Tools using tools

I posted about my project Better Security Questions recently and several people mentioned that instead of actually answering security questions, they generate a random answer in their password managers and use that random answer to log in. Their password tools are now handling all of the logging-in brainpower, rather than the human needing to.

Similarly, as people build new services and APIs and tools, we as an industry aren’t just optimizing for humans knowing how things work or remembering things anymore, but for AI agents and tools to use them.

We have two types of users to build for now, whether we like it or not: humans and machines. Or agents, or assistants, or whatever you want to call them.

Developer experience (DevEx) does still matter, because professional developers ultimately still (for now? Ugh) are making the choices of technologies and tools and architecture. But, now AI experience (AIEx? LLMEx? TexMex?) matters too now, at the rate that we’re going.

How humans and machines find you

I think a lot about this from a content perspective.

With the humans in mind, developer advocates/devtool companies/content engineers/technical content creators (and so on) need to publicly “use all parts of the buffalo” and turn a demo into an open source project, a podcast episode, social media threads, a YouTube video, a blog post, a whitepaper (and so on, again) to cast a wide net for SEO and growth. You have to meet the people where they are, on all the platforms and all the things.

With the machine in mind, you now need to “use all parts of the buffalo” for a different purpose: to seed training data, improve model recall, and ease differentiation from other tools. You have to add things like an llms.txt file or prompting assistance, beyond the content.

How this changes our roles

I think as technical people, we do need to accept that even though it’s early, all signs are pointing to AI being here to stay (and hopefully become a bit more “invisible” than it is right now in our newsfeeds).

As we build, we need to think about how we’ll shape our own developer experience, and the developer experience of other humans. Chances are, a human using your tool/API/utility/library/etc. will have some kind of AI assistant. The humans will lead and they will delegate, and optimizing their experience for both is going to matter.

People will use tools to use tools… and sometimes those people are tools. I’d like to see a machine do an incredibly deep and objectively funny pun like that to end a blog post. HA.