A Decade of AI: My Reading History
368 articles saved from 2014–2026
Since 2014, I've been saving articles about artificial intelligence to a Feedly board—not systematically, just pieces that caught my attention or seemed important at the time. Looking back at this archive reveals how dramatically the AI conversation has evolved, from fringe existential concerns to mainstream disruption.
My first save was prescient: "Artificial Intelligence could kill us all. Meet the man who takes that risk seriously." That 2014 article treated existential risk as an eccentric preoccupation. Today, it's a central concern—and the subject of my forthcoming book with William Leiss.
2014–2016: Early Warnings
In these years, existential risk was a fringe concern discussed mainly by philosophers and a few technologists. Most coverage focused on explaining deep learning to general audiences—what it was, why it mattered, and what it might do. Google announced an "AI kill switch" in 2016, a concept that seemed almost whimsical at the time. The "Canadian Mafia" (Hinton, Bengio, LeCun) emerged as the field's founding fathers, their decades of patient work suddenly vindicated.
Sample articles from this era:
2017–2019: Building Momentum
AI moved from research labs to mainstream headlines. Voice cloning emerged and immediately unsettled people ("I trained an AI to copy my voice and it scared me silly"). Cambridge Analytica exposed AI's potential for mass manipulation. The military began serious AI investment, and questions about Asimov's laws resurfaced as robots entered real-world applications. Geoffrey Hinton—one of the Canadian Mafia—started publicly expressing worry about what he'd helped create.
Sample articles from this era:
2020–2022: Pandemic Acceleration
Sundar Pichai declared AI "more profound than fire." COVID accelerated AI adoption across every sector—chatbots replaced call center workers, AI discovered antibiotics, and facial recognition became ubiquitous. Timnit Gebru's firing from Google sparked a reckoning over AI ethics in big tech. GPT-3 arrived and stunned observers. Meta's language model survived only three days online before being pulled for safety concerns. The foundation model era had begun.
Sample articles from this era:
2023–2024: The ChatGPT Disruption
ChatGPT changed everything. Google scrambled to define what AGI even meant. "Nothing, Forever"—an AI-generated Seinfeld parody—ran 24/7 on Twitch until it went off the rails and got banned. MIRI, the institute that pioneered AI existential risk research, announced it was shutting down. China moved decisively to regulate AI while the West continued to debate. The question shifted from "can AI do this?" to "should AI do this?"
2025–2026: The Agent Era
The discourse exploded—half my total saves are from these two years alone. Alignment researchers warn that AI agents require AI defense. The $3 trillion infrastructure buildout continues despite uncertain profits. "Hype correction" became the dominant theme—even Demis Hassabis called some AI boosterism "embarrassing." Disney licensed 200+ characters to OpenAI for $1 billion; activists launched "Poison Fountain" to contaminate training data in protest. Claude Code reached what Zvi called "critical mass" in developer adoption. The bubble question looms: what happens when it pops?
Sample articles from this era:
This retrospective was generated by analyzing 368 bookmarks exported from my Feedly AI board, with assistance from Claude. The archive spans from March 2014 to January 2026.
Browse the full archive (349 articles with Wayback links) →