Mostly.
Brian Martin
The age of dynamic software is upon us: systems generated to fit a real business, owned by the people who use them, cheap enough to rebuild or replicate when the work changes. Others have circled the same shift under names like disposable software or Software 3.0. Fine. What matters is the economics. With AI in reach, why buy sealed, vendor-priced, one-size-fits-most product when you can duplicate the effect yourself—aligned to your workflows, running under your control?
I switched accounting packages last year because I was done with pricing strong-arm tactics from a vendor with poor management practices. The company we moved to was better than what we left. It still wasn’t what we wanted. It interfered with how we actually work.
Last week I asked Grok to build an accounting package from a clearly defined feature list. I now have that package running locally in Python, not in an NSA-monitored cloud, for less than the annual subscription of any of the alternatives.
Because it is my software, I can replicate it across my businesses and fold cash-flow reporting for my accountant into one clean picture with almost no extra effort. For six companies, the whole stack will have cost less than a single year of the original vendor’s subscription. Around a hundred dollars.
What do these legacy vendors have left to sell? What feature, what integration, what “roadmap” would make me prefer their meter over a tool I own? Decades of mediocrity, priced like scarcity, look different when scarcity is gone.
Dynamic software changes a lot. It does not change everything.
Experiential software (games, for example) still earns its keep. Foundational tools like Python and MySQL still matter, for now. Even those layers will eventually feel the pressure of what people loosely think of as Moore’s law. Moore’s law is not the right instrument, but it is a useful rule of thumb for compounding capability, much of which is riding on hardware progress underneath the models. Will regulatory revenue capture slow the shift? The dinosaurs will try, but this can’t be put back in Pandora’s box.
That is not to say that software architecture or computer science are dead. It is a question of who the field is for. Does it become, again, work for people who actually think, or remain a daycare for adults who learned to rearrange widgets? Most of us, myself included, will find it enough to ask an AI for a working solution and then get back to the business we are actually in rather than continue refining our tech skills. Our knowledge will enable us during the paradigm shift and help us usher in the next tech revolution.
Will “learn to code” go down as one of the dumbest pieces of career advice of its era?
As functional software approaches free, not even a commodity in the old sense, what happens to the venture playbook built on metering it? How far does that logic reach into physical products and services? This brings us to the robots.
Imagine a router: firmware drafted by AI, chip layout largely AI-assisted, board and packaging AI-assisted, with humans still doing high-level marketing and the handshake. How much is that handshake worth when buyers can stand up comparable capability without captive lock-in? Can a hardware vendor still strong-arm an account and repress a competitor when the customer can obtain both units for essentially the same price? What does that do to hardware vendors and to the venture capital that used to sell access to the deal?1
VC has always been about distribution, about forcing a product into a marketplace. That was the essential function of many board seats and many venture relationships: access, placement, and sometimes the quiet suppression of rivals. When the playing field flattens, what is left of that power?
Today I can conceive of a small robot for a particular task and, within days, have a prototype in hand. Volume and sales channels still constrain scale — for now. Are those permanent moats, or fading inconveniences? My robot will also carry the legacy of all the hardware cycles that came before, and have maximum capabilities built in for little additional cost. Self-charging. Standard interfaces. Security.
So where is the durable play? Ownership of the robots and manufactories. To quote Rage Against the Machine on Down Rodeo: “[radio edit] the G-ride, I want the machines that are makin’ ’em.” And now those machines can live in my garage, a virtual/robotic Apple Lisa assembly line.
Brian Martin