I've been thinking about where all this AI development is heading.
Building is becoming easier. Knowing what to trust is another matter entirely.
We're seeing tools like Claude and ChatGPT make it increasingly easy for people to build things that, not so long ago, would have needed a development team, a fair bit of technical knowledge and considerably more time and money.
That's a good thing. It opens doors, gives people the opportunity to experiment and means more ideas can actually make it off the drawing board.
More people can build
But there's something else happening alongside it that I think deserves a bit more attention.
Just because you can build something doesn't mean you've established that it works.
And I don't mean whether it works when the person who built it is using it, or whether it looks impressive in a carefully prepared demo.
Has it been established?
I mean what happens when somebody else gets their hands on it.
Does it do what it's supposed to do? Does it behave consistently? What happens when someone uses it in a way the creator didn't anticipate? Where does it struggle? And are the claims being made about it actually supported by what the product can deliver?
These are the things I'm interested in.
What happens after something has been built
There's a lot of focus on what AI can help us create, and rightly so. But I think we're going to need to pay just as much attention to what happens after something has been built.
Because there's a difference between testing something during development and having someone independent come in, look at it with fresh eyes and put it through its paces.
AI can help build the product. It can help test it too. But I'd argue there's still real value in having a human approach the finished product without being invested in its success.
Not to catch someone out or look for reasons to give it a bad report. Quite the opposite.
To establish what works, what doesn't, where the limitations are and what someone considering using it should realistically expect.
That's the thinking behind CrashTest AI and increasingly where I see the opportunity.
Ready to be trusted?
As the barriers to building AI products continue to fall, the number of products entering the market is only likely to grow.
But how do we know which ones are ready to be trusted?
I don't think a good demo, a convincing landing page or the creator's own testing is enough to answer that question.
And I suspect this becomes a much bigger conversation as AI development continues to accelerate.
AI is making it easier to build. That doesn't automatically make it easier to trust what gets built.
That's something I intend to keep exploring.
Stay tuned.