I learned a long time ago that the signal is never clean.
That was true before AI showed up in a polished interface and started acting like it had invented uncertainty. I learned it on radios, circuit boards, oscilloscopes, frequency counters, wattmeters, signal generators, and the little alignment tools we called tweakers. You would tune a circuit into spec, sit there for seven seconds feeling like a genius, then touch the next potentiometer and watch the first reading wander off like it had unpaid taxes and a drinking problem.
That was electronics. Nothing stayed perfect for long. Tubes warmed up. Components aged. Connections got flaky. Antennas cared about weather, terrain, grounding, physics, and whatever other grudges the universe had filed that day. The manual gave you a diagram and a specification. The equipment gave you whatever it felt like giving you.
You did not get truth. You got readings, evidence, tolerances, and a system that rarely behaved as politely as the documentation suggested it might. You got a spec window, a meter, a schematic, a test bench, and the responsibility to decide whether the thing was actually working or merely producing a number that looked reassuring.
That kind of work changes how you think. It teaches you that “working” and “working correctly” are not the same condition. It teaches you that confidence is cheap, clean output can still be wrong, and systems can lie without having any intention at all. Most of the time, they lie convincingly.
So when generative AI arrived and half the planet began treating it like a digital oracle with venture funding, my reaction was not especially mystical. I thought it was fantastic. I also thought it was dangerous as hell. Both things were obviously true.
To me, AI is another noisy system. It is unusually powerful, strange, fast, and useful, but it is still a system. It has to be tuned, constrained, tested, challenged, and judged. The model gives you signal. The human still has to know what to do with it.
That is where this gets interesting.
The model gives you signal. The human still has to know what to do with it.
Before prompts, there were potentiometers
I did not begin my career in a strategy meeting. I started in rural northeast Arkansas, where work was not discussed in branding language. Work was work. You did it because it needed doing, because money mattered, and because nobody was standing nearby with a thoughtful multiyear development plan designed around your potential.
I graduated from high school at 17. Roughly two months later, I was in basic training at Lackland Air Force Base in San Antonio. There was not much of a pause between those two versions of life.
Part of the decision was education. Part of it was technical training. Part of it was the simple reality that I did not have wealthy parents waiting with a checkbook and a paved runway. Some people get a runway. Some people get gravel and a shove. I got ground radio communications.
That meant air traffic control radio, long-haul HF, tactical communications, NATO systems, big antennas sitting in fields, radio trailers, tubes, patch panels, and test equipment. These systems had to work because the people depending on them did not need a motivational quote. They needed communications.
I spent the first six years of my adult life overseas. Japan came first, then Germany. I met my wife in Japan. We dated there, married just before leaving, and moved to Germany for a NATO special duty assignment. We were young, broke, newly married, and living a life much larger than anything I had imagined as a kid in Arkansas.
That changes a person. Japan changed me. Germany changed me. The Air Force changed me. Electronics changed me. The world became larger than my hometown, my assumptions, my accent, my church, my comfort zone, and the version of reality I had inherited before I was old enough to examine it.
The work also taught me that systems have personalities. Not technically, of course, but spend enough nights in radio rooms, maintenance shops, field sites, and half-documented infrastructure and then tell me machines do not have moods. I will wait.
Some systems fail loudly. Some fail politely. Some fail only when the one person who understands them is on leave. Others fail in ways that make everyone insist the problem cannot exist, which is usually the first useful evidence that it absolutely does.
Electronics teaches fault isolation. You half-split the problem, test the midpoint, move upstream or downstream, and keep narrowing the failure. You stop guessing. You respect the evidence without marrying the first theory that looks attractive. That is how trouble starts, both in troubleshooting and in life.
That mental model followed me into computers. By the early 1990s, I had become one of those people other people found when a machine refused to behave. I could fix things, not because I had been born with mystical diagnostic powers, although that would have saved considerable time, but because radio had already taught me how to reason through failure.
