I’ve done 40+ side quests with AI. The kind of project I mean is small. Automating iMessage invites. Parsing a messy table. None of them were about immediate ROI.
I did them to accumulate pattern recognition and confidence. I kept the wins small on purpose. Those wins build the mental models I want in place before a complex, high-value problem lands.
Some people are waiting for stability before they start. They want a perfect tool. This is my personal take. The waiting is the risky move.
The gap between people building AI fluency and people waiting for stability is widening. I call it a K-shape. Early adopters gain leverage. Late adopters face obsolescence. People who know how to use AI take the jobs of people who don’t.
How long do you plan to wait for the ground to stop shifting?
Correct it where the model can see it
Most users fail to train an AI because they break the loop.
You ask for a draft. Then you copy it into your email client and edit the final 20% there. You send. Done. The AI never saw the correction, so it can’t learn your voice. Next time you get the same gap, and you edit the final 20% again.
Does this sound familiar?
Leave the draft in the chat. Correct it there. Ask for another pass. Correct that one too. Stay for several iterations, until the model learns your style. Do that and one-shot accuracy shows up later. Break the loop and the final 20% stays on you.
Try the 5-Email Rule. Correct the AI in the chat for 5 emails before sending. See how the 6th email sounds like you.
Paste the ugly table
AI is good at messy data. You can skip the pre-cleaning.
This tactic came out of an Excel coaching session. Copy an ugly, unformatted table from a website and paste it straight into an AI tool like Claude Cowork. Leave the formatting alone. The AI recognizes the pattern, the headers and the values, and structures them. From there you can automate the data entry.
Why are you cleaning the table before the model has seen it?
Next time you see a messy table, paste it into AI and ask for a CSV.
Side quests, on purpose
Treat AI learning like a video game side quest.
I know, this is a site about productivity, and here I am talking about video games. Stay with me. A side quest is small and low stakes. You do it for pattern recognition and confidence, long before a complex, high-value problem shows up.
That’s what my 40+ side quests were for. Things like automating iMessage invites. Parsing messy tables. I was collecting mental models, one small win at a time.
Pick one side quest this week. A small automation that saves you 10 minutes and feels fun to build.
The first time will feel slow
Brooks calls this the first time friction rule, and he’s right about the feeling.
When you delegate a task to AI, the first execution always feels slower than doing it yourself. His example is setting up a Zoom meeting. Your hands already know the clicks. Teaching the model takes longer. The sentence that shows up is, “it’s faster if I just do it myself.”
That feeling is the training cost. Day one will feel slow. Accept the initial friction. Stay with the tool. You are building the pattern for background automation later. Judge it on day ten, when the effort can drop to zero.
What are you still clicking through by hand because try number one was slow?
Identify one repetitive task you do manually and force yourself to do it via AI once, even if it takes longer.
When it says it can’t
The most underrated skill for AI adoption is curiosity. The willingness to ask, what if?
An AI tells you it can’t do something. Ask it, can you try it a different way? Then iterate. Those extra passes uncover capabilities a single attempt never shows you. That is the move from tool user to explorer.
A side quest works the same way. The first answer might be a no. The second question is where the capability shows up.
Identify one task you gave up on AI for last year and try it again today.
Build something they can open
If you’re a new graduate, or you’re out looking for a job, a traditional resume is becoming obsolete.
The new competitive advantage is building in public. Create a small, functional tool or app that solves a real problem for the employer you want. That demonstrates AI fluency. It proves you can execute.
A small working tool, built for the kind of problem a team actually has, can carry more weight than another line on a resume.
Build one small tool for your dream employer this week.
Five emails
Stop waiting for the perfect tool. Start building the skill of interaction now.
Open a chat. Bring an email you really have to send. Correct the AI in the chat for 5 emails before you send them. Then read the 6th. See if it sounds like you.
Recommended for you
Want the full system? 25X is the flagship productivity system we teach.
Explore 25X →