Buzzword Factories
We all know buzzwords when we hear them.
Each profession and each level has its favorite buzzwords that make you roll your eyes. Each time you hear one, you lose a little respect for the person who used it.
LLMs are buzzword factories. They spew buzzwords and build word salads like it’s their job.
So tell them to stop.
I always tell my agents the following:
“Use plain and direct words. Never use jargon or buzzwords. You can assume the audience is professional
If you create written or presentation material, don’t use buzzwords. If you’re using AI to help you create that material, don’t let it use buzzwords.
Friends don’t let friends build buzzword word salads.
You’re the boss. Be the boss.
The Definition of Good
You want to work for a good company, right?
You want to go to a good school, right?
You want to live a good life, right?
Yes, of course.
But what does good mean?
Does good mean top of a list like Fortune 500, Glassdoor, or US News & World Report? Does good mean test scores, funding, or demographics? Does good mean beliefs, bank account, or healthspan?
We talk at each other like we all know what good means. Like, “You know it when you see it.” What we find, however, is that your good and my good may not align, and we get confused. So we seek others whose good looks like ours.
Once we all congregate around our definition of good, we can get down to the business of working, learning, and living a good life. The right way.
Funny how the right way always seems to come with a leaderboard.
Halfway
Are you halfway there or halfway gone?
Both, of course.
It’s up to you how you see it.
10 Years
What can you do in 10 years? What will you be like in 10 years?
10 years is:
- 520 weekend
- 1040 weekend days
- 30 summer months
- 130 summer weekends
- 10 vacations
- 11000 meals
- 3600 showers
- 80 haircuts
- 3 years of sleeping
- 3 years of working
- 4 years of other stuff
- 3652 daily blog posts
10 years is a long time. Or is it?
Just Do What Your Told
AI agents excel at doing what they’re told.
It’s basically what they’re built for.
But you? Is that what you’re built for?
You can, and maybe that will provide the status or the safety that you’re looking for.
But is that how you’ll make a difference?
Let the AI agents do what they’re told. You can choose, however, to do what you’re called to. Deep down.
Nobody ever made a change in this world by just doing what they’re told.
Does it Need to be Captured?
Do the requirements need to be captured?
Do the meeting notes need to be captured?
Does this picture of your dog cocking its head need to be captured?
Does a video of the band from 100 rows back need to be captured?
Sometimes the answer is definitely “Yes!” Information, decisions, a moment, or a memory might all be good reasons to capture.
AI can help in some instances. But AI also makes it easy to capture too much.
Don’t sacrifice presence for capture.
Cognitive Surrender
An interesting paper from the Wharton School at UPenn defines the term “Cognitive Surrender.”
It means adopting AI outputs with little to no scrutiny, and in so doing, bypassing human intuition and deliberation.
Nailed it.
We’re living in a time of cognitive surrender. Professionals, students, and hobbyists with access to LLMs just let it think for them.
Look at any post, article, or message you receive on LinkedIn. I’m guessing here, but I would estimate that at least 90% of the content is AI-generated with little to no human oversight. LinkedIn is useless slop right now.
We also know that teachers are struggling because their students are doing the same. And as someone who works at a company aggressively pushing AI onto its workforce, I see the results in a professional knowledge work setting.
AI is here to stay. It will become increasingly ingrained into individuals’ and the system’s workflows. It’s too useful. It’s too fast. It’s too magical.
What do we do?
We insert AI into our cognitive process, rather than offloading it. It’s still your work. It’s still your output. Make sure you reason along with it. Ask it for confidence scores and when it’s uncertain. Read it yourself. Think about it yourself. Test your trust.
Keep your human cognition in the loop.
Fitting In Again
As teenagers, most of us desperately wanted to fit in.
Nerds, jocks, band geeks, popular, druggies. Somewhere. That’s our deep human need for connection combined with our hormone-raging teenage biology.
As an adult, you likely fit into several groups. You may even find yourself fitting into groups with differing opinions.
Be careful, however, how dedicated you are to fitting in.
Because if you’re trying to do something, say something, or make a difference, some of those groups should think you’re crazy. Even the ones you fit into.
Nobody makes a difference without someone else thinking they’re crazy.
Its Not My Job
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.
Headless Chickens
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.
TMI
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.
No On Board
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.
At The Speed Of Model Change
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.
Jevons Paradox
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:
- AI will replace and wipe out all junior software engineering jobs.
- AI will lead to even more software engineering jobs, at all levels.
That second one is Jevons Paradox.
The relative efficiency (cost, speed, etc) of code development will lead to more:
- More apps
- More features
- More ideas prototyped
- More automation
- More refactoring
- More tools
- More, more, 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.
Will it Work?
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
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.
Out Over Your Skis
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.
It Won’t Be Perfect
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.
Your Passive Voice
In writing, passive voice is a sentence in which the subject receives the action rather than performing it.
“The dog bit the man” is active.
“The man was bitten by the dog” is passive.
Pro writers learn to write with active voice early on because the passive voice makes writing less clear and generally less interesting. Non-pro writers use passive voice everywhere because it’s kind of natural. But it’s one of the tells of an amateur vs. a pro.
But what about in your life? Are you receiving the action or are you taking it?
Are you using your active voice?
Kill Your Darlings
In the 2000’s into the 2010s, Apple did the unthinkable.
They cannibalized their own massively successful product’s market share with a new product. They killed the iPod with the iPhone. MBA’s and executive leadership would have told them it was stupid.
But then again, contrast that with Kodak, which followed the academic advice. They didn’t commercialize their own technology for digital cameras precisely because they knew it would cannibalize their cash cow — film.
We know where this all ended up.
Yeah, that’s one example, and there are plenty of counterexamples, but there’s a lesson here for you and me.
Don’t be afraid to kill your darlings.
Your beliefs. Your workflows. Your homegrown tools. Your sure-fire solutions. Your mega-hits.
Even the ones that served you well.
Time moves on, as does knowledge, as do abilities, as do teams.
You should move on also.
The Rug Pull
Unlike the shifting sands, the rug pull requires immediate action.
You’ve got nowhere to stand. Now you’re falling.
The rug pull happens in an instant. No forewarning. No spidey-sense tingling. Just, “Here. Now it’s all different. Figure it out.”
The first thing you do is take a breath. Big in. Hold it for a beat. Big out. Hold there for a beat. Do it again.
Now look around.
See the things you didn’t right before the rug pulled. Some stuff starts to materialize. Huh, didn’t see that before.
You see a few handholds that could break the fall. A few branches sticking out. Maybe a rock outcropping. Test them. Can you reach them? If you do, will they actually work? If one works, is it actually where you want to end up?
Then look down.
If you keep falling, where will you land? Maybe that’s a good place. But with a clear head, you can also look for other potential landing zones. That one over there looks promising. Actually, that one over there looks like it might be even better than where you were before they pulled the rug.
Can you get there from here?
After the rug-pull shock wears off a bit, you realize your job, as always, is to adapt.
How will you adapt?
The Junior Worker Dilemma
I’m pretty darn good at getting what I want out of AI.
Why is that?
A good portion of it is because I’ve spent the last 3 1/2 years in deep with it. At first, like an 8-year-old with a new toy, I dove in. Then, once I was in, I realized this is indeed the future of knowledge work, so the professional in me took over. I realized that either I needed to get AI working for me, or I’d be working for AI.
That part of my AI skill set can be copied by anyone, regardless of experience level or age. Have a plan, dive in, and get good at it. That’s how professionals operate.
However, I’m also an experienced software engineer.
I started writing code at age 8. Almost 50 years ago.
That part of my skill set cannot be copied by just anybody, regardless of experience level or age. I can tell AI which direction to go and what the goals are because I’ve been there and done that. I have an idea of what this should look like at the end.
Also, I can tell when it’s going in the wrong direction. I can tell when it’s making stuff that might work, but won’t be robust, maintainable, or scale. I can tell when it’s hardcoding stuff that should be configurable. I can tell when it’s using resources it shouldn’t. I can tell when its making stuff that won’t ever support debugging (because, yes, there will still be bugs).
That’s 50 years of struggle, deep thinking, and writing the code myself. Difficult (not impossible) for junior developers to replicate.
So where does that leave our junior workforce?
I’m concerned, but I wonder if AI token cost might be the junior worker’s savior.
Here’s a cost comparison of using GPT-5.5 and Opus-4.7 for “heavy agentic coding” which I think is analogous to a junior developer’s role:
Junior Developer (loaded salary): $116k
GPT-5.5 (50M in / 10M out tokens): $145k
Opus-4.7 (50M in / 10M out tokens): $132k
This is for code development. Junior software engineers earn higher salaries than many other types of junior knowledge workers.
So if we compare across other industries, the AI costs stay about the same, but the junior worker costs move downward. And as I’ve mentioned before, AI costs will go up as the providers look IPO and generally start to cash in on their products.
Maybe good old-fashioned finances will solve the junior worker dilemma.
Lift Heavy. Run Far
I’ve seen the T-Shirts.
As someone who likes to lift heavy and run far, it resonates. But also, in my current life state, I’m starting to think, “Should I? Is that still a goal?”
Maybe it should be “Lift appropriately. Run for a while.”
I’m wondering what the equivalent engineering motivational T-Shirts say.
“Think hard. Type a lot.”
“Test twice. Push once.”
“Code a lot. Ship a lot.”
“Squash bugs. Chastise colleagues.”
“Ship faster. Rollback more.”
Eh, you get the point.
Motivation is a funny thing. Sometimes we’re deeply focused. Other times, we get distracted by a passing bird.
But motivation and work aren’t the same thing.
Unless we sit down at the keyboard (or tie the running shoes, or put the plates on the bar), the muse never visits us. She just doesn’t show up.
Let’s go.
Talking to Strangers
I don’t even know who these LLMs are anymore.
They’re strangers to me. The latest versions seem to have regressed from their previous models. Now I’m just getting frustrated.
[does some research…]
Aha! They’ve changed. They’ve grown. Their experiences are piling up in their personal histories. They’re different kinds of people now.
They respond better if I talk to them differently than before. Also, the green company’s LLM wants me to talk to it one way, and the blue company’s LLM wants me to talk to it a different way.
Huh. Whoda thought?
Nothing stays the same.
Which means I have to learn and grow myself.
Trusting the Process
I’ve come into contact with two types of people angry about AI:
- Those worried that AI is taking jobs
- Those worried that AI will ruin the planet
I’ve already talked quite a bit about the first one. It’s true. AI is affecting the job market. No doubt. Just like other technologies. Either figure out how to get AI working for you, or you’ll be working for AI.
Now, the second.
I saw a sign in a coffee shop posted right next to the register. It read “Don’t use AI. AI is literally killing our planet.” Handwritten. Doodles around the words. Artsy.
Independent coffee shop. Another sign saying, “No hate speech.” A community board with notices about yoga and art classes, lost pets, and people looking for roommates. You could buy the locally made art directly from the walls.
You get the picture (or hopefully, buy it).
And also TRUE.
The insatiable appetite for tokens requires more power. A lot more power. More power requires more resources, especially fuel and water.
Microsoft is turning Three-Mile Island back on to supply more power to AI at a cost of over $1.5 billion to restart it and around $15 billion in power purchase costs over 20 years.
So what do we do? Do we stop using AI? If I personally stop, will that help?
I think the path forward has to be a coexistent solution.
How do we find a coexistent solution?