Computers were another signal path with newer and more creative ways to betray you. Bad memory. Bent CPU pins. IRQ conflicts. Sound cards stomping on memory ranges. SCSI termination weirdness. Config.sys. Autoexec.bat. Jumpers. DIP switches. Network cards with unresolved childhood trauma. Printers, which were clearly designed by people seeking revenge against civilization.
There was no AI assistant waiting to explain any of it. There was no modern web search experience. There was no polite chatbot offering three likely causes before confidently inventing a fourth. You read manuals. You dug through Usenet, dial-up BBS communities, technical text files, and whatever useful knowledge other obsessed people had managed to leave behind.
You broke things and rebuilt them. You swapped parts, compared notes, stayed up too late, and learned from technical weirdos who also treated sleep as more of a suggestion than a requirement. That was the school. It was not glamorous, consistently credentialed, or particularly clean, but it was useful.
The Amiga in the German flat
My real computer life began in Germany.
My sponsor there, Don, had an Amiga setup. He may have had several Amigas. I remember an Amiga 500, an Amiga 1000, and perhaps a 2000 later. The exact inventory matters less than what happened when I saw them.
Something lit up. The graphics, the sound, the software, the demos, and the entire creative mess of the machine felt different. One of the first games I remember seeing was Zoom. I looked at it and thought, holy shit, this is cool. That was enough. I had to have one.
This would have been a reasonable instinct if I had also possessed money, restraint, or the judgment expected of a newly married adult. Unfortunately, I had enthusiasm, which history has shown to be much more expensive.
I went to the base exchange, got credit, and bought an Amiga 500 with a Commodore 1084S monitor. I also bought one of those cheap, glue-together fake wood desks because apparently a questionable financial decision becomes legitimate once it has a command center.
We had almost nothing at the time. Borrowed furniture. Bare essentials. Not much of a mattress. A small flat in a German village. A young marriage still learning how to be a marriage. In the middle of that sat a computer setup I could not really afford.
My wife was furious. She was right. That is an important sentence in any honest marriage: she was right.
The purchase was irresponsible, but the machine opened a door in my head that never closed. I went deeper into building PCs, running bulletin boards, living on shell accounts before the web became fashionable, and spending time on text-based IRC. Then came modems, burners, SCSI drives, early servers, Cisco, networking, software collections, and the unruly early computing world that felt less like a consumer market and more like a frontier.
Back then, you did not onboard to technology. You fought your way in. You learned by breaking things, staying up too late, reading ugly documentation written by people who clearly hated paragraphs, and trying again after reality rejected your first theory.
That era produced a certain kind of technical person. Not always polished. Not always credentialed in the approved institutional format. Not always easy to manage, a statement I offer with considerable sympathy to every manager who has tried to domesticate a useful technical maniac.
But these people were resourceful, curious, stubborn, and willing to learn ugly systems from the inside out. They could find the fault after the system, the vendor, the log file, and three meetings full of people had all agreed the fault was impossible.
That kind of person is not obsolete in the AI era. That kind of person may be exactly what AI needs.
The long frustration of seeing what you cannot yet build
Most of my career has been spent in infrastructure: electronics, support engineering, computer systems, endpoint work, enterprise platforms, program and project work, Y2K, networks, servers, automation, and operations. It is the kind of work where everything connects to something else, nothing is simple, and the root cause is usually hiding behind five layers of people saying, “That should not happen.”
I loved a lot of it. I still do. I still love technology, the home lab, and the satisfaction of building systems, automations, storage, services, and tools that solve real problems. I still enjoy making the machine do something useful.
But one frustration followed me through most of those years. I had ideas I could not fully execute. Not small ideas, either. I could see systems, products, workflows, automation models, service designs, interfaces, governance models, and operating patterns in my head with more clarity than I could make visible to anyone else.
I could explain the architecture. I could describe the business need, identify the failure modes, anticipate where the system would break, predict who would resist it, and tell you why the first implementation would probably be wrong. Some technical people understood what I was seeing. Others did not.
The problem was not the idea. The problem was getting enough of it out of my head and into the world.
