I'm writing to think, learn, and remember in public. I'll be here everyday.
I have a rule on my teams. We never use the phrase, “That’s not my job.”
That’s because I want to foster attitudes of ownership, providing value, and a sense of “we’re all in this together.”
However, there are some things that aren’t my job.
I can’t make you successful. I can’t make you happy. I can’t make you care.
That’s your job.
A chicken with no head isn’t thinking.
Its muscles and nerves are firing due to stored energy. The spinal cord just randomly triggers until that stored energy runs out. There is no rhyme or reason. No purpose. Plus, that energy runs out quickly.
Running around like a chicken with no head doesn’t help anything, and you can’t put it back on.
But if you still have a head, you can choose differently.
You can stop and think. You can apply method and process. You can try something and see what happens. Deliberately.
Keep your head.
AI agents overshare.
They create more documents, more paragraphs, more explanations, and more words. And if they’re writing code, they create more code, more comments, more tests, and more review items. More, more, more.
So much so that you can’t help but say, to yourself or out loud, “Shutup!”
I’ve said to my agent, “Too much! Be succinct, sparse, and more efficient. If you can use 2 words instead of 3, do that. Do not overexplain.”
We all have a friend or family member who is the same. How do you handle them?
It’s a different approach. We can be direct, terse, and even a bit rude with our agents. We wouldn’t use the same approach with a person. We’d probably try to be kinder.
With people, kindness means patience. With agents, kindness means constraints.
It feels good to have everyone on board.
And that’s the right thing to do sometimes. But also, real change might require pushing through when not everyone is on board. In fact, you might want to look for the kind of change that not everyone is on board with.
Somebody is always protecting something. Their status. Their expertise. Their kingdom. Whoever that is won’t be on board.
If change is the goal, not everyone will be on board. By definition.
The models, they are a-changin’.
Therefore, I’ve updated the ”How to Talk to AI” page.
Some terminology: The term “frontier model” is just fancy AI-speak for “the latest model.” The one’s that OpenAI, Anthropic, and Google announce in the press.
Enjoy.
Most people expected that LED lighting would lead to a significant reduction in global lighting power consumption.
Why wouldn’t you? Seems reasonable. The Occam’s Razor outcome. A 60W incandescent bulb costs about $10 a year to run versus $1 for a similar LED version.
But instead, we just use more lighting. Why not? It’s stupid cheap now, so it’s no big deal to add more street lights, office lighting, home lighting, brighter signage, larger displays, etc.
Overall, LED lighting has not significantly lowered the total amount of power that we apply towards lighting in the world.
That’s Jevons Paradox.
Two basic predictions exist today in the software world:
That second one is Jevons Paradox.
The relative efficiency (cost, speed, etc) of code development will lead to more:
And the unstoppable train of more will inevitably lead to more software jobs.
Personally, I think AI has unlocked a million ideas I have for little side projects around my house — power outage detection, power monitoring, 3D printing, etc. I’ve had these ideas for a long time, but the amount of work I knew I’d have to put in was a barrier. Now, it’s a piece of cake, all with a $20 a month subscription.
Machines provide leverage. AI provides leverage.
Figure out how to get AI working for you, or you’ll be working for AI.
Asking that’s a good start.
Not having an answer is a better start.
It might work. We think so. We don’t really know. We’re going to have to try.
It might be just the right thing to do.
Problem or Situation?
Seth Godin’s new one, The Knot, distinguishes between problems and situations.
A problem has a solution. You can find the solution. You can implement the change.
A situation has no solution. Why? It could be the people involved, timing, culture, or simply “that’s the way it is and no amount of anything you do will make a difference.”
The magic come when we learn to tell the difference.
Is this a problem or a situation?
Dig in on problems. Move on from situations.
You’ve got several tabs open with several agents working on task lists you’ve given them.
It’s neat and tidy in the beginning. You’ve given each one a structured and well-thought-out set of instructions. You’ve got a handle on it.
But as each one starts its job, the information load creeps up. Each one gives you updates full of jargon. They’re asking you questions — permission, this or that, etc. They get needy.
AI makes it easy to get out over your skis.
You need to land it safely. You can’t crash.
So rebalance. Pull it back a bit.
Multitasking is a myth for humans. One thing at a time.
Focus.
No, it won’t be perfect.
But you can make it better.
It’s worth making better. You’re worth making it better.
Better gets us closer.
