An Inevitability of AI Doing Something

Conversations here usually start with a whiteboard and often a sketch or several. There’s some back and forth about everything from aesthetics to outcomes, with usually some kind of boundary offered (“do you want this now, cheaply, or well”). All in all, the conversation around shaping systems and behaviors to intended goals follows a fairly straightforward conversation. The rise of AI tooling seems as if it changes this process, but does it?

Been lightly playing with Replit for an idea stuck in the brain for sometime, and it’s mostly just fine. The ability to use it on iPadOS is a delightful break from some of the other LLM-coders. And the closeness to “building” to “analyzing the user’s intent” is a delightful change from bending into the technical architecture for every feature or purpose. That said, this doesn’t seem like a futuristic shape of building, deploying, and maintaining applications as much as this tooling feels something more “on the way to a final form.”

In conversations and prognostications about what comes after smartphones, there is also a “what comes after apps” context. Where anyone with a handle of natural language and their focused imagination would simply be able to cast their words and gestures into a pane (canvas, glass, camera, and/or microphone) to get what they desire. What does anyone desire? Reduced friction. Reputation being enhanced. Endorsement. Compensation.

They want to better ensure their survival.

Many folks are correct in saying that learning the how and why of machine learning tooling is needed now. Many anchor this in learning the models, or learning what exposed/sold/open sourced models can do. Some anchor this in the many shapes of generative programming around those models - prompt engineering on one end, workflow management on another, and a gradient of usage in between. And this works for now. It doesn’t create what will be… nor can it. Tooling of this moment is only good enough to ensure compliance and comfort for this part of the change. Tools are not eternal.

As a result of using AI tools, some work will change forever. Much like inter-office email changed as soon as email became as easy to finance as a computer on every desk, there’s some seemingly necessary behavior which will be changed because of AI tooling. It won’t be covered in a golden folder. Nor will it require signatures. It might have its own language, security, and likely, domains previously only imagined in the past’s view of the future. It might not even be in a language currently known.

Change is inevitable. And for better, worse, and different, the increasing use of this type of math (models, tokens, calculus, and energy) will be a pallet for some, and a museum for others. It might also birth moselums for what is held dear right now. Learn what you can… adapt, build, and reinforce. And then see what about AI tools actually work forward.