I was not a full-stack developer or a polished designer. I was not the person who could sit down alone and convert an entire concept into a beautiful application, an elegant pitch, a clean prototype, a persuasive narrative, and a credible implementation path. I could see the thing. I could not always build enough of it for everyone else to see it too.
Living with that gap for decades gets irritating. There is a particular kind of pain in seeing a future you cannot yet make legible. Then, years later, somebody builds something adjacent to what you had imagined and the market suddenly calls it visionary. You sit there thinking, I saw that. I knew something like that should exist. I just could not drag it out of my skull fast enough.
That is not bitterness. Not entirely. Fine, partly. There is no reason to ruin an honest point by pretending I am a saint.
Mostly, it is recognition. Ideas are not enough. Taste is not enough. Technical judgment is not enough. Being right inside your own head is not enough. Eventually, the thing has to become visible. It needs to become an artifact, a prototype, a system, a story, a decision, a page, a workflow, or a model that other people can inspect, challenge, improve, reject, adopt, or build upon.
That was the gap. AI started closing it.
AI did not give me ideas. It gave my ideas hands.
AI did not make me creative or technical. It did not give me decades of troubleshooting scars, radio calibration habits, infrastructure judgment, or a healthy suspicion of elegant nonsense. What it changed was the distance between thought and artifact.
That is the whole game.
For most of my career, I could see the thing before I could build it. Now I can build enough of it to discover whether the idea deserves to live. That is not a minor productivity gain. It is a career event, and perhaps a life event.
With AI, I can prototype, code with assistance, shape product concepts, pressure-test architecture, generate interface directions, build pitch narratives, write governance patterns, and simulate objections. I can turn a messy brain dump into a structured article, working page, validator, prompt package, business case, runbook, or implementation blueprint.
None of this happens perfectly or magically. It still requires review. The machine gets things wrong, sometimes confidently and sometimes beautifully. Occasionally it produces the kind of smooth wrongness that makes you want to unplug civilization and start again with better documentation.
But it gets me moving, and that matters. AI collapses the distance between thought and action. For people who already have domain knowledge, judgment, taste, scars, and stubborn curiosity, that collapse can be profound. Dormant ideas can become prototypes. Imagination can produce evidence. Instincts that once sat in the mental parking lot can become systems that other people can evaluate.
I am not excited because AI can write a bland email. The world already had plenty of bland emails. We were not facing a shortage. No civilization has ever collapsed because it lacked one more person “circling back.”
I am excited because AI lets me turn systems in my head into working artifacts before the energy dies. That is the difference. That is the ignition.
AI collapses the distance between thought and action.
The model is not the differentiator
This is the part people keep missing: the model is not the differentiator.
Access is not competence. A prompt box is not a strategy. A subscription is not a capability. Almost everyone can open roughly the same model and ask it for a plan, script, summary, pitch, design, workflow, or miracle with bullet points. The resulting quality varies wildly, which should tell us something.
AI amplifies the person using it. It amplifies clear thinking, domain knowledge, good taste, structured reasoning, strong questions, and the ability to compare an output against reality. It also amplifies laziness, nonsense, shallow strategy, cargo-cult leadership, fake expertise, and executive theater at machine speed. That last category is going to be expensive.
People use AI in very different ways. Some treat it like a calculator. Some use it as a search engine with confidence issues, a ghostwriter, a junior analyst, a design partner, an engineering assistant, a mirror, or a slot machine for corporate vocabulary. That final use should probably require licensing.
The tool is powerful. The operator still matters.
If you have spent years troubleshooting systems where one bad assumption could waste hours, break production, or strand people inside a problem nobody wanted to own, AI feels strangely familiar. You learn to interrogate the output. What did it miss? Which hidden dependency did I fail to state? What constraint did I assume without saying? Which source is carrying the claim? What sounds too confident? What would prove this wrong? What happens if this goes into production and the adult supervision leaves the room?
That is not merely prompt engineering. It is engineering judgment.
Do not rule out the old hats
Companies need to be careful about a lazy assumption that has attached itself to the AI conversation: the future supposedly belongs to the youngest people in the room because they grew up closer to the newest tools.