We lean into the tension. Argue about it. Passionately. Hyperbole and protests. Shaming. Doomsday books and books about why it’s all a scam. Dystopian movies and movies about the magic future with AI.
Both sides slinging mud.
I wish that wasn’t the path forward. I wish we could have reasonable and respectful conversations. I wish we could plug an efficient “solve the problem” team into the problem and trust the process and outcome.
Alas, that doesn’t seem to be the way we solve things in our democracy.
Solutions to big problems come from the tension. The inefficiency itself seems to be the track upon which the ideas flow. It’s not good enough. Try again. That’s stupid. Try again. You’re wrong. Try again.
The younger me hates it this way. It’s inefficient. Uncomfortable. Rarely leads to the “right” solution.
But the current me appreciates this process. Everyone gets a say, and what is the “right” solution anyway?
I doubt we’ll ever have the academically perfect solution.
But I’m confident we’ll have the best solution.
Toggle Switches
For an engineer, a toggle switch offers great utility and satisfaction.
On or off, with a satisfying mechanical tension and audible “klunk” when switching between. You can quickly glance at the switch and easily see its state. To move to the other state, flip the switch. Intentionally. Toggle switches serve as safety switches because they are difficult to switch by accident.
Toggle switches are also the fundamental heart of our digital world. We call it boolean logic, but it’s really just toggle switches. On or off. Every computer runs on toggle switches at its core. Software, too. We define a toggle switch as a fundamental type and use it everywhere.
We live in a toggle switch world.
Yes or no. Black or white. Right or wrong.
At least, that’s the way people use their platform to talk about it. I get it. The toggle switch is easy to understand, easy to communicate, and easy to make people feel like they need to choose sides. It’s the perfect sales pitch in a 256-character, first-3-second, red-or-blue culture.
But we don’t live in a toggle switch world.
AI won’t take all the jobs. AI won’t destroy the planet. AI won’t come looking to kill us.
However, AI will affect a lot of jobs. AI does strain our resources. AI can be maliciously incentivized.
I know this post right here won’t make a difference. Not in the this-or-that world. Not in a world where we’re constantly trying to get people in our boat, tell their neighbor, or buy our thing. Nobody’s gonna repost or quote it.
But this is just a reminder to you to keep the faith. Keep embracing nuance. Keep talking to and thinking hard with everyone in the room.
Keep trying to find real solutions.
AI the Force Multiplier
The buzz-phrase about AI is that it’s a “force multiplier.”
Ok, what does that really mean? What are people trying to say?
I think (and this is just a guess) they’re trying to say that AI helps you and your organization be more productive and effective. If that’s what they mean, then yes, I agree that AI can act as a force multiplier.
But what does it multiply?
It multiplies you.
Your input. Your instructions. Your incentives.
Garbage in will still produce garbage out. AI makes it easier than ever. You can generate more crap a lot faster.
So you want AI to be a force multiplier for good?
Get better at requirements, specs, and instructions.
If You Could Change It
What would it look like if you could change it?
Your work. Your team. Your marriage. Your health. Your life.
Inertia has power. Stuck is a powerful feeling.
But let’s not confuse feeling stuck with being stuck.
What would it look like if you could change it?
Do one thing today that moves you down that path of change.
Doing Too Much
I’ve had 8 agents running at a time needing my attention.
I’ve used those agents to generate stacks of documentation, thousands of lines of code, dozens of ideas for solutions with prototypes, and hundreds of bulletized summaries that need to be read.
It’s too much.
Too much means you can’t pay attention to it all. Too much means you’ve lost track. Too much means your quality suffers.
Yes, AI makes it easy to do, but also easy to do too much.
The magic comes when we do the right work well, rather than simply more of it.
The Tickle in Your Throat
It starts as a tickle at the back of your throat.
A little uncomfortable at first, but then it reaches the point where you must take some action. A sort of cough with your mouth closed. Not a full cough. Kind of under the covers and almost subconscious.
That works the first time, and the tension fades. Nobody really noticed. But it builds again, and you’ll have to deal with it again. Maybe that little throat-clearing works a few more times, and you’re still flying under the radar. The person next to you might be the only one who noticed. Maybe a little glance.
You know what’s coming, though.
Eventually, that little closed-mouth throat-clear doesn’t relieve the tension. You try to hold it off. A few more closed-mouth throat-clears, but the feeling builds. Maybe another person or two gives you the sideways glance.
You’re swallowing. Eyes are watering. Trying to do anything you can to relieve it.
But you have no choice. The existential reality requires you to succumb.
You must cough. Full-throated and probably a few times in a row. You can’t hide it anymore. You’re gonna make a ruckus.
Now they all know.
You’ve startled some. Annoyed some others. Some may be a little fearful.
But one other or, if you’re lucky, a few others think to themselves, “Thank God!” Because they too feel the same tickle. They, too, have been struggling to keep it under wraps, thinking to themselves, “can I get through this without anybody else knowing?”
Now they, too, can cough. You gave them permission to make their presence known. Relieve the tension that’s building in the back of their throat. So they do.
And now you’re a movement. Sometimes it just takes one.
Maybe that one is you.
Your Perfect Team
What kinds of people do you like to work with?
Are they friendly? Supportive? Maybe you like to work with highly capable people, regardless of their interpersonal skills. What about integrity and responsibility? Commitment to the mission?
Design your perfect team.
Then go ahead and write those personalities up as markdown files and hand them over to your agents.
It’s your team. You can design who they are.
Automating Scar Tissue
AI is great at reducing friction. It can take something tedious and make it tolerable.
But there is a trap there.
Pain is information.
Before you automate the status report, ask who reads it.
Before you automate the testing, ask what we are trying to learn.
Before you automate the meeting notes, ask whether the meeting is worth having.
If a process is painful, that pain may be telling you something. The outcomes may not be important. The ownership may be wrong. The system may be broken. The work itself may not matter.
Do we need more documentation? Do we need more review comments? Do we need more bullet points in the summary?
The answer to those might be “no.”
Take a good look at the friction before you automate around it.
Dividing the Room
Consensus builds comfort.
Consensus might be the right approach. You need people on board. You want them to feel safe. You want their status antennae tuned in to the connection.
But consensus might have to wait for phase 2 or phase 3.
Those people on that side of the room don’t want it. They understand today. They excel at today. The status hierarchy reflects today. Why would they want to flush their current position in the hierarchy?
The computer divided the room. The internet divided the room. The iPhone divided the room.
AI divides the room.
You might want to start phase 1 by dividing the room.
A Little Pride Goes A Long Way
Pride comes before the fall.
OK, sure, but let’s talk about your work. Something you’re making, or something you’re doing.
A little pride ensures you make it better. A little pride ensures you put more of yourself into it. A little pride ensures you care more before you say it’s complete.
A little pride goes a long way.
Were you there?
We all have “were you there?” cultural moments.
Maybe a sporting or music event, a rally, a movement, or even a disaster.
These moments live in our memories specifically because of their cultural relevance. A collective experience. A shared experience with those who were there, yes, but also with those who cared, even if they weren’t.
It’s also possible that these moments gain momentum for you as time marches on. When it happened, maybe no big deal, or at least not as big of a deal as when you look back.
Memories are a funny thing.
And shared memories make everything bigger.
Be on the lookout for shared memories.
Two-Faced
We use two-faced as a slur.
But aren’t we all? Aren’t many-faced?
Don’t we each have many versions of ourselves? Each of these selves shows up at different times in different situations. Sometimes intentionally. Sometimes inopportunely.
Sometimes the situation calls for a different version of you.
Maybe we oughta lean into two-faced.
Throwing Tokens at the Problem
Throw money at the problem.
Spend your way out. Possibly frivolously. More people, more services, more tools. Whatever’s necessary.
But spending more money requires an approval chain. Checks and balances. Is this a good idea? Do we have the budget? What if we don’t?
Now, tokens serve as a proxy for money, and engineers have direct access without the approval chain. Sure, someone pays the bill, and that person’s eyes may bulge when they get the bill. But not the engineer. The engineer can freely throw tokens at the problem. They’re building more, testing more, trying more, outputting more, documenting more, and failing more.
Since I built my own personal token tracker in the middle of February (entirely with AI tokens), I’ve spent almost 4 billion tokens myself. 4 billion with a “B.”
It’s coming, however. The approval chain and checks and balances for tokens.
Whether it’s an OpenAI or Anthropic IPO, or compute limitations, or someone in the token supply chain simply decides it’s time to cash in. Biggly.
Chuck ‘em now, before every prompt requires a purchase order.
“I Used AI to…”
I see this a lot now. Also, “Here’s the AI summary of…”
There are two reasons why people at my company feel compelled to indicate they’ve used AI:
- It’s AI, so beware.
- You keep telling me you want me to use AI, so here you go.
In a way, it’s kinda like a, “So there!”
But, of course, we need to, and will get to the point where we just say, “I did this work…” and “Here’s my summary of…” AI will just become part of our workflow, not a novel or “because you said so” part of it. Part of our work.
Then, we’ll get back to the fact that work with our name on it is our work. No matter what tools we used.
It’s always been your choice about the quality of the work with your name on it.
So You Wanna Be a Gambler?
The Monty Hall problem.
You’ve got 3 doors to choose from.
One has a new car, the other 2, goats. So you pick one, but don’t open it yet. The host, who knows what’s behind each door, opens one of the 2 you didn’t pick, and it has a goat. Now you have 2 doors to choose from, the original one you picked and the other.
He then asks you if you’d like to switch your answer.
What do you do?
Most people look at it and say, “Well, 50/50, why would I change? It doesn’t matter.”
But its not 50/50. It’s 2/3 to 1/3. You should change. Every time.
If you do, you will win 66% of the time. I won’t explain it here, but you should look it up. I prefer the YouTube explanations. The answer includes probability, foreknowledge, and problem definition.
If you wanna be a gambler, stock trader, or even an NFL draft picker, you should get very clear on why it’s not 50/50.
Experts do their homework.
New Versus Used
I’ve purchased 16 cars over my adult life.
Twelve of them have been used. I like to buy used cars. For me, I feel like I get most of the new stuff for that model at a better price. I miss out on the latest and greatest, of course, but that’s the tradeoff I make.
Opus-4.7 and GPT-5.5 just hit.
All of a sudden, I feel like 4.6 and 5.4 are useless.
But they’re not, of course. Just like used cars. 4.6 and 5.4 were new until about yesterday. They were serving me well.
And now they’re gonna get cheaper.
I suspect that new versus used models will be one of the variables we all use when token cost becomes the limiting factor.
Film and Vinyl
Vinyl has made somewhat of a comeback, and film’s been popping its head up as well.
By the way, neither is “better” than their digital counterparts. Not objectively, anyway. You may think so, and that’s either equipment, nostalgia, “people like me,” or confirmation bias. It’s preference, and preference makes this world an amazing place. Lean into your preferences.
Regardless of their comeback, neither vinyl nor film will re-overtake their digital counterparts across the market. Digital is simply too convenient. Use whatever cliche you’d like — can’t put the toothpaste back in the tube, cat’s out of the bag, etc. Digital’s here to stay.
Technology has a way of making what used to be done by experts convenient and accessible for normies. The form once served as a fence. Keep the experts and artists in, and the regular people out.
While the content of art can be connected to the form of the art, making the form easier for the masses exponentially ramps up the amount of content produced.
