Don't chase tools before you know how you work

It is June 2026 and a new frontier model shows up roughly every two days. Twelve of them landed in the first two weeks of this month alone. Open Facebook or TikTok and there are hundreds of accounts running the same play: a screaming title, a thumbnail with an arrow and a shocked face, this one repo that changes everything, this one tool you are an idiot for not using yet. A person who is new to all of this opens the feed in the morning and feels behind by lunch.

If that person is you, here is the honest version: most of it is not for you yet, and chasing it will slow you down.

The tool is downstream of the need. You cannot pick the right one for a problem you have not found. Before you know your own workflow, your own taste, what you are actually trying to make and where your real friction is, the trending tool is just another tab you will never open again. You are shopping before you know what you came to buy.

I can put a number on the alternative. I measured my own year of Claude Code: 483 sessions and just under 25 million tokens, most of it the model's own output. What produced that number was not the tool of the week. It was a workflow I settled into early and barely changed. Gate the decisions instead of reading every line. Keep memory in files. Ship small things and finish them. The total is large because the habit was steady, not because I kept swapping my setup for whatever was loudest that week.

The model itself stopped being the bottleneck a while ago. Fable 5 landed as the smartest thing available and I am still gating decisions the same way I did six months ago. Each new release moves a benchmark and changes almost nothing about how I actually spend a working night. The treadmill runs whether or not I step on it.

We have seen this movie. 2020 and 2021: gamefi, crypto, NFTs. Hundreds of posts a day about technology that was going to change the world, the same breathless tone, the same gurus. Time did the grading. Most of it was bullshit. AI is not that. It is real and it is already in my daily work. But it is young. The trending-est thing this week is a few months old. Nobody, not the accounts selling it and not me, knows which parts are durable and which are this cycle's JPEG. When you cannot tell what lasts, you do not want your whole approach pinned to a guess that is ninety days old.

So anchor to the thing that does last. That is not the tool. It is how you think, how you check work, what you are trying to build, where your taste sits. That compounds across every model and every framework that comes and goes. A tool you adopt because a thumbnail told you to, before you had a use for it, just sits in your stack as noise you now have to maintain.

I still read the release notes. I try a few things. But I start from a problem I actually have, not a tool someone is selling, and I keep the workflow that works until something genuinely beats it, not just out-markets it. The feed resets every morning. The way I work does not.