Sometimes the youngest person will be the best person. Plenty of younger workers are fast, creative, fearless, and technically sharp. This does not need to become a generational cage match. The internet already has more than enough nonsense, and most of it has comments enabled.
Age is not the variable that matters most. Learning velocity matters. Judgment, domain depth, curiosity, pattern recognition, operational scar tissue, and the ability to work inside ambiguity all matter. So does the ability to recognize when a system is producing polished garbage, a skill organizations will need more than they currently want to admit.
That is why I would not rule out the old hats. Some of us have worked with probabilistic behavior, tolerances, noisy signals, operational constraints, brittle infrastructure, incomplete documentation, and high-consequence environments for decades. We may not describe those habits with fashionable vocabulary. We may not call ourselves “AI native,” mostly because some of us still have a small shred of dignity left. The mental model is there.
The real question is whether experience keeps learning or starts to calcify.
There are two very different late-career paths through the AI era. One is defensive. The work changes, the tools change, and the language changes, so the experienced person retreats into resentment. Everything new becomes stupid. Everyone younger becomes naive. Every tool becomes a toy. Every change becomes theater. That path is real, and it will not end well.
The second path is leverage. The experienced person keeps learning. New tools meet old judgment. Decades of tacit knowledge become more executable than they have ever been. That is the path I care about.
The risks are real. A June 30, 2026 brief from Boston College’s Center for Retirement Research found that workers age 55 and older are just as exposed to AI as mid-career workers. It also found that workers in jobs with high AI exposure have experienced a rise in job exits since generative AI use surged. The researchers were careful about the early evidence, and everyone else should be too, but the direction is not a cute trend. It is a warning flare.
AARP’s updated 2026 research shows the tension from another angle. Familiarity with workplace AI among workers age 50 and older rose to 52 percent in its third wave. Only 12 percent reported taking AI training or classes for work, while 49 percent said they wanted to learn more about using AI on the job. Older workers continue to see AI as both an opportunity and a threat, with concern outweighing optimism. That is not evidence that older workers cannot adapt. It is evidence of a training and access gap sitting next to a substantial appetite to learn.
So no, this is not a story made entirely of inspiration and sunrise photos. AI can push people out, expose skill gaps, and provide one more excuse for lazy age bias wrapped in innovation language.
That is only half the story. The other half is leverage.
People who understand the business, the risks, the exceptions, the data, the customers, the systems, the politics, the buried operational debt, and the real cost of failure are not optional in serious AI work. They are often the difference between a demo and a capability.
A demo has to impress people for ten minutes. A capability has to survive reality. Those are different sports.
Retirement got interrupted
Before AI pulled me back into the deep end, I had started thinking seriously about retirement. Not because I hated the work, had nothing left to contribute, or believed I was finished. I was bored. That is the honest answer.
Infrastructure work matters. Support engineering matters. Enterprise platforms matter. Keeping real systems alive matters. But after enough years, a person starts doing the math. Sixty-two? Sixty-five? Sixty-seven? You begin thinking about the glide path and when it might make sense to hang it up from corporate America.
You also start asking whether the next version of the work will feed the same part of you that first came alive around radios, Amigas, networks, servers, and all the strange machinery that carried you this far. For me, the answer had started to look uncertain.
Then AI changed the equation. Not because it was shiny. Shiny wears off.
AI got my attention because it made old ideas reachable. It moved the work from maintenance mode back into invention mode. It gave me one operating surface where infrastructure, architecture, automation, product thinking, design, writing, governance, and problem-solving could meet.
That is why I can spend absurd hours on this without experiencing it as normal work. Yes, that probably sounds unhealthy. Some of it probably is. Obsession rarely arrives wearing a balanced lifestyle and a sleep tracker.
There is still something deeply energizing about reaching the later part of a career and discovering that the thing you may have been preparing for is not behind you. It may still be ahead. I did not expect that.
I thought AI might help me finish the race more efficiently. Instead, it made me wonder whether I was done racing at all.