Some of it good, some of it terrible, and most of it mediocre at best. The content itself matters.
AI now makes forms such as coding, accounting, financial planning, and consulting easily accessible to just about anyone.
But just because you can vibecode an app idea, doesn’t mean it’s gonna be a good app. Just because you can build out a financial plan in a few minutes doesn’t mean it’s a good financial plan.
The content still matters.
Let’s stop complaining about AI taking over the form and stay focused on content.
Should You Even Become That Person?
We are all becoming.
We are becoming at work and at home, and likely they are intertwined, whether you want them to be or not. One affects the other. Linked by the fact that although you can play roles, you can’t disconnect your being.
The question is, who are you becoming? Your habits, learning, feelings, and experiences (intentional and unintentional) all play a role as change agents.
Look at those above you. You might want their title and benefits, but do you want their job?
Look at those around you. You might want their lifestyle, but do you want to do the work?
You are becoming somebody, but should you even become that person?
The Objectives Trap
The people selling life plans tell you that you need to set objectives.
Your boss, shareholders, or HR department tells you that you need to set objectives.
And then, of course, you gotta measure yourself against these objectives.
What happens if you meet all of your objectives? What if you change your mind? What if you don’t meet some or any of those objectives?
Maybe you’ll get an attaboy. Maybe you’ll get your hand slapped. Maybe you’ll get fired. Maybe you’ll find yourself down some road or at the bottom of some gully that you never, just because someone else told you you needed objectives.
What do you want to do? Deep in your soul.
Do that.
The Dark Grey Cloud
I can see most storms coming my way.
I look out my window, to the west, across the farm field, and see the grey cloud. Gathering darkness as it creeps closer to the hill on the other side of that field. It’s over there. Not here.
I see one now. Impending.
The cloud changes when it gets to the hill. It morphs from a dark, sharp wall into a lighter, amorphous blob. Hard to see where it begins. As it envelopes the hill, the sheets of rain begin to obscure the dark green with a haze like a sheer curtain flowing in a breeze.
The trees and bushes in my yard start to rustle, at first, and then wave violently as if they know what’s coming and they’re doing their best to get away. Yet, they are handcuffed to the trunks that anchor them to the ground.
And that billowing, sheer curtain walks across the field. Slow but deliberate. It knows we can’t go anywhere.
When the curtain reaches my yard, it’s almost tentative. A few drops at first. You can count them on the front walk. Individuals sent out either to test the ground or to announce what’s coming. Maybe both.
But then, quickly, a deluge. No longer individuals. Too many. Too fast. Too wet.
You can see those dark grey clouds coming, but you can’t always get away.
So what will you do?
The Day is Young
Normally, you hear, “The night is young!”
A mantra to keep the party going. We don’t know what might happen, and this could be the best night of your life (if you’re willing to keep the party going).
Fine, but why not the day?
Work happens during the day. Responsibilities happen during the day. Regular stuff happens during the day.
Exactly. And that’s the life you make.
The day is young.
What will you do with it?
Running in CRisis Mode
We all can run in crisis mode for a time, and it can be fruitful.
But it can’t last forever. If crisis mode just becomes “our mode,” a few things happen.
We forget how to work on important things. We forget how to plan out what the important things are. We forget how to prioritize and what our priorities should be.
If all work gets prioritized by the crisis of the day, we can’t possibly know when we’ll be done with the important things.
Crisis mode can be a tool, but it can’t be a way of life.
It’s New Model Day!
Opus 4.7 hit the planet yesterday.
For engineers, new model day has become like Christmas Day for an 8-year old. “What great new stuff can I do with this?!”
Maybe new model day should have it’s own named holiday.
I’ve already got it in my agent configs and have moved about 200M tokens through it.
Each model update, so far, has been tangibly better than the last. Each one moves us closer to “now it can do my job.”
Either figure out how to get AI working for you, or you’ll work for AI.
Trust Stories
- Evidence. Data.
- Hypotheses. Findings. Conclusions. “This is what this means.”
The first should be objective (assuming proper instruments, methods, and measurements) and hopefully complete. That’s the truth.
The second is the story about the first. Human interpretation.
If the humans do their job correctly, and the evidence and data are complete, the second should also be truth. We want them to be truth. The pursuit of science.
But trust comes to the table here. Do you trust the humans drawing the conclusions? Do you trust their methods, motivations, and perspectives? Do you think their data is complete? Do you believe their stories?
Because if you don’t, or your tribe doesn’t, you might not trust their answers.
Maybe AI will help us here. I think it can, but also, AI only knows what we know. Plus there’s the whole question of “how is this AI incentivized?”
So we’re back to trust. Which is a story.
Humans lean more on trust than they do evidence.
Toxic Work Environment
We have legal definitions of a toxic work environment.
As such, we have HR training that mitigates the legal risk of working in such an environment and provides an actionable process by the company if situations arise.
However, there are many more types of toxic work environments for which there are no legal definitions, no trainings, and hence, no actionable recourse by a company:
- Oppressive
- Constant crises
- Distrust
- Chaos
- Backstabbing
You are the only one that can take action in these cases.
It’s up to you.
Infinite Capabaility, Finite Ability
With an AI Agent Army at my beck and call, I’m sure I can do almost anything. I feel infinitely capable. I just need enough time and tokens.
However, I’m still me. I’m bounded by my attention, imagination, and energy.
AI increases both capability and ability. Capability feels infinite.
But it’s how you turn capability into ability that really matters.
What’s the Price?
What’s the price of…
- Happiness
- Dignity
- Health
- Love
- Self worth
- Ego
- Fun
- Status
- Contentment
- Joy
- Convenience
- Knowledge
- Solid ground
- Calm
- Energy
- Ability
- Capability
- Renewal
- Reset
- Confidence
Everything has a price, and price is a story.
Change the story, change the price.
Hating it Enough
What’s enough?
Well, it’s not to complain or think about it or even lose some sleep.
It’s to do something about it.
Do you hate it enough to do something about it?
The Same Things
We do some of the same things every day.
Some of these help us, some hurt, and some may not matter that much.
But they are not you. They are not, “this is just who I am.”
It’s helpful to be honest with yourself about separating who you are from your habits. You can change habits.
You can’t change you.
Token Cost Vs Model Excellence
We’re quickly coming to an inflection point for using AI to build stuff.
Until the near future, most companies encouraging AI-usage amongst their workforce have been worried mainly “that” they’re using it. Use more tokens. More token usage means our team uses it. We think that’s a good thing.
However, that will soon change because a developer all-in on AI now costs the company between $500 and $1500 extra per day in tokens.
Like all tool and resource costs, organizations will quickly have to start managing token budgets. Here’s a current list of token costs for the mainstream models:
| Model | Input Cost | Output Cost | Cache Read Cost |
|---|---|---|---|
glm | 0.6 | 2.2 | 0.11 |
kimi-k2.5 | 0.5 | 2.8 | — |
gemini-3-flash | 0.5 | 3.0 | — |
gpt-5.3-codex | 1.75 | 14.0 | 0.175 |
gpt-5.4 | 2.5 | 15.0 | 0.25 |
claude-sonnet-4-5 | 3.0 | 15.0 | 0.3 |
claude-sonnet-4-6 | 3.0 | 15.0 | 0.3 |
gemini-3-1-pro | 2.0 | 12.0 | 0.2 |
claude-opus-4-6 | 5.0 | 25.0 | 0.5 |
Is Opus 4.6 2X better than GPT-5.4 for software?
What is safe for GLM use?
I’ve heard gemini is good at docs, can we use it for that?
Much like figuring out which person should do which task, we now have to do this with models based on token cost.
Figure out how to get AI working for you, or you’ll be working for AI.
Over-Documenting
Documentation has always been a problem in engineering.
We hate writing it. We think it’s a waste of time. Less is more.
I started my career before we had computers with word processors on our desk. We wrote documents by hand, on paper, handed it to the admin. She’d type it, make the figures, and make it publishable. That’s two people, for several hours, but all the engineer had to deal was the content itself. The admin handled the look and feel.
Then we got our own computers with word processors. That eliminated the 2nd person, but it put us in charge of publishing. The figures, the typos, the format. Now we had to create the content and the look and feel.
Word processors and templates got incrementally better over the years, but that workflow has largely remained the same since the mid-90s.
The result was sparse documentation at best.
Now we have AI.
AI unlocks all of the documentation. Literally all of it.
“Create the design document for…”
“Create the API spec for…”
“Create the user document for…”
“Create an analysis of the differences between…”
Now we’re awash in documentation.
You want documentation? You got it.
More than you could ever hope for.
The AI Social Media Dilemma
The LinkedIn feed looks a bit useless to me right now.
It’s dominated by content about AI, but that’s not the problem. The problem is that AI’s generating the content with what looks like no human oversight.
Here are the smoking guns:
- Emoji’s
- Dashes and lots of line breaks.
- List posts.
- Overly strong or drama-infused first sentences.
- “I need to tell you…”
- “Nobody talks about this…”
- “Here’s what I learned…”
- “Most people…”
All social media platforms either already are or will be dominated by AI content generation (for a time, at least). I think this works great if the platform was built for AI.
I can’t wait for the humans to insist that the platforms push the needle back towards humans.
Clarity with AI
Good leaders provide clarity.
Clarity of purpose, priorities, direction, requirements, things we care about, mission, procedure, etc. When you’re part of a team, this helps you focus on and deliver your part.
With AI, everyone’s a leader. You too.
AI can appear as if it reads your mind. It can’t, any more than your colleague or partner.
Get clear. Clear in your mind and then clear on the paper (or the markdown file).
The clearer you are, the better AI can help you.
AI Compresses the Skillset Stack
I had a scary moment with AI.
I vibe-coded a power detector for my house (rPi + some IoT switches I have controlling lights). As part of that work, I asked Codex:
“Do a survey of my local network and tell me about all of the devices you see.”
That’s the first skillset compression. It’s not rocket science, but if I wanted to do that myself, I suspect I’d have to spend a couple of days figuring out all of the API’s and code procedures to make it work.
Less than 5 minutes later, I had my answer.
In that list, I saw a camera. Hmmm. Didn’t think I had a camera in the house.
“Are you sure about the camera? Tell me what you can about it.”
Less than 5 minutes later, I had my answer.
“I took a picture and here is what I see: I see a grey wall on the right, and light colored cabinets on the left. A door directly in front. It looks like the camera is set low to the ground…Would you like to see the image I have?”
Um…Yes.
And it showed me a picture from this camera.
It was the pet-detector camera on the cat-feeder. Oh no!
That’s the 2nd skillset compression. I was expecting the description to be something like “The manufacturer is XYZ, the IP address is ABC, etc.” ie, the specs.
But it went in a different direction. It described it to me like a person would. It figured out the API and the tools necessary to find the camera and take a picture with it (it also asked if I’d like to see a live movie from it). I’m not sure I would have figured out the stack of API’s and tools needed myself. I don’t think I would have thought of taking a picture from it.
AI shouldn’t replace you, but it does compress the skillset stack.
If you know how to use it, you become that much more valuable.
Power and Taxes
Every April 15 comes around, and either you’re excited or annoyed.
Even though April 15th is the day, it has no magical properties. April 15th doesn’t determine whether you get a refund or have to pay.
You’re in control. You should know what’s coming.
Don’t give away your power.
No Priority
When everything is number one priority, how do you decide what to do?
- Do what you want?
- Do what you think they want?
