Axios CEO: AI 'Way Better Than Most Think' After Year Test
Get the Tech newsletter
Daily tech — startups, AI labs, chips, the launches that shape the next decade. Free.
- Jim VandeHei spent a year using AI obsessively across all Axios departments — testing OpenAI's Codex and Anthropic's Claude Code — and personally uses ChatGPT or Claude 1-2 hours daily via a phone-controlled 'AI personal operating system' tied to an always-on computer running multiple agents
- VandeHei assesses AI is 'smarter than 95% of the people on 95% of topics, 95% of the time,' and called it 'the smartest doctor I've met' after feeding it all his medical records, MRIs, blood work, diet and supplements — with his human doctor agreeing with AI diagnoses 'almost every time'
- VandeHei's view on job losses shifted from expecting 'subtraction' (cost-cutting, shrinking headcount) to believing the bigger opportunity is new revenue lines, with Axios now exploring three projects that 'simply weren't possible without AI'
- Business gains remain overhyped for now: AI hits internal walls on human use, security, system connections, and data access decisions, and agent-to-agent collaboration is 'a mess' that must be fixed before companies can experience the individual-level gains VandeHei describes
- Psychological effects are real: at age 55, VandeHei reports waking at 3 a.m. jazzed to test new ideas and accomplishing more than ever, but also shorter sleep and more anxiety — echoing the 'AI-pilled' Silicon Valley burnout stories he linked to
- A 'new class of super-worker' is emerging at Axios — rank-and-file workers whose brains are 'wired for AI' — and VandeHei urges every company to provide AI training, arguing the difference between power users and dabblers is the real dividing line
Why it matters: Coming from a media CEO who deployed AI company-wide and personally, VandeHei's account offers rare bottom-up evidence that AI's biggest near-term payoff may be new revenue lines rather than headcount cuts — a notable reversal of the 'subtraction' narrative he himself held a year ago. His warning that agent-to-agent infrastructure remains 'a mess' is a concrete deployment constraint every enterprise buyer should weigh before scaling.
Ask SkimNews
