Thursday Thoughts: The AI Gut Check for Startups
At a data and AI event earlier this week, I heard one of those off-the-cuff observations that just sticks with you. A CEO was sitting on an industry panel, and someone asked him: how do you survive and thrive as a startup in the AI world?
His answer was disarmingly simple:
"If you wake up every morning hoping the AI models got smarter, you have a good business model. If you wake up every day hoping they got dumber — or stayed the same — you have a bad business model."
I haven't been able to stop thinking about it since.
The Wrapper Problem
There's been a lot of debate lately about whether being a "wrapper" is a viable business. You know the archetype — a product that adds value primarily by packaging basic LLM capabilities into a nicer interface or workflow.
My honest take? The models will eat that business. Everything that lives at the surface level of what a frontier model can do will eventually become the frontier model. It gets commoditized, absorbed, and shipped as a default feature. If your moat is just "we made GPT easier to use," that moat gets filled in pretty quickly.
Surfing the Frontier vs. Racing It
What you actually want is a business that surfs the frontier — one where as the underlying model gets smarter, your product gets smarter too. Your value compounds with every new model release instead of eroding.
That's a fundamentally different design philosophy. It means your product isn't competing with the model. It's extending it. It's taking that raw capability and doing something with it that the model alone can't do — whether that's deep integrations, proprietary context, specialized workflows, or something else entirely.
I'll be honest — this was a useful gut check for me personally. I thought about what we've built at Coder and whether it passes this test. I won't turn this into a product pitch, but I do think we've structured things so that better models make our product better, not redundant. That's the north star.
The Wrong Question About AI and SaaS
There's a lot of hand-wringing right now about whether AI is "eating SaaS." I think that framing is a distraction. It sends founders down a defensive path — trying to figure out how to survive the onslaught rather than how to ride it.
The more productive question is:
Does your business extend the value of frontier models, or does it try to stay ahead of them?
If it's the latter, the clock is ticking. The frontier moves fast, and staying ahead of it as a startup is an exhausting, expensive, losing game.
If it's the former — if you're genuinely amplifying what these models can do in a way that grows with them — that's where durability lives.
Simple litmus test, but I think it cuts right to the heart of what makes an AI-era business worth building. Worth asking yourself honestly: which camp are you in?