Experience still has to earn the room
None of this means experience deserves automatic reverence. Experience can be powerful. It can also become a museum exhibit with opinions.
Years alone do not make a person valuable. Scars do not automatically produce wisdom, and seniority is not the same thing as judgment. We have all met people with twenty years of experience who really had one year of experience repeated twenty times, usually with a larger calendar and stronger coffee.
Experience matters only when it keeps metabolizing new reality. That is the standard.
If AI changes the work, experienced people have to change with it. They do not need to pretend they are twenty-five, chase every new tool, or post synthetic enthusiasm on LinkedIn and call it transformation. They need to do the work.
Learn the tools. Build real things. Understand the failure modes and the governance problems. Know where the data comes from. Know what should never be automated, what must remain human, what can be accelerated safely, and what requires evidence before anyone calls it progress. Learn when the model is helping and when it is merely producing expensive fog.
That last distinction may become one of the most important skills in the enterprise. The surface is getting easier while the danger gets quieter. The output looks better. The reasoning underneath may not be better at all. That is a dangerous combination.
The Federal Reserve noted in April 2026 that Census Bureau survey data showed about 18 percent of firms had adopted AI by the end of 2025. It also emphasized that adoption estimates vary according to the respondent, unit of analysis, and wording of the question. The broad point is not that one number settles anything. It is that adoption is moving while institutional maturity remains uneven.
That matches what I see. People are experimenting faster than organizations are adapting. The toy phase is easy. The governance phase is harder. Production is where the bodies are buried.
That is where experienced operators matter.
This is not young versus old
The age framing needs care because young versus old is the easiest possible story, which is usually a good reason to distrust it.
The World Economic Forum reported in June 2026 that more than one in three young workers globally are in occupations with medium to high exposure to AI-driven task change. Early-career workers are being hit too. In some cases, the ladder is moving before they can get a proper foot on the first rung.
That matters. If AI consumes too much entry-level work, companies may eventually discover that senior people do not hatch fully formed from quarterly planning decks. Someone has to learn the basics, make mistakes, do junior work, receive correction, and remain in the system long enough to become capable of handling senior work.
Apparently this needs saying because civilization has a habit of removing the bridge and then acting surprised when nobody can cross the river.
The meaningful divide is not young versus old. It is adaptive versus static, curious versus performative, disciplined versus careless, builder versus passenger.
Some young workers will become extraordinary with AI. Others will become shallow and prompt-dependent. Some older workers will be pushed aside unfairly. Others will refuse to move. Some will become more valuable than ever because they know things the model does not know, and they know how to use the model without trusting it blindly.
That is the interesting part. AI does not erase experience. It exposes the quality of it.
A life measured in signals
I have spent a long time around technology. The path ran from rural Arkansas to the Air Force, from Japan to Germany to California, from radios to computers, from oscilloscopes to Amigas, and from bulletin boards to enterprise infrastructure. It ran from config.sys to cloud, and from half-split troubleshooting to AI agents.
It has been a strange arc. A good one, mostly. It was not always easy, elegant, or cheap, as my wife would still be fully justified in mentioning. But it has been mine.
After all these years, I feel oddly early again. That is the part I did not expect.
AI has made me feel something close to what I felt when I first saw that Amiga in Germany: curious, restless, slightly unreasonable, convinced that something important is happening, and unwilling to sit politely on the sidelines while other people define it badly.
Some of these ideas may age well. Some may not. That is fine. Every signal exists in time. Every reading has context. Every system changes. You tune as well as you can with the instruments you have, then you test again.
Here is what I believe right now.
The future of AI will not belong simply to the people with the newest tools, youngest résumés, loudest opinions, or cleanest personal brands. It will belong to people who can think clearly under uncertainty, combine imagination with discipline, and remember that a beautiful output can still be wrong. It will belong to people who continue learning without surrendering judgment.
Some of those people are old hats. Do not rule them out.
Some of us have been tuning noisy systems our whole lives.
AI did not give my ideas life.
It gave them hands.
Some of us have been tuning noisy systems our whole lives.