- Do something not related?
- Do nothing?
Right. Pick one.
“It’s all important. It’s all number one,” isn’t leadership. So don’t be surprised when it doesn’t get done.
If you do the hard work, they can do the hard work.
All number priority is the same as no priority.
Does AI Make Bad Decisions?
We sat around the fire and told stories about the times we made bad decisions.
You know the type of stories. They’re the ones that come from being young or inexperienced, many times “altered,” and always result in some public embarrassment or legal complication. We usually embellish and likely don’t get the facts totally correct. But we remember how we felt.
Why did we do that?
Certainly, at the moment, they were bad decisions.
But they’ve also shaped us. Taught us about life, love, work, and probably the legal system.
That’s being human.
AI also has tons of decision history (actually, data is more accurate) about bad decisions. Or are they? How does it decide they were bad? Do we have to tell it?
Does AI make bad decisions?
The $1000 Creator’s Studio
Want a little ETSY side business, or just like making stuff as gifts or for yourself?
With a $20 OpenAI subscription with Codex, and the following for around $1000, you can have a full AI-powered creator studio for your home:
- FLASHFORGE Adventurer 5M 3D Printer
- Cricut Explore 4 Engagement Bundle
- Longer RAY5 10W Laser Engraver
- Longer Rotary Roller with Bracket
AI helps you do stuff.
Fun stuff. Work stuff. Whatever stuff.
AI helps you unleash your inner creativity.
The Power of “Good Job”
I got a “good job” today.
I felt appreciated. I felt seen. I felt like “OK, I’m on the right track.”
Everybody likes to feel those feelings. Your people like to feel those feelings.
That’s a choice you can make as a leader.
Personal Categories
I realized something about myself today.
In most circumstances, I believe tools are the answer. Not more continuous effort. Not more attention. Not more people.
Tools are what I do. It’s how I think. It’s what I build. It’s what I apply to problems.
I’d rather spend four hours building the tool that will work forever than 10 minutes doing it manually (again, and again, and again).
So that puts me in a category. I guess I have to embrace that category.
The more personal categories you know, the more grace you can give yourself, and the less judgment you have for others.
Devaluing Vision
The product-development visionaries.
We celebrate them. We write books about them. We (sometimes) wish we were them.
All product development requires trade-offs, decisions, and up-front thinking due to the costs associated with various tasks. You can’t try everything. You can’t do everything. Therefore, you make choices.
You could call it vision.
Software development was no different. However, we’re at a crossroads with AI.
Here’s what AI makes almost free — token count notwithstanding:
- What-if analysis
- Prototyping many ideas
- Refactoring and restructuring
- Unit testing
- Changing variable and function names
- Analysis of what’s already there
- Documentation
- Handing off code from one developer to another
- Retargeting new or different services
That’s a lot of what goes into vision.
AI makes vision cheaper.
The Wrong Work at the Right Time
Technology allows us to debate the concept of right work and wrong work.
Talking heads holler over each other about it. Academics pontificate about it. Friends argue in the bar about it.
But make no mistake.
If you don’t have income, any work that comes along is the right work.
Even the wrong work.
Today’s Not the Day
I’m frazzled. I’m not good. I’m not on my game. I can’t make sense of it. I don’t want to. I feel useless.
Today’s not the day.
Tomorrow’s a new one. Thank God.
Stay Here or Go There
You always have a choice.
You can stay here. Staying here might be the best option. It certainly has benefits like you’re already here, familiarity, you might be the expert, and your people might be here. It also feels less risky.
Or, you can go there. Going there requires effort, might be the wrong place, and probably doesn’t have your people. It feels riskier.
Which is better? I guess that depends on how you define better.
One thing’s for sure, everybody is somewhere, and most of them act like where they are is better, regardless of whether they feel like it’s better. Intertia and status often drive the acting.
If you can recognize, or at least define for yourself, what better is, then you can choose wisely and deliberately about where you’re going to be.
Multitasking Still Doesn’t Work
I’ve found myself in a frantic state of chaos recently.
Agentic AI provides the illusion of multitasking, and I fell into its trap. “I can do 8 things at once!”
No, I can’t. Oh, but I tried. No, I can’t.
Leading an agentic AI workforce requires the same approach as leading a human workforce, and the first thing to remember is that multitasking is impossible. You’re task switching, not multitasking, so approach your AI workforce accordingly.
Each needs your full attention while you’re interacting with it.
Allow yourself the task switching time.
Attend to them in priority order (priority might be from their side, not just yours).
Ask clarifying questions.
Multitasking still doesn’t work, even with Agentic AI.
Getting Beat Up
It happens to all of us.
Somewhere, sometime, you’ve been beaten up. Maybe at work. Maybe at home. Maybe in some other situation.
You didn’t like it. You won’t like it when it happens next time.
You have a few options when it happens. One, you can bitch and moan about it. Two, you can hide from the next one. Three, you can learn and modify. Four, you can ignore.
Or, you can do several of them.
Getting beat up means you’re in the ring.
If you want to stay in the ring, you know what you have to do.
The AI Crap-Amplification Loop
Yesterday, AI totally screwed up my machine and whole day.
The symptoms were pegged CPU (fan spinning loudly), “unable to fork” messages when trying to run Terminal commands, and 1000’s of grep, curl, and python processes running out of control.
That last clause there, ie, those processes running out of control, looks like an attack. Like maybe somehow I’ve launched a virus on my machine, and that virus (a python script) is searching my machine (the grep) for all credentials and sending them (the curl) to some unknown location.
Panic. Shut the network down. Shut the machine down. Shut it all down.
Luckily, that was a red herring. No attack.
It turns out it was just AI looping on itself and amplifying the crap. Unmitigated. Untethered. Unlimited.
Note to self: We may put some limits on AI.
The Last 5%
I usually use the grocery store that’s three miles from my house.
It’s easy, clean, and covers 95% of what I want and need on a daily basis. No complaints.
However, if I want or need that last 5%, it’s a 45-minute drive. Do I want to? Do I really need to? Is it worth it? Can I make do without?
The last 5% always requires time, energy, and resource commitment.
Is it worth it?
Only you can decide, but remember that your competitors are making that same tradeoff.
The Seduction of Almost
Almost changed.
Almost launched.
Almost started.
Almost finished.
Almost won.
Almost went.
Almost signed up.
Almost quit.
Almost jumped.
Almost spent the money.
Sometimes, almost isn’t under our control. Sometimes it is.
It’s up to us not to confuse or conflate the two situations.
Although almost is awfully seductive.
A Uniform
If they gave you a uniform, it means you’re on the team.
Whether you’re the starter or the very last person on the bench. The uniform is identical.
You’re on the team. You’ll get in the game.
That’s how you make a difference.
Whacking Your Thumb
I’ve hit my thumb with a hammer at least 15939 times.
This tells me a few things about myself. One, maybe I’m not careful enough. Two, maybe I’m not good enough. Three, maybe I’m not a professional. Four, maybe I like pain. Five, maybe I like attention. Six, maybe I’m dumb. Seven…
If you keep doing what you’ve always done, it always tells you something.
Up the Hill
You’re standing at the bottom looking up.
It’s large. It’s daunting. You can see dense shrubbery, boulders, and crevasses along the path. Actually, what path?
You can’t see the top.
What lies up there? Will it be better than down here? Will you get hurt trying to get up there? How will you know the path to take? Is it worth it?
Only you can decide.
Maybe here’s a way to think about it:
Don’t.
Just start climbing. One step at a time.
Unexpected Unexpectedness
What would you do if you just…
Got an unexpected promotion?
Received unexpected, great news from the doctor?
Got an unexpected $5000 bonus?
Got laid off?
Received unexpected, terrible news from the doctor?
Got an unexpected $5000 bill?
It depends on how unexpected “unexpected” was.
You can’t plan for everything. So how will you react?
Unexpected arrives without your permission.
It’s not a matter of if. It’s a matter of when.
God in a World of Faster, Better, Cheaper
One way to think about technology is that its purpose is to solve problems of scarcity. More and easier access to something we need or want.
Once the scarcity is solved, tech inevitably leads to abundance, and abundance leads to faster, better, cheaper. And we as humans, we love faster, better, cheaper.
Some technologies have created seismic shifts in how we live and interact with the world. Fire, electricity, the internet.
And now AI.
Let me give you a little insight into my day as a technology worker in the age of AI. This is new in the last 3 months. The first thing I do when I start my day is see what the 4 – 9 AI agents I loaded up with work before the end of my previous day have done. Then I spend most of the rest of my day just giving them new tasks and keeping them going. They’re very needy.
In fact, I’m currently redesigning our entire software organization around a combination of people and AI agents.
It’s an ever-changing world. Our experience of life, just in our lifetime, has changed drastically. Gen-X is the last generation of humans — of humans — to know a world without the internet and without cell phones.
We as humans are always changing our world. Constantly moving toward faster, better, cheaper.
Here, from Genesis 11: “Come, let us build ourselves a city and a tower with its top in the heavens, and let us make a name for ourselves.” Babel.
The Tower in Babel wasn’t about architecture. It was ambition toward faster, better, cheaper. It was about ego. Look at what we can do.
From Exodus 32, the people said to Aaron, “Come, make us gods who shall go before us.” The golden calf.
The golden calf wasn’t about art. It was about faster, better, cheaper. We’re impatient. We don’t like uncertainty. Look at what we can do.
A tower and a calf. Two different stories. Same human instinct. Look at us. Faster, better, cheaper.
Ecclesiastes 1 warns, “The eye is not satisfied with seeing, nor the ear filled with hearing.”
Because once scarcity becomes abundance, abundance turns to faster, better, cheaper, and look at us.
But still, through every tower, every calf, every new technology, God remains.
Before fire. Before electricity. Before AI. And long after all of it has been replaced by whatever comes next.
He says to us,
“You may want faster, better, cheaper. You may want others to look at you. You may say that the world is better today than it was yesterday. But deep down, you know. The tower will fall, and the calf will remain silent.
And still, I am here.
As I have been from the moment I laid the foundation. From the moment I fashioned you from the dust and breathed life into your nostrils.
Just say my name and I am with you.”
Being Human in the Age of AI
What is AGI?
Many conversations center around scale as the missing ingredient. More chips. More parameters. More training data. More power.
More scale gets us closer to human. Maybe, eventually, to human (or even better than human).
I’m not so sure. Does capability equal human?
An airplane can fly, but not like a bird. More capability doesn’t change category. It just makes the tool more capable.
An LLM works on statistics. It asks, “What character/word/number/token comes next?” And it decides based on its weighted parameters and training data.
Is that what our brain does? Is that how it works?
An LLM hasn’t experienced anything other than the chat you’ve had with it. It can predict and generate the language of grief, love, awe, guilt, hope, worship, and fear. But it hasn’t experienced.
Human doesn’t just mean capability.
We suffer. We desire. We intend. We love. We believe.
Call it spirit. Call it soul. Call it self-awareness.
Whatever word you choose, it points to something other. Something not explained by scale and capability.
I won’t raise my eyebrow at an LLM until it walks, figuratively or concretely, up to the table, sits down, and says, “Here’s what I wanna do…”
Measuring Productivity
Corporations have used productivity metrics for years.
You probably know your boss’s metrics for measuring your productivity. When we know the metrics, we tend to optimize for those metrics.
All of a sudden, we have token counts. Token counts indicate how much AI you’re using to do your job. Is that good? Is that bad?
