At the start of the year, I made seven predictions about how 2025 would unfold. Six months in, it’s time to mark my own work.
No AI wall (Grade: A)
I had foreseen that we’d see no slowdown in AI capability growth. Ten-million-token models exist (Meta, Mistral, Gemini). Benchmark scores are rising sharply - Frontier Maths, RKGI, MMLU. O3 is 25% on Frontier Math where GPT-4 was 2%. There’s been real progress. That said, some models are inconsistent and overthink things. Ask me how Flash 2.5 wrecked one of our workflows.Warp-speed deployment (Grade: A)
I said usage would skyrocket and token prices would fall. We’ve seen 30x–50x increases in demand (Google, Microsoft, China). Cursor passed $100m ARR. Other verticals are following fast. Agents are everywhere. Nothing here was overhyped.Bots would outtalk humans (Grade: A-)
We modelled this crossover point a while back. By word count, bots will be ahead of us sometime this summer. LLMs now consume and produce more text than most of the internet. It’s a shift in who is shaping language online.Waymo overtakes Uber in SF (Grade: B+)
They’ve beaten Lyft, and expanded to New York. They might overtake Uber by year-end, but not we’re not there yet. Tesla’s now in the mix with its Austin robo-taxi fleet, and that changes the game.Climate extremes intensify (Grade: A)
This one was almost too easy. The world’s on fire, literally. Record temperatures, terrifying anomaly charts, more billion-dollar climate events. This was baked in.Solar keeps breaking records (Grade: A-)
Solar PV output is up 30%, led by rapid buildouts in China, Vietnam, and Pakistan. Political resistance remains, but the economics are unstoppable..EVs shift up a gear (Grade: A-)
EV growth is steady, with Indonesia jumping from 20% to 80% in two years and BYD pushing into Europe. Tesla’s stumble gave rivals room, and legacy European carmakers have lost their cost leadership.
What I missed
Although we analysed these trends in Exponential View, I failed to include the following in my predictions:
The capex boom around AI datacentres, chips, and grid infrastructure.
The rise of humanoid robots. See Figure, BMW, Amazon.
The workforce impact. Salesforce, Microsoft, Google all report AI doing 20–50% of work. Some CEOs predict job losses of 10–50%. I should’ve seen that coming.
Looking ahead
By 2027–28 we’ll hit turbulence. Startups will be four or five years old, scaling fast, challenging incumbents. Incumbents like JP Morgan or Moderna, if they’ve built right, will start to reap results. It’ll be the moment vertical AI hits the real economy in force. That’s when things get interesting.
One question for you: We’re now closer to 2050 than to 2000. How does that change how you think about the world?
Let me know in the comments!
📅 Catch me live every Friday at 9 am PT | 12 pm ET | 5 pm UK.
Share this post