If you think talking in plain English to an LLM (large language model) will produce a coded solution that does what you want, does it well and will stand the test of time, you’re wrong.
When anyone first tries to learn to code, it’s hard. Everything’s alien, and even though the programming language uses English words, it might as well be hieroglyphics. Many people give up, and even the ones who persevere, still long for the ability to just say what they want without having to shape it as code.
So has it always been. Machine code programmers longed to enter the assembly language they’d scribbled down on paper. Assembly language programmer’s wanted something to input their pseudo code and have it turned into assembler. Now people writing in high level languages long for something closer to English.
Then along come LLMs. Finally we can just say what we want and the machine does the rest. To the casual observer it looks like that’s exactly what we have now, but they’d be wrong. You see, the hard part of learning that first programming language wasn’t the language at all. Any decent programmer can learn their second, third and fourth languages in no time at all. A week is all it takes to start working effectively in a new language. So why did the first time take so long?
Because you weren’t just learning a programming language – you were also learning how to programme. You were learning how to take a vague idea and shape it into something detailed and coherent that actually solved the problem. The language wasn’t the problem – that was just how you explained your answer to the computer. That’s why you can’t just talk to an LLM and get a working solution. It doesn’t understand English. Many would have you believe you’re now the product manager and your agents are the developers, but they’re wrong.
Contrary to the biased and, to be quite frank, offensive attitudes of some non-developers you can talk plain English to human developers. You might think you’re being clear and concise, but trust me, you’re not. That’s why you found coding difficult – it forces you to be intricately accurate, and that’s incredibly hard.
Luckily for you, those developers will take that vague, hand waving, English and turn it into something concrete; something implementable in code. To achieve that, they’ll do a lot of thinking that you haven’t done and they’ll probably ask a lot of questions. You might even change your mind about what’s needed as a result. Then they’ll write the code. If you hired the right developers it’ll work and you’ll have a great product.
LLMs change nothing. You still need to explain what you want in concise, unambiguous instructions and to do that you have to know how to programme. That’s what an experienced developer is doing when they drive their coding agents. That’s not plain English they’re typing. It’s code that looks like English.
That’s why AI doesn’t replace developers. If you thought it did, and you got rid of them, you’re in for a very nasty surprise.<\/p>\n","protected":false},"excerpt":{"rendered":"
If you think talking in plain English to an LLM (large language model) will produce a coded solution that does what you want, does it well and will stand the test of time, you’re wrong. This post attempts to explain why.<\/p>\n","protected":false},"author":1,"featured_media":217,"comment_status":"closed","ping_status":"open","sticky":false,"template":"","format":"standard","meta":{"footnotes":""},"categories":[9,1,23,46,60,30,10],"tags":[6,27,36,115,31],"class_list":["post-430","post","type-post","status-publish","format-standard","has-post-thumbnail","hentry","category-artificial-intelligence","category-business","category-culture","category-discrimination","category-disruption","category-software","category-technology","tag-ai","tag-business","tag-communication","tag-programming","tag-software"],"aioseo_notices":[],"aioseo_head":"\n\t\t\n\t\n\t\n\t\n\t\n\t\n\t\n\t\t\n\t\t\n\t\t\n\t\t\n\t\t\n\t\t\n\t\t\n\t\t\n\t\t\n\t\t\n\t\t\n\t\t\n\t\t\n\t\t\n\t\t\n\t\t\n\t\t\n\t\t\n\t\t\n\t\t\n\t\t\n\t\t\n\t\t