We Need to Stop Pretending We Know What's Coming

Yesterday, Anthropic announced Project Glasswing and dropped Claude Mythos Preview into the hands of a small, closed group of tech and finance companies. The model was not released to the public. Not because it was not ready, but because it was, by Anthropic's own words, too capable to safely put out there.

It had already found thousands of high-severity software vulnerabilities on its own, including a 27-year-old bug in OpenBSD and a 16-year-old bug in FFmpeg that human security researchers had collectively missed for decades.

I sat with that for a while. A model so good that its own creators decided the world was not ready for it yet. That is not a normal product launch. That is a different kind of moment entirely.

And it made me realize something I think a lot of us have been quietly feeling but not saying out loud. We have officially lost the ability to predict what stays and what goes in this space. That is not a complaint. It is just the truth. The old mental models are not working anymore.

The confident takes, the listicles about AI-proof careers, the reassurances that some categories of work are too human to be touched, all of it has been running on assumptions that the last 12 months have quietly dismantled.

Cybersecurity was supposed to be one of the safest fields. Adversarial thinking, creative threat modeling, deep expertise. The argument was solid. And then Mythos walked through it without breaking a sweat.

I am not saying this to be dramatic. I am saying it because I think there is something freeing about admitting it. When you stop pretending you can predict the shape of the next two years, you stop optimizing for the wrong things. You stop trying to future-proof a specific skill set and start thinking about something more fundamental.

Which brings me to something I have been thinking about for a while now, and this Mythos moment just made it feel more urgent. We need to seriously change the way we think about productivity.

The current definition is almost entirely transactional. I did a task, I produced an output, I got paid, therefore I was productive. That equation made sense for a long time.

But it is increasingly a trap. Because if that is your only frame, then the moment AI can do the task faster and cheaper, your entire sense of self-worth and contribution collapses with it.

Here is a different way to think about it. You can sit on your couch, watch a film you love, and in between scenes give a well-thought-out instruction to an AI that produces something genuinely useful. Is that productive? By the old definition, barely. By any honest modern definition, absolutely. The value was in the clarity of your thinking, the quality of your direction, the judgment behind the instruction. The execution was handled. That is not laziness. That is a new kind of leverage, and we have not built a language for it yet because we are still using a vocabulary that was designed for a world of manual, linear output.

What this era actually demands is something people have historically treated as a luxury. Meta skills. The ability to think across domains, to make connections that are not obvious, to direct and discern rather than just execute. And alongside that, what I would call the softer things. Playing an instrument. Making something with your hands. Reading fiction. Going for a long walk without a podcast in your ears. These are not distractions from productivity. They are the foundation of it. Because the one thing that makes a human genuinely irreplaceable right now is not a hard skill that can be replicated. It is a mind that is rested, curious, and intact.

I genuinely believe that mental clarity is the most underrated competitive advantage of this decade. Not a specific programming language. Not a certification. Not a tool. A mind that is not burnt out, not anxious, not running on empty, that mind can direct AI, generate ideas, build relationships, ask the right questions, and adapt to whatever comes next. A mind that is depleted cannot do any of that, no matter how good the tools around it are.

So when people ask me what skills to develop right now, I have stopped giving the obvious answers. Learn to think well. Learn to rest without guilt. Learn something purely because it brings you joy, whether that is piano or painting or cooking or chess. Not as a productivity hack. Just because a person who is alive in that way is genuinely more capable of navigating a world that none of us can fully predict. Project Glasswing reminded me of that. The pace is not slowing down. But the people who will move through it well are not the ones who saw every turn coming. They are the ones who stayed grounded enough to keep thinking clearly when no one else could.