UIs with buttons and charts are not cool anymore. Haven’t you heard?
All that matters is AI agents now. APIs, CLIs and MCPs – that’s all you need.
It’s all about the data. Just get me the data, bro. Where is your MCP?
AI can make charts and tables for me. I don’t need your dumb dashboard anymore.
Stop slowing me down with all this UI cruft.
Browser? No thanks Boomer.
Just let me hook the data up to my agentic workflow.
You should build software for agents now, not humans!
—
Or so the discourse goes.
But what does that look like?
What does it really mean to “build for AI”?
Does that just mean some data connector app that lets you run queries or perform some actions via an API, CLI, and MCP?
Do I really not even have to build a UI anymore? What does that mean about the value of my product or what I can charge for it now?
What about this data… What specific data is most useful? Do I own or control this data? What if I don’t? What types of questions are people going to ask the UI that I need to account for in my MCP and AI tools?
I see a gap in the market where competitors are not using AI effectively. I could build something and maybe get traction in the short term, but what if the owner of that data or the competition in that space release their own AI integrations, making my app unnecessary?
What if Anthropic or OpenAI build a thing for that type of data or that use case and integrate it directly into their own apps (like Claude Design)? The AI labs have already shown they are more than willing to expand well beyond just being model providers. Why would their ambitions stop at whatever my idea is?
What does a “relatively safe” idea even look like anymore? Is there even such a thing?
What could I possibly build that won’t eventually be swallowed up by AI?
—
This thought loop has run in my head countless times the past few months.
Sure, I can build “X, but for AI agents”, or “Y, but with AI integrated into it“. But how long is that really going to last? Is the vertical large enough for me to make a viable business out of it, but small enough that the labs companies will ignore?
I can build all my ideas faster now using AI, but it’s less clear than ever what that thing should even be.
—
This AI tidal wave hitting my profession so directly makes me anxious. I have embraced AI more enthusiastically than most other developers I know, but even I find myself often pondering all these unknowns and not being able to arrive at any solid answers.
On one hand, AI is an amazing tool. I am really excited about the ability to build anything I want, quicker and easier than ever. It unlocks so much more than any other technology I have ever seen in my life.
At the same time though, I am also increasingly unsure what exactly to build that would be competitive in the AI age. Many of my current ideas are no longer viable in the same way they were before. They look more like traditional SaaS ideas, or simple tools for common tasks that AI now makes trivial. My ideas were not imagined in a world built for AI agents. I’m going back to the drawing board with all of my old ideas. Some might still work with tweaks. Most I discard entirely.
The only sure thing seems to be helping other companies implement AI.
The future of software is a blank page. A complete paradigm shift. The only thing I am certain of is that it won’t look anything like the past.
The dust will settle eventually, but the future is very unclear in software right now, and we’re all feeling it.