Dev,Dr,Dad,Sim,Cloud,Crypto,CTO. map≠territory. Forgiveness+discernment. Jesus showed how to live. Universe is relational. E8 universe + geometry. Let's collab!
Mark Cuban just told every software company on Earth they’re already dead. The people inside them are still building roadmaps.
Cuban: “Software is dead because everything’s going to be customized to your unique utilization.”
Every SaaS company was built on one bet. Software stays rigid. Humans stay adaptable. You learn the tool. You bend to it. You pay for someone else’s version of your solution.
AI just inverted that. The tool bends to you or it dies.
Cuban: “33 million companies aren’t going to have AI budgets, aren’t going to have AI experts.”
33 million businesses feel something shifting beneath them. None can name it.
The distance between what AI can do and what small companies can access is the most mispriced gap in markets today.
Not a technology problem. A translation problem.
Cuban: “Learn all you can about AI but learn more on how to implement them in companies.”
Everyone is racing to build intelligence. Almost nobody is racing to deploy it where the pain is deepest.
The person who walks into a 40-person company and rewires their entire operation captures more value than the team that trained the model.
Understanding pain is now worth more than building intelligence.
Cuban: “Every single job available for kids coming out of school because every single company needs that.”
The most important career of the next decade has no title. No degree path. No university knows it needs to exist yet.
It belongs to whoever learns two languages fluently. The language of a business that can’t articulate what’s breaking. And the language of an AI that doesn’t know where to aim.
33 million companies. Zero translators. Whoever arrives first doesn’t enter a market. They create one.
"LLMs are weak at durable truth, authority, repeatability, and accountability. This codebase is trying to supply exactly those missing pieces." - Codex
@unclebobmartin I've noticed AI tends to go for the over the top implementation/solution. Good in certain contexts. You need to do a joint discovery/scoping session before build begins.
LLMs are great when you have tests. Watching them work they do a form of tree search. They don't show the kinds of "generalized experience" in their approach but by going down a path and then backing out constantly it can eventually find the path through.
I wonder if they don't know this because most of it is tacit knowledge and programmers only commit the happy paths. I assume they are trained on lived examples and thus will naturally get better with time...only because people are using it. Sucks if you aren't codex or Claude at this stage maybe...