Musings on No-Code and LLM Prompt Skills

I’ve been giving a lot of thought to “no-code” recently. The term is back in vogue. And for good reason. Over the years, I’ve had a front-row seat for the tarnishing of the “no-code” term. In 2022, “no-code” was often synonymous with “very complicated user interface”.? So much so, that learning to do a little scripting was often the better choice.

But that world is gone. Obliterated. A logic-competent, strong communicator can make magic with an LLM-enabled product. A no-code product. Even at Okareo, a developer tools company, we are providing customers the option to use natural language to evaluate how a model is expected to behave. Of course you can still upload code, but developers are leaning into the narrative form.

Recently we have had the privilege to partner with Wayfound.? The amazing Wayfound team is building an agent platform that business users can train to have meaningful, action-full conversations with customers. These capabilities were impossible in 2022. The key is clear expository writing. We know because Okareo auto-evaluates agent behavior.

Now, I’m not one of those folks prognosticating the end of coding - at least not yet. But I am seeing a rise in language-based customization powered by LLMs. Increasingly borad and meaningful customization of software behavior through no-code, all-language behavioral description feels on-track to become a norm.

So that was a lot of preamble to where I’ve been noodling around. With two young adults in and nearing college, I’ve been giving a lot of thought to the next generation of critical skills. Is this a moment to focus on a writing and philosophical metalogic education with some systems knowledge thrown in. Is that the new powerhouse skillset? Is this a time to dive deeply into abstract subject matter expertise? Perhaps depth and narrative for subject-specific outcomes will be more powerful than strict engineering. That was pretty abstract. Learn to write. Learn a subject deeply. Learn to think about systems. Don't worry about coding.

Consider the parallel between prompts and first-time engineering manager instructions. The common and unhappy pattern of the first time engineering manager is to apply coding habits to management. We call it micro-management. The same habit of applying code thinking to prompts is also common. In management, experience brings space to trust the team member’s independent thought which leads to creativity and innovation (not to mention happy team-mates). Perhaps the same is true with AI/LLMs.

If we shift our thinking to the LLM as team-mates, then we would approach prompts differently. Instead of long micro-managing prompts with highly specific instructions, exceptions, and pseudo-code, we should provide contextual information and broad guidance. Certainly, when I take a step back and give the LLM some “space” the results are often far better than expected. To do this, well, I just wish I’d spent more time learning to write.

So, aside from only giving advice to my kids when asked (it’s a journey), my noodling bet is that no-code customization is soon to be a norm; that expository writing skills are about to be critical skills for all; and that if we want happy AIs, we should give them a little more freedom of choice in how they accomplish our asks. And yes, I want happy AIs.