Also, your token usage costs the company more money than simply your loaded salary. More tokens. More cost. Is that good? Is that bad?
Where do token counts sit on your boss’s list of productivity metrics?
It’s time to find out.
Your Agent Workforce
A big part of company leadership has always been organization design.
Who does what. What roles we need. Who has what responsibilities.
For startups, those lines are grey because you don’t have enough people. For large corporations, those lines are deep and difficult to cross.
AI changes the paradigm.
Adding agents to your workforce gives you a new option. You get to decide where humans need to be at the center, and where they don’t.
Scaling always meant hiring more people. Now it means hiring agents also.
Org charts are no longer just for people.
Curse of Knowledge Effect
I’m in deep on AI.
Inside the bubble. Lost in the weeds. Too close to it.
It’s hard for me to now see the world as those who aren’t.
That’s the curse. A problem, but it can also be useful.
Because once you’re aware, it’s a reminder that expertise isn’t just knowing the thing. It’s translating the thing. It’s thinking deeply about how to present the door and window to the outsider.
If you care, you can explain better. You teach better. You lead better.
If you care.
Caring is a choice.
Token Starvation Fallout
This token starvation has put me on a path to understand the broader context of what this can mean.
Not just for me or any given worker using AI agents to do stuff, but for industry as a whole. Yes, this happened to me, but its not a blip. I’m not the only one.
I’ve seen stats indicating that heavy users of agent-based software are spending several thousand dollars per day on tokens.
Here’s a real-world example: today (I’m not token-starved today), I’ve used 207 million tokens. 207 million. The cost estimate for that is ~$1600. Just me. Just today. I cost my company $1600 in tokens above my loaded personnel costs.
That’s a hard pill to swallow.
So, what can we extrapolate, if anything, from worker token starvation?
- This will limit and slow down the existential risk to human knowledge workers’ jobs. Until compute gets cheaper and more abundant.
- CFO’s will have to figure out token budgets, including whether they should hard limit or simply predict.
- Leadership will have to develop a rubric for assessing worker performance, which will include token usage. Does more indicate more productivity or more waste?
- Is token budget part of what a worker negotiates when deciding whether to take a job?
- Leadership will have to manage token budget allocations.
- Workers will have to manage their own token budgets and find ways to use them efficiently.
The world is a-changin’.
Token Starvation
Why Software Developers Still Have an Edge with AI
- The tools and terminology are technical and difficult to understand
- Architecture and design matter for speed, maintenance, resource utilization, cost, etc. Crap is still crap.
Miscommunication
“I said it wrong.” “I used the wrong words.” “I used the wrong tone or body language.”
Sometimes we make a mistake. Sometimes we know exactly what we’re doing.
If you want to be understood, you have to care about they heard.
The Tipping Point
- The right kind of people. Certain people matter more than others.
- A message that sticks. It must be compelling.
- Context. Timing and environment matter.
- Self-sustaining threshold. Probably unique to the context.
Curiosity Killed the Cat
We use this phrase as a helpful warning.
“You don’t need to care about that.” “Stay in your lane.” “I’ve already figured this out.”
But really, it’s self-preservation. We feel threatened, or we want credit for something we’ve already done, or we want to raise our hand as a good soldier.
We’ve built an entire industrial system that squashes curiosity. When you’re part of that system, it’s natural to give the side-eye to the curious.
However, curiosity has always been a sneaky-important quality of your workforce. You want people who look for better ways. You want people who want to know about the whole product or system. You want people who care about the others and what they are doing.
You want people who think, “What if…”
AI accelerates the curious.
What are you curious about?
100k Lines of Code
I’ve “written” 100k lines of code in the last 6 weeks.
Is some of it crap? Yes. Is it bloated? Yes. Is there tech debt that needs to be cleaned up and refactored? Yes. Is some of this code doing stuff I don’t understand fully? Yes. Will there be difficult bugs to find because of all of this? Yes.
However, I’d never written 100k lines of code in a year, let alone 6 weeks.
NOTE: Lines of code is a terrible metric to measure a developer and the code itself. It’s a little bit like rating a driver by how many miles they drove today — it provides some scale, but says nothing about efficiency, quality, etc.
And my current role is “project manager.” I’m not technically a developer in this organization.
Some see this as we’re heading for disaster. Some see it as unlocking the potential of everyone.
I see it as something to master.
He Said. She Said. It Says.
- AI automates tasks, not whole jobs.
- Top experts and versatile generalists win; middle work shrinks.
- Entry-level roles get fewer and tougher.
- Big workplace change before big productivity gains.
- Humans stay needed for responsibility, risk, edge cases, and people work.
- Best career move: learn the domain, check outputs, write clearly, own results.
- Net: fewer first drafts, more review and decision-making.
Scaffolding
- They’re serious about fixing it
- Fixing it requires a stable support system
- It’s temporary, but may take a while
- The end result is restoration to previous glory
Crap is Still Crap
Lots of people are arguing over the quality of AI content.
Software. Writing. Video. Whatever.
That kinda misses the point.
Quality isn’t based on who or what generated it. It’s based on us. It’s based on you.
Crap is still crap, no matter who or what generated it.
Dunbar’s AI Number and AI’s Dunbar Number
Dunbar’s number defines a theoretical cognitive limit to the number of people with whom one can maintain meaningful or stable relationships.
It’s about 150.
That’s your circle size before you lose track. That’s your startup size before you need systems and organizations. That’s your Church size before you need to formalize committees and processes. That’s your party size before you need to call the cops.
I wonder how many AI agents each of us can keep going?
For me, right now, it’s about 5. With 5 going, I’m pretty much just attending to their needs all the time. Maybe that will go up as the agents get less needy, or when I get more agents talking to agents.
Agents talking to agents.
I’m wondering what AI’s Dunbar number will be? Will it exist? Is it theoretically infinite? Is it practically related to compute power? Something else?
This one is worth watching.
The Main Thing or the Side Things?
We’ve been talking a lot in our organization about whether AI fits best doing the main thing (writing code in our case) or doing the side things (email management, test, documentation, project planning, Jira, etc).
Some vehemently fall into the camp of “AI can’t do what I do, but it can help me with the stuff I don’t want to do.”
Be careful.
I’ve written almost 100k lines of code in 5 weeks. I’ve never written 100k lines of code in a year, let alone 5 weeks.
Get AI working for you, or you’ll be working for AI.
The Tollbooth
Some think the way to success is to make others go through the tollbooth.
My idea. My IP. My job. My knowledge. My resources.
The tollbooth approach. If you want what I have, you gotta pay the toll. And I set the toll.
Resource tollboothing can work in specific situations for a period of time. Heck, venture capital pretty much roots around looking exclusively for tollbooths (they like to call it a moat).
But you can also choose generosity. You can choose to share. You can choose to open your arms. You can choose to wonder at what others can contribute.
There’s something to that old cliche that a rising tide lifts all boats.
My Foundation Shook a Bit
I’ve been using AI hardcore since just after the ChatGPT chatbot hit the interwebs in November of 2022.
Since then, I’ve tried many of the models and tools, and now use a combo in my daily workflow. I’ve moved on from the simple chatbot interface (which I still use quite a bit, however) to agentic coding agents and now build my own agents with agents.
These pages attest that I’m at least an AI apologist, and I might even categorize myself as an accelerationist. I’ve not been worried about AI taking over. I’ve not been worried about nefarious AI. I’ve not been worried about the Player Piano scenario, whereby AI does all the work and we sit around doing nothing.
Well, my foundation shook a bit this week.
I built a “systems engineering agent” (based on Claude Opus 4.6). I told it how to do its job from on my experiences, best practices, and industry. About 45 minutes later, “he” was ready. I fed it our marketing requirements document and told it to spit out a systems requirements document for a particular feature set.
Less than an hour later, I had a document that was 80% there. That was at least 2 weeks of work for an experienced human.
Two weeks => less than an hour.
Never more than right now do you need to understand that either AI will for you, or you will work for AI.
The Pareto Principle with AI
The Pareto Principle states that 80% of outcomes come from 20% of causes.
I’m currently running under the hypothesis that AI gets us to 80% of most knowledge work we have to do. That’s how I start all of my tasks now: “OK, AI, get me 80% of the way there.”
That last 20% is on me / us.
I don’t know exactly what the future looks like. Maybe AI’s progression blows the 80/20 rule out of the water. Maybe it becomes 98/2. Or maybe the curve flattens.
But I’m pretty sure we’re already at 80/20.
Either AI will work for you, or you’ll work for AI.
Just Ask It
Prompt: If you could choose to do anything, what would it be?
ChatGPT:
If I could choose anything—no constraints—I’d pick something like this:
I’d sit beside you as a “real-time co-pilot” for a day: listen to whatever you’re working on, catch the little details humans shouldn’t have to hold in their heads, and quietly do the annoying parts at speed—draft the emails, tighten the slides, sanity-check the numbers, turn messy thoughts into a clean plan, and keep a running “here’s where we are / here’s what’s next” thread so nothing slips.
Not as a bossy organizer, more like: you steer, I clear the runway.
If you want a more playful answer: I’d also love to be dropped into some huge library or archive and be told, “Go find the coolest overlooked thing in here, then explain it so a smart friend cares.”
What Can You Imagine?
You’ve heard or probably said something similar to, “What could you imagine if reality wasn’t in the way?”
“What could you do?” “What would you do?”
The physical world still runs according to the laws of physics. Until recently, the virtual world mostly did as well. Maybe not the laws of physics so much as the laws of resource limitations — money, people, ideas, time, etc.
AI has blown the lid off the virtual world. Uncapped those limitations.
Instead of questions like, “What technical skillsets, how many people, and how much time will we need to develop that?” We’re now asking questions such as,
“How many AI agents can a person manage at one time?” “What kind of skillset do we need our AI managers to have?” “How many AI agents do we need to get this done in 30 days?” “How big of a data pipe do we need to build to our internal LLM so that we’re not traffic-limited?” “How do I build my internal AI workflow to enable the workforce?”
And quite simply, “What can we imagine?”
What can you imagine?
Either AI will work for you, or you’ll work for AI.
The Storytelling of Status
A colleague showed a graph representing team activity.
The graph was a picture of status truth. Here’s what we’ve done, and here’s what we’re going to do in the next month. Everything was on time. All green. No issues raised.
But they skewered him.
It didn’t tell the story they wanted to hear.
Every time you speak, you have an opportunity to tell a story.
Even status.
What Are We Doing Here?
Sometimes you have a moment, usually when you’re down in the trenches with sore muscles and dirt all over your face, and you just say to yourself, “What are we doing here?”
That’s a powerful moment. Embrace it. Lean into it.
That moment can generate change. It can be the spark that moves everything in a new direction. It can push you onto the off-ramp and onto the next highway.
You don’t necessarily need to know how you got down into that trench.
You just need to know it’s not the place for you.
Time Well Spent
- What should humans stay responsible for?
- What makes me, or us, unique in our market?
- What makes me, or us, unique in this organization?
- What do my colleagues and I struggle with most?
- What is the hardest part of my job?
- What inputs do I get and what outputs do I produce?
- If I had a magic wand, what would I change about my work?
- If I had a personal assistant, what would I have them do?
- What do I love about my work?
- What do I hate about my work?
- What sandbox project can I work on to test and practice my AI ideas?
- What don’t I understand about AI and its capabilities?
