I Got a Five-Minute Fix by Complaining Out Loud

I do my real work on my desktop. Every build check meant grabbing the laptop, opening it, and logging in. I described the whole frustrating situation to Claude. It proposed something I hadn’t thought of. A fiddly laptop routine turned into a five-minute fix.

If you can say a friction point in plain language, an AI will often find a workaround you didn’t know existed.

Years before I used a password manager, I watched someone log into a site with a wildly complex password in one click. My mind was blown. I’d been accepting the annoyance as a fact of life. You don’t know what you don’t know. Most professionals can’t yet imagine which of their daily friction points an AI assistant can dissolve, until they watch someone do it. That’s the whole reason to keep watching demos, podcasts, and screen-shares.

The podcast was publishing fine. Nobody’s process was broken. Brooks was still spending about half a day per episode naming files, moving files, adding images, reading timestamps, and Googling every tool link for the show-notes post. He could have automated that years ago with Hazel rules, AppleScript, and Python. The figuring-out cost never seemed worth it.

AI eliminates that exploration tax

You describe what you want and it builds it. The biggest friction it removes is the friction of not bothering to remove friction.

After every recording he asked, what’s one tweak? He automated one annoying step, generating project thumbnails. Mid-solution, he realized six other small manual finish-up tasks could ride along. One Claude command now does seven things he had always done by hand. He never considered any of them worth automating individually. The 20% wrap-up work around a project vanished. Half a day of grunt work turned into 15 minutes of supervision. He drops the file in and tells it to go.

Each fix revealed the adjacent one

Friction removal compounds. Brooks adds a caveat, and I agree with it. Improvements need a payoff. Don’t tinker just because you can. That’s kaizen: continuous improvement as a standing mindset.

When an AI spits out twenty suggestions, the temptation is an impact/effort matrix. I call that procrastination in disguise. On your first pass, behavior beats optimization. Tackle any friction point, even a tiny one. Get the reps. Build the muscle. Small wins first, because a finished micro-fix beats a perfectly ranked plan that never executes. Keep it atomic.

I use the Three Times Rule

It’s my golden rule. If something annoys me three times or more, I fix the root cause. An app that keeps crashing? Update it or switch tools. The third annoyance is the trigger, so I don’t debate whether it matters enough.

I picked this up from a classic idea in operations. Every system has exactly one constraint, and removing it lifts everything else. I’ve been car-free for over a decade, so distance is my constraint on the gym. A gym 30 minutes away means I won’t go. Ask yourself: if what weren’t there, I’d do this all the time? Whatever you name is the friction point.

You can’t audit yourself from the inside

I saw that demoing my workflow on a friend’s TV over AirPlay. Granola, Wispr Flow, files and folders. Watching from the audience side, I thought: why does moving this file take so long? I’d done it hundreds of times and absorbed the slowness. Demo one workflow this week, or record a Loom of yourself working. Write down every time you think, why is this slow?

Brooks hired a virtual assistant and, while training her, realized his own process was completely cockamamie. He’d solved the problem, celebrated, and stopped asking if it was still the best way. It kept working just well enough. Workflows decay quietly. An outsider’s questions expose them. Document one workflow as if you’re training a new hire.

That training is the audit

When the question is what AI could do, record the work first, Loom or similar, and turn those recordings into a playbook. Then hand that playbook to Claude or ChatGPT to streamline it, run it, and let you verify the output. You can’t remove friction you don’t understand.

Desktop apps like Claude or ChatGPT save every conversation as a session file in a hidden folder, a detailed log of how you actually work. Start a new session and say: go through all my past session files and tell me what workflow tips and suggestions you have. It’s a neutral third party that has watched you work all week, and it will surface friction you can’t see. Use the desktop app.

A browser is sandboxed and can’t read those files

I haven’t made a formal mind map in four or five years. Verbalizing the problem replaced outlining for me. Instant feedback. Back and forth. The AI remembers everything, and I don’t manage a file. If I want it visual, I say “create a mock-up of what we just talked about” and I get an image or an HTML page.

Brooks noticed the same shift when he opened MindNode for the first time in months. He used to pause a training video and build the map. Pause, type, pause, type. Now he drops the transcript into an AI tool and works from there. Someone had to watch the video before anything happened. That queue collapsed. Which apps have you stopped opening? That’s the evidence a workflow already changed.

Pick one thing that annoyed you three times this month

Describe it to Claude in plain language, the same way I described the laptop. Ask for one workaround. Finish that fix before you rank a list.