A rational conversation on where AI is actually going | Benedict Evans

TL;DR

• 1

  AI is huge, but think internet or mobile, not instant sci-fi rupture: Evans says AI is "as big a deal as the internet or mobile and only as big a deal," which is still enormous, but a better frame than calling every new model the industrial revolution.

• 2

  We are in the 1997 phase, not the mature phase: Outside tech, adoption is still patchy, with some survey data showing only about 15 to 20 percent of even 13 to 18 year olds using AI daily and many people still barely touching it.

• 3

  The job apocalypse story misses how companies actually work: Enterprise software rollouts take years, not weeks, and even OpenAI and Anthropic are adding headcount while hiring more forward-deployed engineers and services help, not fewer people.

• 4

  Automating tasks is not the same as replacing jobs: Evans points to accountants, whose numbers kept rising through spreadsheets, databases, ERP, and cloud, and argues the key question is "what's the task and what's the job?"

• 5

  Model labs may not keep the profits: If foundation models remain interchangeable and lack strong network effects, Evans thinks the economics could look more like cloud or telecoms, where infrastructure is essential but margins are thin and more value sits higher up the stack.

• 6

  The practical advice is simple: use the tools instead of performing moral purity online: His blunt recommendation is to stop saying AI is evil on BlueSky and instead immerse yourself enough to understand what it can do today, because that is what makes you a stronger hire.