Everyone has a prediction about where AI is going and its very different depending on who you ask. But they all have one thing in common: it’s just a hypothesis.
Here's the truth: AI development right now is so fast and so unpredictable that making confident 5-year forecasts is impossible. Even the people building it can't tell you exactly where this ends up.
But that doesn't mean you're helpless. It just changes what the right move looks like.
What the data actually says
A new Anthropic research paper (published this week) tried to measure how much AI is already affecting jobs. A few things stood out to me:
No mass unemployment yet. There's no measurable increase in joblessness for highly AI-exposed workers since ChatGPT launched.
But hiring for young people is slowing. Workers aged 22–25 are about 14% less likely to get hired into AI-exposed roles compared to 2022. The door isn't closed but it's getting narrower.
The most exposed jobs aren't the ones you'd expect. Computer programmers, financial analysts, customer service reps. These are highly educated and well-paid roles, not just low-skilled work.
No reason to ring the alarm bells yet, but the signs are clear.
What (I think) to actually do right now
Since we can't predict it, the goal is not to bet on the right skill. The goal is to stay hard to make obsolete.
1) Stay versatile, not hyper-specialised
The people who will struggle most are those who can only do one narrow thing that an AI can now replicate. The people who will thrive can combine things (e.g. coding, communication, marketing, creativity)
2) Stay genuinely curious about a lot of things (polymath)
This sounds vague but it's actually strategic. A person who is interested in ML, economics, biology, and design sees connections that a pure specialist misses. Multi-curious people build mental models that transfer. That's hard to automate.
3) Stay close to the technology
You don't need to master every new tool or skill coming out right now. But you can't afford to be a passive observer either. Use the tools, break them, understand their limits. The people ahead right now are not the smartest, they just picked up the tools 6 months before everyone else did.
4) Prioritise skills that compound
Writing clearly. Thinking critically. Breaking down problems. Learning fast. These become more important as AI automates the mundane tasks.
The takeaway
Don't try to predict where AI is going. That's a losing game.
Instead: stay versatile, stay curious, stay close to the tools. Build the kind of foundation that works in multiple future.
The uncertainty is real. But uncertainty is also an advantage if you move while others freeze.
One exception worth mentioning: if you're building a company, deep specialisation can actually be a competitive advantage because you need to be the best at something specific to win. But that's a different game entirely, and it applies to a small minority. For everyone else, the versatility advice stands.
💡 My Recommendation of the week
The Anthropic labor market paper I referenced above is actually worth reading if you want the full picture:
Have a great week,
Chris