ποΈ Driving stick shift makes you better at A.I.
βThe more you know about how something works, the more you can push its limits.β - Ayrton Senna
π This issue at a glance (2-second summary)
Learn A.I. on one platform at one speed, and you're stuck in one car. Learn the fundamentals and you can drive anything.
Where did you learn how to drive?
Last week was my first time in Oklahoma, and I ran a double header. First Tulsa, then Oklahoma City.
In both rooms, I taught A.I. using the metaphor of driving a car.
Every A.I. tool has four parts that determine performance:
- Make. The platform. ChatGPT. Claude. Copilot. Gemini. Grok.
- Model. The engine inside the platform. Opus. Sonnet. GPT-5.6. Heavy engines for heavy lifts.
- Gear. The effort level. Low gear for a quick email. Higher gear for a strategic plan. Ultra-thinking for coding and multi-step processes.
- Governor. The permissions and guardrails. A rental car tops out at 85 because the rental company set it there. A corporate Copilot tops out wherever IT drew the fence line.
With practice, you learn the right combination to get what you want from almost any A.I. platform. With the right settings and permissions, A.I. can do almost anything a human can do on a computer.
After the OKC session, one of the attendees offered to drive me to dinner. Ironically, he drives a stick shift sports car. It had been years since I drove stick. I enjoyed watching the metaphor come to life in real time.
Here's why this matters to you right now.
The whole A.I. industry just shifted gears. For most of 2026, the message from big tech was simple: burn more tokens. Meta and Disney built internal leaderboards ranking employees by A.I. usage. Nvidia's CEO cheered it on. The industry even coined a word for it. Tokenmaxxing. More tokens equals more productivity.
Then the receipts came in.
Microsoft started cutting Claude Code subscriptions. Uber's COO said the company blew its entire annual A.I. budget in four months. IBM found the average developer now spends under $150 a month on A.I. and ships better work than the leaderboard chasers.
The new word is valuemaxxing. Justify every token with a measurable business outcome. Code shipped. Bugs fixed. Deals closed. Hours returned to the human.
Here's where the stick shift comes back. A valuemaxxing driver knows when to shift up, when to shift down, and that running Opus on routine spreadsheet work is like driving a Rolls Royce to the grocery store. Yes, it gets the job done. But you're paying luxury prices for a $4 gallon of milk.
The tools will always change, so I teach the fundamentals. Make. Model. Gear. Governor.
Once your team gets this, they can still drive next year when the car on the lot changes.
Winners take action
Pick one task you've already run successfully with A.I. Next time it comes up, run it on a different model and a different effort level. See if you can match the quality while saving tokens. It may take a few tries to dial in. The payoff: you stop over-driving tasks that don't need it, and you learn which gear actually fits the job.
The tools you need
Challenge: Your team is repeating the same multi-step process every week: editing files, publishing content, moving data between apps. A.I. could handle it, but writing out every step every time defeats the purpose.
Solution: Record it once. Let the A.I. learn by watching.
Both Claude and ChatGPT now have a Record a Skill feature. You perform the task on your computer one time while the A.I. watches your screen and listens to you talk through it. When you're done, it turns the whole thing into a reusable skill you can run on new inputs with a single prompt.
In Claude Desktop: open Cowork, click the + in the composer, and choose Record a Skill.
In ChatGPT Desktop: go to Settings > Plugins > Add > Record a Skill.
Note that both of these require you to actually download the desktop application, not use the web version.
I used this feature to create a skill for my entire podcast editing and publishing workflow. Remove long silences. Find the hook. Trim the audio. Add the intro and outro. Format the show notes. Upload to the host. Schedule the post. I walked through it once, hit save, and now the whole thing runs on its own, using different apps on my computer for ~24 minutes while I do something else (that process used to take me an hour).
Try this in 10 minutes:
- Pick one weekly task that touches two or more apps on your computer.
- Open Claude Cowork, or go to ChatGPT Settings > Plugins > Add. Choose Record a Skill.
- Hit record. Perform the task once, narrating what you're doing and why.
- Review the skill draft the A.I. produces. Save it.
- Next week, say βrun the podcast skillβ (or whatever you named yours) to let it run while you move on to higher-value work.
- Check its output and fine-tune as needed.
This week on the pod
On King Moves episode 152, Does Every Piece of Feedback Mean You Need to Change?, I bring Justin something that stung me earlier this year.
I taught a private workshop for a company, and the employees were required to attend. The boss let them join Zoom cameras-off. I opened with the foundations before dropping into their specific tool, the way I always do, which is what the boss and I agreed on. A small group of the employees complained because my opening screens weren't in their exact tool. Those few voices also happened to be the loudest voices in the room.
Dozens of workshops, thousands of reviews, and I'd never heard that complaint before. I sat with it for days.
Justin's reframe helped me look at things through a different lens, and his coaching may help you too.
If you've ever let a single piece of criticism rent space in your head, this one's for you.
Stop playing life on hard mode. Automate your success.
Until next week,
Ethan King
A.I. Automation for Business Growth
Keynote Speaker | Author | CEO | Strategist
P.S. Catch me live at these upcoming events:
Public:
- Build My AI Team Live, your AI-employee-building walkthrough on your computer: Oct 26 and 27, 4β6pm ET both days
EO chapters:
- EO Fort Worth, Part 2: Oct 20
- EO New Jersey, Part 2: Oct 29
- EO El Salvador, 2-day workshop: Nov 4β5
- EO Iowa, Part 2: Nov 12
P.P.S. Want to bring the private training to your team? Just reply to this email to learn more.