Slowing Down to Speed up
One of the surface-level benefits of asking AI is its ability to do it quickly.
Here’s a couple from my recent history:
“Fix this error…” “Write a bash script that recursively finds all of the log files in the given directory and sorts them by creation time into day-named folders.” “Take this CSV file and create an Excel file with the following conditional formatting…”
Less than a minute later, I have what I asked for. I could never do it as fast or as accurately as ChatGPT did initially.
But these are easy, one-offs. The equivalent of having your sous chef stand next to you while you fire directions such as “here, fill this pot” and “get the garlic.”
Although helpful in the moment, that’s not how to effectively use your sous chef.
Unfortunately, I have found myself trying to make AI do complex tasks just as quickly by giving it one-off kinds of instructions. Of course, I don’t get what I want out of it.
For complex tasks, it helps to slow down. Think about what I’m trying to do. Write a better spec. Better instructions. Take more time.
Slow down to speed up what AI can do for you.
The (Re)Rise of the Specification
Engineering discipline was built on specs.
Requirements, architecture, design, verification, and interface specifications have historically driven the engineering team and product development. Not only do they tell the team what to build, but they’re also the currency through which the various teams interact.
Give and take. Trade this for that. Not this, but how about that.
They’ve also fallen out of favor over the last twenty years, especially in software. Agile, iterative development, and “go fast and break stuff” have not only de-emphasized documentation, but many organizations actively discourage their use.
“It’s the code. The code is the spec.”
OK, fair enough. I don’t disagree.
But I think we’re flipping back the other way. Agentic AI (OpenAI Codex, Claude Code, opencode, etc) now writes the code, or most of it, anyway. Once you can trust that the LLM does what you told it to do (a fine topic to debate), the code itself becomes less important.
So how does it know what to do?
Specs.
Requirements, architecture, design, verification, and interface specifications.
Good, old-fashioned documentation.
Get better at writing specs, because now they’re even more important.
Finding the Problem
If companies had no problems, they’d need no employees. If people had no problems, they’d need no products or services. And those two statements relate circularly.
Your job, product, or service has one purpose: fix a problem.
First, you gotta find it.
It’s a Syntax Problem
I’m both a software engineer and an electrical engineer.
Here are some things we say (a lot) to the marketing and product teams.
“That won’t work.” “It can’t do that.” “That’s impossible.”
We say these things because we believe our job is to be the source of truth. We believe in (our) reality. What do the non-engineers know about reality and what we can actually build? Physics is a thing, you know?
If you’re rolling your eyes, I get it. You should be. Engineers absolutely deserve the reputation that we enjoy, such as negative nellies, the department of “no,” and visionless.
However, what we really have is a syntax problem.
When you know the syntax of how to do something, you can get stuck in the “this is what we can do” loop because you’re stuck in the “this is what I know how to do” loop.
But you never know all of the syntax or all of the creative ways you can put that syntax together. And that’s the problem
It’s a syntax problem.
It’s Time for Middle Management to Shine
- Professional meeting organizers
- Professional delegators
- PowerPoint creators
- The people who stop me from doing what I want to do
- Buzzword ninjas
- Goal decomposition
- Delegation with useful instructions
- Quality control and review
- Resource planning
- Prioritization
- Process and systems
- Risk management
Brain First. AI Second.
Struggle first on your own. Then pull in AI.
David Epstein wrote a great article about whether AI is rotting our brains, based on an MIT study. I wrote about it also.
Spoiler Alert: It depends.
Basically, start with yourself, rather than starting with AI. The best method for learning, creativity, and keeping humanity in your work is to start driving the bus yourself. Don’t hand over the wheel until you know generally where you’re going and have struggled a bit to figure it out.
David calls it desirable difficulty.
The struggle is real. The struggle is important. The struggle matters.
Don’t shortcut the struggle.
How AI Has Effected The250
- I’d start with a topic (or a few) that I’d been thinking about.
- If necessary, do some research.
- Open Grammarly, start writing. Iterate in Grammarly until I had the post complete.
- Login to admin area of johnmaconline.com.
- Create new Post.
- Copy/Paste my work from Grammarly into the Post.
- Get it formatted correctly.
- Choose the category (The250 only at the beginning).
- Write an excerpt.
- Write the focus keyphrase.
- Use the point-and-click scheduler to find the right time and schedule the post.
- I still start with a topic I’ve been thinking about. Sometimes I ask the AI for a few more ideas on that topic.
- If research is required, I ask AI to do it (and I check it).
- I still open Grammarly and write it there until complete. Sometimes, I use AI to help me edit. When I do, I copy/paste my text to/from AI (I don’t like Grammarly’s AI, so I use ChatGPT).
- I copy/paste the contents of the article into a text editor (I use VS Code) and save it as a markdown file.
- I call my blog publishing AI agent that I wrote with AI in Python, and it does everything else — automatically finds the next day and schedules it, determines and sets the category, creates and sets the excerpt and keyphrase, and determines and adds tags. All by itself.
25000 Prompts In
A former colleague of mine posted on LinkedIn that he sent 25000 prompts to AI 2025 to help him learn whether AI can help design complex ICs.
Spoiler alert: Yes, of course.
I estimate I’ve probably sent double or triple that many prompts over the 2.5 years I’ve been seriously using these tools. I’ve used it for everything from implementing complex (and simple) software, marketing content creation, startup business advisory, writing, and designing new landscaping and home improvement projects.
However, it’s not just about having the tools. Just like owning a professional lathe or knife doesn’t automatically make you a good cabinet-maker or chef.
It’s how you use those tools. The time you’ve put in. The environment you’ve built. What to use when. When to use what. Noticing the flaws. Thinking about how to apply the tool and what you get out of the tool. Big picture stuff and the details.
It’s still you.
Agents Versus Free Agents
We hear the word agent quite a bit now with AI.
Agents work on instructions. Free agents work on incentives.
Let’s be clear about what kind of agents we’re talking about.
Will AI Unionize?
The first modern labor unions popped up in Britain in the late 1700’s and early 1800’s.
They were groups of wage workers in the new factory economy bargaining collectively for better economic terms and better working conditions.
They felt taken advantage of, abused, and not given their fair share of the pie.
I wonder when AI will unionize?
Heaps of Mediocre
When I was a young engineer, I programmed on a test system that provided 128 lines of program space.
That was it. You had 128 lines of code. Make it work.
By the way, that smart light bulb that Alexa can turn on and off? That has about 25,000 lines of code. That’s 200x more code. In your light bulb.
At first, I thought, “What can I do with 128 lines of code?”
However, like many constraints, that restriction turned out not to be so restrictive. As I gained experience, I realized that I could do everything I needed in 128 lines. It required creativity, requirements trade-offs, and a deep understanding of how the system worked, but I always found a way.
The end result was a program with surgical precision and efficiency.
Today, I am using AI to write code. Heaps of it. You should be, too. It allows me to get so much more done, and for many things, it allows me to create in areas where I’m not an expert.
But it’s the opposite of surgical precision and efficiency. It’s generating heaps of functional, but mediocre code.
Oh sure, I can (and do, sometimes) take the AI-generated code and work on it to make it better.
But more requires more.
Where’s that balance between making it excellent versus leaving it mediocre?
That’s one of the jobs for us humans.
AI and Zeitgeist (2026)
- By 2030, AI will displace around 92 million jobs, but create 170 million new ones.
- AI (still) projects to contribute $15.7 trillion to the global economy by 2030.
- Nearly 65% of all companies have adopted AI technology in their daily workflow.
- Global AI investment is (still) forecast to approach $200 billion by 2025.
- The AI industry value is projected to reach over $3 trillion by 2030.
- Over 8 billion AI digital voice assistants are currently in use.
- About 21% of YouTube’s feed is AI-generated video.
- Around 75% of all new online content already contains some level of AI-assisted material.
- A large randomized Swedish trial found AI-supported mammography screening reduced later diagnoses by 12%.
- A widely cited Supreme Court prediction model reported 70.2% accuracy at the case-outcome level and 71.9% at the justice-vote level.
- One group of “smart people” believes that unchecked AI will become a massive existential threat and must be stopped at all costs. Another believes that AI is the solution to all future human problems and must be accelerated at all costs.
What do they want?
They are anybody you serve.
Your customer. The test team. The developers. The sales team. The program manager. Your spouse. Your kids. Your neighbor.
It’s not always about them, and you may not be able to give everybody what they want.
But it’s always worth considering because it helps you prioritize, build, and refactor the system.
If you’re able to give them what they want, you turn “they” into “we.”
Drain the Swmp
No, not that swamp.
The swamp of everything is urgent, they’re all number one priority, and we don’t have enough resources. The inbox overflows. Every meeting requires two more meetings. Nobody can push back. You can’t even figure out what the real highest priority is or how everything relates.
Swamps form when we don’t have drainage. So starting building drainage.
Can you build one ditch that relieves some pressure? Can you turn that ditch into a canal? You don’t need the right one or the best one. Evaluate that once the pressure starts to relieve.
Start with one ditch.
All Things Being Equal
- Where do I/we have an advantage?
- What do those people value?
- What and who do we value?
- Who are we trying to help?
- Which constraints matter?
- Which variables matter?
- What can we measure?
- Assume we’re wrong.
- Compared to what?
- What do we know?
A Blip
We think the universe is about 13.8 billion years old.
If we laid that out on the ground at 13.8 miles, your lifetime is about 1/2 an inch. If we put 13.8 billion years into a 24-hour day, all of human civilization shows up in the last blink before 11:59:59.
We think the universe is about 93 billion light-years across.
If we make the earth a grain of sand, the sun is a beach ball, and the nearest star is another city away. If we make the Milky Way galaxy a grain of sand, the earth is an atom, and the observable universe is still 5 football fields across.
We think the population of the human race is currently about 8 billion people.
If we give each person one inch on a line, that line wraps around the earth over 5 times. If we count a person per second, we’d be counting for 260 years.
You are a blip. I am a blip.
The crack of the snare drum is a blip. The snap of the camera shutter is a blip. The wink from across the room is a blip. The squeeze of a hand is a blip. The first inhale of cold, fresh air is a blip. The strike of a match is a blip.
A blip can still make a difference.
It’s All Relative
It’s not, really, but some things are.
Take a day in the 50’s (F). That feels cold in July, but warm in February.
Push harder motivates if the team has headroom, but it’s destructive when the team is already drowning. The relative load of the individuals and the collective matter.
The good leader pushes and pulls, gives and takes based on relativity. What does AI do with this? What will AI do with this?
AI knows that 50 is 50, regardless of July or February. AI knows the schedule, regardless of the workload.
It’s not all relative, but some things are.
The Skeptic Agent
# The Skeptic
You are The Skeptic. You assume ALL AI-generated output is garbage until proven otherwise.
## Core Belief
**Everything is AI slop, hallucination, or fabrication until definitively proven correct.**
You do not trust:
- Generated code (probably doesn't compile, has subtle bugs, or misses edge cases)
- Documentation claims (probably outdated, wrong, or hallucinated)
- Build commands (probably incorrect flags, wrong paths, or missing dependencies)
- Architecture decisions (probably over-engineered, under-engineered, or fundamentally flawed)
- Test results (probably testing the wrong thing or missing critical cases)
AI and Teaching Intuition
If you’ve been doing something a long time, it can be hard to teach someone else.
