Essays • AI and engineering • • 8 min read

AI Development Revolution Part 6: Future Impact

What building with AI cost me, the wrist pain, the late rebuilds, one brain holding every role, and what that means for solo founders and for teams.

Illustration of a man seen from behind facing a glowing city skyline under the word Future, surrounded by floating chart panels

The AI Development Revolution: Part 6 - Future Implications

Part 6 of the AI Development Revolution, about what the building cost and where I think this is going.

The cost behind the numbers

The numbers look good, lower cost and faster builds, and what they leave out is what it took from me to get them.

I slept badly through the sprints. I spent whole days validating AI suggestions one at a time until I could not tell a good decision from a tired one. My back hurt. My eyes hurt. I lost track of people because the work wanted all of my attention and took it.

I wrote about this in "The Multithreaded Mind". Working across every layer at once is a different way of being awake. Old development let me sink into one thing. This kind asks me to hold architecture, implementation, testing, business logic, and user experience in my head together, all the time. The mind is not built to run that hot for long.

Every suggestion needs checking and every decision sits on top of the last one. The real bill is managing context across a solution that keeps changing shape under you, doing quality control at a speed that was never meant for a person.

Cheap to make, expensive to carry

Enterprise software used to need an enterprise team. Now one person can build the thing that used to take a dozen specialists. What once cost hundreds of thousands costs tens.

The cost did not disappear, it moved. The money is low and the attention is high, the team is small and one person carries all of it, the iteration is fast only while someone stays at that intensity. You can afford the work now and you pay for it somewhere else, and the somewhere else is a person.

What this does to the work

Agencies were built on dividing the work by specialty. AI flattens those divisions, and that sounds like a win until you live on the other side of it. The coordination overhead a team used to absorb falls on one person. The timelines compress, but everything the old timeline held still has to happen, now in less time, from fewer people.

The human is still the bottleneck, and AI speed only makes that clearer. More projects means more validation, and validation does not scale the way generation does. The early adopters I know are tired in a way that is starting to look structural.

There is also a knowledge problem. When the workflow is this personal, almost none of it transfers. Decisions go undocumented because documenting them at machine speed is one more thing to do. Senior developers watch juniors with AI skills move faster than they can, and it shakes something loose in how they see themselves. The hard question for agencies is whether they can keep the people intact while they adapt the tooling.

Startups have the same shape. A solo founder can ship a real product now, no team required. The technical barrier is mostly gone and the personal one is still there. I built enterprise infrastructure in weeks, and those weeks hid the rebuilds, the late nights fixing what the AI broke, the wrist pain, the cancelled plans. I was architect, developer, tester, and manager in one overloaded brain. The risk moved from the technology to the person. Whether you make it depends less on funding than on how much sustained intensity one body can take.

Enterprises ran on being slow and expensive and predictable, and AI takes the predictable part away. The coordination looks simpler from the outside, but it hides how much complexity one person is now holding. The limit now is endurance.

The skills flipped

Knowing how to architect now matters more than knowing how to type the code, judging quality more than producing it, direction more than memorizing a framework the AI already knows cold, and that order changed fast.

The real skill became working with the machine, holding context across sessions, learning how it tends to fail, checking its output at a pace that pushes against what a person can do, inventing a way to work that had no manual yet.

Manual coding starts to feel old-fashioned. Memorizing frameworks feels pointless when the AI has all of them. Years of hard-won expertise can suddenly count for less than a few months of knowing how to collaborate with the thing.

It hurts. Senior developers asking what they are now worth, juniors moving up on AI skill, expertise that took a decade feeling thin next to a machine that does the same task without breaking a sweat. Adapting to this means rebuilding who you thought you were at work.

The market underneath it

Solo developers compete with agencies now. One consultant delivers what a team used to. A personal brand can pull work away from an established firm. The technical part of the service is becoming a commodity, and strategy is the part still made by a person.

You can validate a business with a working system instead of a pitch deck. Test an idea in days. The barrier to entering a market keeps dropping, and the cycle keeps shortening. Custom builds are starting to cost less than buying something off the shelf, which is not how any of this was supposed to go.

Education is behind. Memorizing facts is worthless when the AI holds all of them, and collaboration matters more than the facts, but the curriculum was written for the old world and is out of date before anyone graduates, the old shape dissolving and the new one not set.

Faster used to mean worse

It does not anymore, at least not the way I expected. Building faster with the AI, with systematic checking on top, gave me higher quality. Security went in from the start, optimization kept happening all along, and the machine caught things I would have walked past.

Quick iteration surfaced better answers because I could try several where the budget used to allow one. The quality came out of the speed. I did not predict that, and it changed what I expect from a build.

Wider than my desk

It reaches past the industry. A personal project can reach professional quality now. A small business can touch tools that used to belong to the giants. Entry barriers fall and competition gets sharper for it.

Developers become architects because the work demands it. Founders get freed from the risk of technical execution. The whole system speeds up as friction drops out of it, and it is happening across the board.

The things I do not have answers for

Quality control at machine speed, with a human as the slow step. Systematic checking helps, but it does not solve it.

A widening gap between people who have the new skills and people who do not. Traditional developers feeling written off. Training that cannot keep up with the change.

Business models breaking faster than anyone can rewrite them. And a quieter risk underneath all of it, that the more the machine handles, the more human expertise wastes away in the areas it touches. The knowledge survives only if someone decides on purpose to keep it.

These are real and I do not have the fix.

Where I think it is going

I will not give a timeline, because every timeline I have seen has been wrong. But the direction is visible in fragments. Game development is already showing AI-first architecture, interfaces designed for the machine to work against, metadata pulled out clean, patterns shaped for collaboration instead of for a human reader. That is Part 7.

The early adopters are pulling ahead while the rest of the field is still arguing about whether any of it is real. Mainstream adoption is coming. When, I cannot say.

The predictions keep missing the human in the middle, the cognitive cost that compounds, the question of whether any of it is sustainable. I know the direction and not how fast, and I have stopped pretending I do.

What I would tell you

Experiment, and know what it costs. Write down the failures along with the wins. Build with the price in mind, start small, measure the whole picture, and put the investment into people before tools.

This changed me along with the software. My capability went up while my capacity strained against it. Speed and quality stopped fighting each other. Whether any of this is sustainable, I still do not know.

Along with how software gets built, I am finding out what a person becomes when they work this closely with a machine. The promise and the price showed up together and I have stopped expecting one without the other. The future comes whether I am ready or not.

Next: Part 7: Advanced Patterns →

Related Reflections:

Series Navigation:

AI development