Have you ever carefully written an article or long-form post only for somebody to run it through an AI detector and say, “This was written by AI”?
Frustrating, right?
This is how I beat AI detectors...
Over the last several months, a lot of my fellow programmers, engineers, etc. have been talking about token use like it's a badge of honor. I even have one client who thinks it's a game to see who can use the most tokens.
Can we please stop with all that?
Don't let the industry fool you into wearing out your card on token burn. "This isn't keep up with the Joneses the AI remix". There is no pat on the back worth going broke because you spend bigger than everyone else.
Be smart with your money please.
It's not about how many tokens you use, but how much productivity you get from the tokens you use.
If you're using an agent or agents, most of your spend will come from the thought cycles and the iterations.
A lot of us will give agents an idea and we'll go back and forth solidifying the idea, and then the scope changes slightly, or we remember something we forgot, and we keep the turns and the cycles going.
Sure, it's productive, but when that bill hits? Ouch!
See, as time goes by, the better and more accurate you should be at prompting, which means the less "figuring out" the agent has to do, which means the less you spend, and the more you produce.
Here's what I do.
I do all of the heavy brain work in browser first. All of the idea refinement, all of the planning, the scoping, the brain-dumping and brainstorming in the browser.
Then, when I have it, I'll have a prompt generated to hand off to the agent - no wasted turns, no wasteful token burn.
If any of you are in that boat where your spend is high and you'd like to bring it down? That's my personal strategy.
I ran up almost 250 million tokens last month, but my spend? Barely $200 bucks.
And the amount I got done? Insane.
I wrote about my strategy in detail here, but it's not required reading. I already gave the "meat of it" in this post.