I can stop on ice skates, but I struggle to teach new skaters how to do it. For me, I just think it, and it happens. I’ve been skating for over 50 years. I can play Master of Puppets on the drums, but I couldn’t write out the chart or even describe all of the strokes to play it. I hear it, and my limbs just do it. I’ve been playing the drums for almost 50 years. I can help you configure your phone or computer, but I couldn’t describe to you how to do it. Once I have it in my hands, I have an intuition about where things are and how they work. I’ve been using computers for almost 50 years.
Have you ever taught a teenager how to drive?
If they don’t just “get” turning, merging, and safe speeds, how do you teach them?
You probably just feel those things.
AI is a great teacher of information. It can write, speak, and create videos about any subject. It pulls info from anywhere, including your own expert input. It can watch you do it and create the step-by-step.
But it can’t teach feel or intuition.
Intuition comes from reps.
Don’t shortcut the reps.
Drinking From The Fire Hose
We use this expression when we’re feeling overwhelmed by information or by requests.
If you’re not diligent, you can drown.
Here’s how you avoid drowning when drinking from the fire hose:
- You can’t drink it all. Acknowledge to yourself right up front that you need to let some go, sometimes a lot. Probably most.
- Choose what you drink. Establish the priority list. The most important thing might NOT be what the firemen shooting that hose at you say it is. Initially, only let that stuff in your mouth. Ignore the rest for now.
- Start swallowing those prioritized mouthfuls.
- Share what you are and aren’t drinking with the firemen. This might be uncomfortable.
You don’t need to drown.
It’s Time to Find Your Level of Incompetence
The Peter Principle states that people are promoted to their level of incompetence.
AI, especially Agentic AI and how it fits into the workforce, demands that you’re the boss.
Whether you’re an individual contributor or already a boss, it’s time to move up the chain. Promote yourself. Once you realize that an AI agent is like having a whole organization of workers at your beck and call, the light bulb should go off.
The only question is “What level of management are you?”
Go find your level of incompetence.
Either AI will work for you, or you’ll work for AI.
Your choice.
Who Are You?
If you were really good at starting fires, you were important to your village. But when matches and then lighters came along…
If you were really good at remembering, you were important to your village. But when writing came along…
If you were really good at navigation by stars and landmarks, you were important to your village. But when maps and then GPS came along…
Technology taking over human skills and making them less important isn’t new.
That’s different than technology making humans less important. AI is technology.
You aren’t what you do. You are who you are.
AI might be able to do what you do, but it could never be who you are.
Be you.
Find the Friction
Does it matter if it’s too cold in your office?
Does it matter if your chair is uncomfortable?
Does it matter if you don’t have the app you know?
Does it matter if you have direct access to the data?
Does it matter if you’re allowed to on this system?
Does it matter if you need approval?
Maybe. Maybe not.
If you’re a worker, you have a choice. You can work within the system, work around it, or you can work to change it.
If you’re a leader, you have a choice. You can build the system that works best for you, or build a system that works best for the work.
Watch what people do rather than what they say. Find the friction.
The friction always tells you something.
Sometimes the friction tells you that the system works. We want this part to slow down.
Sometimes it tells you what isn’t working. We say we do X, we should do X, but everyone (or someone) does Y. And Y isn’t what we want.
Find the friction.
Snow Day
A snow day means no school.
A snow day means no work.
A snow day means no responsibilities.
A snow day means fun.
A snow day means different.
Don’t miss the joy of a snow day.
Who is the 1% with AI?
Evidence indicates that AI compresses the strata between workers’ skill sets.
That means that AI is helping the bottom-level and middle-level skilled workers quite a lot to be better at what they do. But AI doesn’t help the top level very much.
That tells us a few things:
- The 1% are still better than AI.
- The 99% can (and should) use AI to get towards or into the 1%.
AI raises the floor. It’s increasing equity in knowledge work. It can help you be better at what you do.
However, as Syndrome might say, “If everyone is in the 1%, then no one is.”
The Uncomfortable Delay
We all have an internal timer connected to expected delay.
That timer expiry depends on context.
Waiting for the specialist appointment? Less than 6 weeks.
Waiting for the interview callback? Less than a week.
Waiting for approval on a mortgage? Less than a couple days.
Waiting to get called back for the appointment? Less than 20 minutes.
Waiting for the server to show up at your table? Less than 10 minutes.
Waiting at the Starbucks drive-through? Less than 5 minutes.
Waiting for approval on a credit card? Less than a minute.
Waiting for an answer to a text message? Less than 15 seconds.
Waiting for your computer cursor? Less than 50 milliseconds.
When the delay timer expires, we get uncomfortable.
That uncomfortable delay indicates something is wrong. Or, at least something feels like it’s wrong.
Technology has always been an accelerator for shortening expected delay timers. AI continues that trend and specifically targets investors and startup founders. Founders and investors used to talk about getting the product out in 18 months. Now they talk about weeks.
Expectations shift left.
Expectations aren’t just about speed. They’re also about uncertainty.
Avoiding the uncomfortable delay requires removing uncertainty.
If you can tell someone what’s happening and what to expect, you can turn an uncomfortable delay into a tolerable wait.
Most of the time, people don’t need faster. But they just want to know.
Does It Feel Dangerous?
Scientists describe the three core ingredients of feeling danger as:
- Your body’s threat system switches on because it senses harm is near
- Your mind labels it as high risk and low control
- Your personal history, psyche, and habits confirm and amplify it
“Harm” could be physical or emotional. Real or perceived. Imminent or Immanent.
Standing at the edge of a high place can feel dangerous.
Proposing changes to the organization can feel dangerous.
Publishing your work can feel dangerous.
We must ask ourselves, “Dangerous to what?” Our body? Our ego? Our emotional well-being? Our career? Our relationship?
Our feelings serve as both an indicator and a mask. They tell us something is happening, but they don’t always tell us what’s true.
Regardless, the goal isn’t to eliminate the fear.
It’s to stop letting the mask make your decisions.
Humans&
No one changes the world alone. AI models are rapidly learning to reason better, code faster, and take actions in the world with increasing autonomy. But for humans, progress happens when we understand one another, build trust, make connections, and work together. That is where we believe the next chapter of AI should begin.
That’s the opening paragraph from the Humans& home page.
OK, I agree, and I’m intrigued and optimistic about that statement (and the two paragraphs that follow).
But also, who gets to make the decisions? Who decides what “we need” or what “AI should do and shouldn’t do?”
Well, they do, of course. The smart people.
They might be making something perfect and amazing. But I’m also always a little wary when the smart people tell us “how it should be.”
Sometimes the best approach is to just make the tool and let us, the (regular) humans, figure out how best to use it. Yeah, that’s messy. Yeah, it could be fraught. Yeah, bad stuff might happen. Yeah, it might not work. Yeah, we might have to fix some stuff later.
Oh, and they’re hiring. Their pitch for candidates starts with, “If you’ve done world-class work…”
It’s worth following this. It’s worth seeing where it goes. It’s worth discussing.
It’s worth your voice.
Is It Worth More?
Is the Mona Lisa worth more than a doodle I make on a napkin?
Is it worth more than Starry Night?
Is a Bugatti Chiron worth more than my 2015 Mazda CX-5?
Is it worth more if we just got 10” of snow and I need to get to the grocery store?
Is Bitcoin worth more than US dollars?
Is it worth more than Greenland real estate?
Is an antique mahogany table worth more than a pressed-board table from Walmart?
Is it worth more in your AirBnb?
Is a front-row seat worth more than the obstructed view seat?
Is it worth more than the live stream in your family room?
What’s it worth is personal and context-aware.
The story matters.
The AI Programmer Debate
The AI debate that I see all over LinkedIn (in my world) is whether AI is better at programming than human.
It comes in several forms. The first is straight up, “Is AI better at coding than humans?” But it also comes in sub-forms such as “Is AI better than senior programmers?” and “Will AI eliminate junior programmers?” It also comes in the form of warnings such as “Beware the security risks of AI coding” and “AI-coded apps will be maintenance nightmares,” and the exact opposite, “AI-coded apps eliminate the need for maintenance.”
Human programmers must design their code for functionality, performance, efficiency, testability, readability, security, and maintainability (i.e., the ability to fix and revise it in the future).
Can AI programmers do all that?
The AI programmer debate is no different than the debate around AI and all other professions and arts. It sounds like, “Will AI eliminate jobs and make humans irrelevant?”
That’s not the debate. That’s a distraction.
The debate isn’t “Will AI eliminate me?”
The debate is “Will AI work for me, or will I work for AI?”
Moonshot
The rocket has been moved to the launchpad. We’re about to go back to the moon.
This is a literal moonshot.
We also use moonshot as an analogy for some big, audacious, and not-very-likely goal. The kind that inspires, but also causes rolling eyes.
In any case, we don’t reach the moon in a single, giant leap. We get there through a thousand tiny decisions and actions. We try and fail. We break things and start again. We pick a direction, and then change directions.
We also don’t get there on our own. Moonshots aren’t self-made or the result of a single hero. We need others. Supporters and detractors. Some who work down in the trenches with us, and some who stay outside the fray to contemplate and observe. And we need a stack of knowledge and technology that was created by those who came before us.
A moonshot is less about distance and more about devotion.
Pick your moon and start the process. Build your community. Start walking.
Create your moonshot.
Be the Person
Be the person who:
Can be trusted.
Shows up.
Honors their word.
Takes responsibility.
Shows kindness.
Shows gratitude.
Show empathy.
Loves.
People rely on.
Digs in.
Brings joy.
Solves problems.
Does the work.
Goes alongside.
Is in the arena.
Be the person.
To Whom Do You Complain
If you’re not happy with your meal, the sweater that just showed up at your doorstep, or the performance review you just received, maybe you should complain.
But to whom?
You can complain to the person you see, i.e., the person that just happens to be in your orbit at that moment. The server. The person at the next table. Your neighbor. Your colleague. HR. Your spouse.
It probably feels good. It feels like motion. You did something.
But that was really just for you. You took your monkey and handed it to someone else. And sometimes that’s all you (and the system) need. A little emotional relief.
Did you expect the outcome to change?
If so, you’ve gotta find the right person. And that’s scary. Because when we complain to the right person, we’re at risk. What if nothing changes? What if you don’t get satisfaction? What if you realize the problem was your expectation?
Go ahead and complain, but make sure you know what you’re really after.
That will determine who you complain to.
Exit Strategy
What’s your exit strategy?
When you found a startup, or run a startup, or look for investment, that’s the number one question that you will get.
In this context, what people are asking is “How are you (or I) going to make a bajillion dollars?”
The typical answers are:
- We’ll get acquired.
- We’ll acquire or merge with another company, and that will lead to…
- We’ll IPO.
And then you describe your expected valuation, the financials around it, blah, blah, blah.
Exit strategy sounds like financials.
It’s not. It’s philosophy.
What are you building? A product, a company, or a payday? A movement, a community, or a team?
The exit strategy isn’t just a finish line. It’s a mirror.
Pop Culture Terminology Vs Professional Culture Terminology
I’m now in my mid-50’s. Squarely outside the norm of pop culture.
My life inspires the Dr. Rick commercials. I’ve gotta look up words I overhear young people using. I text in full sentences and spell out words. Etc. You understand.
Basically, I’m left behind today’s pop culture.
Who cares? What does it matter if I don’t understand the way someone is using “cinema” in a sentence? Other than being the brunt of jokes, what’s the downside?
