (IEEE Spectrum, May 2023) — the story behind the 1956 group photo shown on screen, written by the wife of participant Ray Solomonoff, who's writing a book about that summer
  3. Pew Research Center (June 2026) — What do Americans think AI is? — 5,119 adults; chatbots 29%, robots & sci-fi 8%, images/video 7%
  4. Pew Research Center (April 2025) — How the U.S. public and AI experts view artificial intelligence
  5. Tom F.A. Watts — Terminator at 40: this sci-fi B-movie still shapes how we view the threat of AI (The Conversation, 2024) — read on screen in the movies chapter
  6. A Final Reckoning on AI (Modern Age Journal, 2025) — Mission: Impossible's sentient rogue AI as the 2025 reboot of the 1968 HAL fear
  7. IBM — The evolution of AI agents — how "agent" traveled from Russell & Norvig's 1995 textbook to today
  8. CSIS — Lost in Definition: How Confusion Over Agentic AI Risks Governance (2026)
  9. SEC press release (March 2024) — SEC charges two investment advisers with AI washing — $400K combined penalties
  10. Reuters — New wave of investor lawsuits targets AI washing (Feb 2025)
  11. IBM — AI vs. machine learning vs. deep learning vs. neural networks — the vocabulary split
  12. LLM vs. generative AI vs. AGI (AI TLDR) — where AGI fits, and the ARC-AGI-2 gap (AI ~4%, humans ~100%)
  13. Fortune (March 2026) — Jensen Huang says AGI is achieved — nobody agrees what AGI means
  14. Financial Times (June 2025) — Why Big Tech cannot agree on AGI
  15. Ian Bogost — 'Artificial Intelligence' Has Become Meaningless (The Atlantic, 2017)
  16. Forbes (Aug 2023) — Is It Time We Stop Saying AI?
  17. IBM — Deep Blue — computer chess history behind the 1996 and 1997 Kasparov matches
  18. Challenger, Gray & Christmas job-cut reports — the AI-cited layoff percentages quoted in this episode come from their monthly reports rather than one canonical page, so I show the numbers on screen instead of linking a source I can't pin exactly
  19. Yuval Harari, Tanner Lectures on Human Values — the agency excerpt I react to; there's no stable public transcript, so it's described rather than linked
  20. + the full chapter list and timestamps in the video description
How I cite: I read the 1956 Dartmouth proposal straight from McCarthy's own words, linked above. Primary sources linked wherever they exist; anything I'm not reading from a source is clearly flagged as my opinion. This episode is commentary and criticism, not financial, medical, or legal advice.
Community

Compare notes with me all week

I build all of this in the open, with a crew of AI agents. If you're directing agents too — or wondering where to start — the Discord is where we trade what actually works.