Warp’s Post

If you squint, the AI dev tools are sort of like a compiler, which isn't a new idea. They take some form of human language and translate it into software. They're arguably the most advanced compiler-like tool ever written. But, I imagine that that's been said about a lot of compilers in their time. In the meantime, coding is often the easiest part of engineering. Building useful software is more of a discovery process, either through technical limitations or product requirements. Even something like translating unknown production issues into bug fixes doesn't lend itself to the current crop of AI Engineering tools. In the meantime, having AI tools to help me auto-complete documentation, write boiler-plate UI, or help code around some obscure hardware edge case, isn't the worst thing.

As an engineer, I’ve come to think of LLMs like a souped-up keyboard with an infinite number of macros. Say you want it to respond with “The quick brown fox jumps over the lazy dog.” You could prompt “respond only with ‘The quick brown fox jumps over the lazy dog’” or just “respond only with the most common English sentence containing one of each letter from A-Z.” But you could also prompt with “respond only with the top 10 most common English sentences containing one of each letter from A-Z.” I just ran that through Claude, and here’s what I got: The quick brown fox jumps over the lazy dog. Pack my box with five dozen liquor jugs. Sphinx of black quartz, judge my vow. How vexingly quick daft zebras jump! The five boxing wizards jump quickly. Jackdaws love my big sphinx of quartz. Quick zephyrs blow, vexing daft Jim. Two driven jocks help fax my big quiz. Five quacking zephyrs jolt my wax bed. Waltz, nymph, for quick jigs vex Bud. A lot of people seem to think LLM are one-shot tools, like a search engine that can also compose sea shanties. But once you understand how they work and how to coax different kinds of output from different models, it starts to feel like Just Another Macro.

This tracks. Would love to hear your thoughts on test automation in the LLM dev future.

point 2 is spot on

This is gold

#5 is 100% spot on.