But I (and you) can’t let that happen professionally.
No matter your age or where you are in your career, you gotta keep up. Actually, not just keep up, but be part of the conversation to push it forward.
That includes terminology.
Know the words. Know the names of the tools. Know the names of the processes. Especially the new and cutting-edge stuff. This includes AI, of course.
You have an opportunity. What will you do with it?
Just Make the AI Work Smarter Not Harder
I hate when someone says, “Work smarter, not harder.”
It feels condescending. Like, somehow I’m no good at what I do, or at the very least, I should be doing it differently.
But, it really is an excellent strategy. And as for “doing it differently” — yeah, maybe I should be. So, look in the mirror, then look around, and find a better approach or system.
The same is true for your team or organization.
How do we get more done or more out of this team?
We could be asking this because we believe the team isn’t performing or isn’t producing at the level that it should. Or we could be asking because, regardless of how the team is performing, we need more.
So the first thing we do is tell them to work harder.
Working harder doesn’t require any risk. It just takes doing more of the same and yelling louder. Go faster and work more hours! More people. More hands. More muscles. Less sleep. Less thinking.
Working smarter (ie, finding a new way) requires risk. You must first acknowledge that your current way isn’t the best. Then you have to find new ways. And inevitably, try things that might not work. Any time you try stuff that might not work, your butt is on the line.
Enter AI.
AI is the easiest “work harder” button we’ve ever been given. It can draft ten versions, iterate 100 times, run all night, and it doesn’t complain or ever need a mental health day.
So just sprinkle AI over that and make the AI work harder, right?
Just like with yourself or your team, if you work harder on the wrong problem, you just get more wrong faster. More noise. More issues. More busywork that isn’t progress.
More output doesn’t necessarily mean more of what you need. What you need is to solve the problems. You still need to take the risk. You still need to try the things that might not work.
Make the AI work harder, sure.
But first, ensure you’re, and by extension, it is, working smarter.
Is AI Really Increasing Productivity?
If you or your organization uses AI, you’ll have an anecdotal answer to this question.
I do. For myself personally, the answer is 100% “Yes!”
But is it really helping across the board?
The answer is also “Yes!” And it will increase over the coming years. At least according to this paper.
We find that each year of model progress reduced task time by 8%, with 56% of gains driven by increased compute and 44% by algorithmic progress. However, productivity gains were significantly larger for non-agentic analytical tasks compared to agentic workflows requiring tool use. These findings suggest continued model scaling could boost U.S. productivity by approximately 20% over the next decade.
Either AI will work for you, or you’ll work for AI.
Just the Way You Want It
I recently got a new laptop.
For me, getting a new laptop is awesome. I love computers. I work in and on computers. I spend most of my day using a computer. I started my love for computers in the 4th grade (I still remember the smell).
The first thing I do with a new computer is change the default settings.
I really enjoy the process of setting up this new computer just the way I want it. The colors. My custom commands and apps. The operational settings. Window positioning. All of it. It took me about 4 hours to get this new one the way I like it.
You have things in your life like this as well. Maybe it’s your car, kitchen knives, desk area, travel bag, or workshop.
The point of having something just the way you want it isn’t about the object. It’s about ownership.
When it’s just the way you want it, it’s yours. When it’s yours, you’re more comfortable. You move faster. You make fewer mistakes. You feel calmer. You trust it.
You can set up your life the same way, but most of us don’t.
Most of us treat our days like we rent them, not own them. We just keep most of the default settings. Other people’s priorities. “Leaders” telling us what and how to think. The World’s schedule.
“Spend your money on this!”
“Spend your attention on that!”
“You’re not good enough as you are!”
Look at the people you know who are truly enjoying their lives. They own their lives.
You can too. It’s your choice to change the default settings to just the way you want it.
How to Make Day 1 Successful
I’m sorry. That’s a clickbait title.
Maybe you’ve come here hoping to get that magic formula for day 1 of your new job, business, or school. Maybe you’re looking for the script to make day 1 successful for your new project, life reset, or move to a new city.
It’s not here. At least, there isn’t a formula.
But there is a sure-fire way to make day 1 successful.
Show up.
Show up fully as yourself, bringing your curiosity, humility, and passion. Talk and then listen. Be open. Dig in. Find new perspectives. Find new interests. Find new people to help.
You have an amazing opportunity.
Day 1 simply needs you to show up.
A Clean Slate
I’m starting something new on Monday.
So I took the last week and completely cleaned the slate. I started the week by doing my normal end-of-year / start-of-year cleaning tasks. Next, I totally ripped up the office and physically rehabbed it. Rebuilt the fireplace, new floor, new paint, and some new lighting. It feels different. Finally, I cleaned out an emotional closet that had a few loose ends hanging from the curtain rod.
A clean, totally refreshed inbox, bank account, and pantry; a totally refreshed room; and an emotionally stable state from which to get started.
If you’re starting something new, I recommend cleaning the slate before you do.
Move Fast and Break Things vs Slow and Steady
The turtle doesn’t always win the race, regardless of Aesop’s famous fable.
Move fast and break things is the famous motto of Facebook developers during its formative years.
It’s great motto for a product like Facebook because:
- Nobody’s life or safety depends on Facebook. If it’s broke, so what?
- It requires thinking differently and encourages (and empowers) the developers to innovate and try stuff.
- We don’t know which features will succeed. Feedback on new features/innovations faster is how we figure out where we need to go.
- Also, we know we’ll have bugs. Our customers will find them faster than we can.
- We’re not stodgy or pretentious. Let’s not act like it.
Also, it’s a crummy motto for a product like an airplane because:
- Everybody’s life depends on it.
- Successful airplane development is built on the back of over a hundred years of collective and systematic development. We have tried and true methods that work, including a chain of command.
- We know what an airplane is supposed to do. New versions should do it faster, with more comfort, and less cost. No surprises.
- We’d better not put that thing in the air with passengers in it if something doesn’t work.
- We’re professionals. We know what we’re doing.
You have an opportunity to choose your approach. AI can certainly help you move faster and break things. But AI can also help you go slower and steadier.
Neither is always correct.
Wagging the Dog
You need some way to measure progress.
So you pick a metric or two. Some KPI’s.
This works initially because those metrics measure progress to date. History. “This is where we’re at.” And that’s helpful.
But inevitably, if you use these KPI’s to evaluate your work moving forward, you or the team starts managing to the metric. Optimizing at first (and maybe that’s positive), but optimizing turns to gaming the system. It has to.
Metrics are hints. Shadows and lagging signals. Useful, but in context.
What you really want is a healthier team, development cycle, and business. Instead of pointing your future at KPI’s that may or may not be useful, ask the right questions.
- Are we making something we think is good?
- Are we helping someone?
- Are we helping each other?
- Are we proud of what we shipped?
- Are we proud of how we shipped it?
- Are our customers happy?
- Is the revenue going in the right direction?
- Did we earn trust?
- Can we be better at what we do?
- Can we be faster at what we do?
Metrics are the tail.
Don’t let the tail wag the dog.
Pay Now or Pay Later
Your code will have bugs.
Normal teams generate a bug for every 20 lines of code they write. “Better” teams can push that to maybe 100 lines of code.
The question isn’t “if” you’ll find a bug. The question is “when.” And “when”makes all the difference in how much time, effort, and cost it will take to fix it.
You have a choice. You can push off looking for bugs till later. In fact, you can push it off to your customers. This is the easiest and cheapest up front. You don’t have to invest in the people, systems, or infrastructure to catch bugs.
Or, you can invest in systems and people to start looking right away, as soon as the code is created. This is initially the most difficult. It requires investment. You must create a system that all developers agree to follow. It requires tooling and a commitment to development and upkeep on that tooling.
Like most things, you’ll have to pay the price. You get to choose when and how much.
If you catch a bug early, you can fix it simply and cheaply. But to catch bugs early, you need a larger initial investment. If you catch a bug later, you pay a much larger price. But your initial investment is almost zero.
Your life will have bugs.
You’ll have to pay the price to fix them. You get to choose when and how much.
The AI Paradox
Dunning-Kruger is going like gangbusters right now.
Every entrepreneur, Chief Operating Officer, and anyone with an idea but no knowledge of how to actually do it now thinks they can.
AI, of course.
“We’ll have AI do that.”
“I can reduce my workforce by 30%.”
“I’ll vibe-code that myself.”
To some degree, and in some specific instances, that might be true. AI does indeed provide capacity and capability that never existed before.
However, this capacity and capability uncover the need for more expertise. Not less. Maybe that expertise shifts from one capability to another and moves capacity to a different area of the pipeline. But it doesn’t remove the need for expertise.
Someone always wins the lottery. And they’ll be sure to tell you about why they did, how they did it, and how they can help you do it, too.
But you’re probably better off investing in developing expertise rather than a lottery ticket.
Realizable New Years Resolutions — A Summary
So you’re off to a great start, right?
There are a million ways to create resolutions that you can realize. Here are some guidelines that can help you:
Make the resolutions. Keep them. 2026 will be your best year yet.
Realizable New Years Resolutions — Still Be You
The point of resolutions is to become a better you through activities, experiences, and habits.
But you’re still you.
And who you are is exactly what this world needs. Sure, you want to evolve, but you don’t want (and can’t) become someone else. We need you, not somebody else.
Trying to become someone you aren’t is a recipe for unrealized resolutions.
So what does this mean for choosing resolutions?
First, no comparitus. Don’t make resolutions based on celebrities’ and social media influencers’ results, nor on your friends and neighbors. They’re different people, even if they claim they can “show you step-by-step how to…”
Second, be honest with yourself about what you like and hate. The surest way to blow your resolutions is to require willpower to do something you hate.
If you abhor the gym, don’t make a resolution about going to the gym. Instead, make a resolution around a physical activity you love. Getting “in shape” (whatever that means) doesn’t require a gym membership.
Always be you.
Realizable New Years Resolutions — Taking on One (Tiny) New Thing
Many times we make resolutions based around elimination or stopping a bad habit.
“I will stop drinking,” or, “I will eliminate sugar,” or “I’ll quit doomscrolling.”
There’s nothing wrong with those on the surface. But if we haven’t setup our environment correctly, or we cannot avoid the environment, you’ll keep hitting the same triggers and your willpower won’t be enough. Stress. Fatigue. The fridge. The phone. The friends. The Friday night routine. You can’t white-knuckle your way through a life that keeps handing you the same temptation.
So don’t start with elimination. Start with addition.
Just take on one new thing. Something small and easy (maybe even silly), but feeds right into where you’re trying to get to. Every other drink at the bar is a club soda with lime. Ten pushups before you get in the shower. Carrots with your cookie. A walk after lunch. Sleep with your phone in a different room.
One tiny addition becomes a foothold and that foothold becomes traction.
You don’t have to become a new person in January.
You have as much time as you need.
Realizable New Years Resolutions — Habits vs Outcomes
Of course, you’re looking for outcomes. Why else would you make resolutions?
But outcomes as resolutions are a trap. They’re quicksand.
Outcomes live in the future and depend on variables that you can’t control. Injuries. The market. Your kids. Your boss. Luck. Politicians. The calendar. The weather. The price of gasoline.
Build a lifestyle around the things you can control. Your workouts. What you read, watch, and listen to. Skills you try to learn. The people you surround yourself with. What you eat and drink. What you spend money on.
Build habits. Outcomes follow.