I have been writing code for 45 years. I started when "memory" was something you counted in bytes, not gigabytes, and when a "bug" was nearly liter...
For further actions, you may consider blocking this person and/or reporting abuse
Great point, although computers don't always seem to behave purely logical and reproducible from a human obverver's point of view. Computer interaction often sens unintelligible cryptic error codes hard to trace down to a root cause, and race conditions cause erratic behavior. Still I agree that generative AI is much worse and less predictable.
Best article I have ever read on this topic. Very well put.
Wow, you just hit on something so cute I havenβt been able to articulate! π€
The framing of prompting as negotiating rather than programming is the sharpest observation here. After 45 years of deterministic feedback, being asked to "massage" a probabilistic system is a fundamentally different cognitive contract β not just a new skill to learn. fwiw I've found that the more structured and explicit you make your prompts (almost like writing a spec), the more predictable the output becomes. Which is kind of ironic: the people who think most like compilers may actually end up being the best prompt engineers, once the tooling catches up to reward precision over persuasion.
I'm implementing an AI code-gen platform, and my experiences is that citizen_can_ code, fairly complex things, and will such eat from our market - But you come much further if you actually know some coding, and I'm not sure if that will ever change.
I started programing 54 years ago, so I know that I started when "memory" was something you counted in bytes, not gigabytes, and when a "bug" was nearly literal is exagerated by some 30 years. The famous bug incident happended in 1947, debugging first occurs in print in the early 1950s .
I have now been using copilot to help me build a nostalgia project (a complier and IDE for the CDL2 language) for about 1/2 year. It is tremendously helpful ... assuming that it is treated like a junior programmer supervised by one with a lot of experience ... me. Some examples.
I wrote a code generator for the project that generates PowerShell code (because that made it relatively easy to debug the generated code). I then asked copilot to generate the C# code generator. That took about 3 minutes. It had two bugs which copilot helped me track down.
About two hours ago I decided I wanted name completion in commands for the IDE. The command syntax is quite elaborate, It needs three separate kinds of completion. With copilot's help (and it did need quite a bit of explanation) the job was done in two hours. I think it would have taken me a day or two if working alone.
Memory was jealously counted in bytes even 40 years ago. Do you not remember having to write safety checks around malloc() for even simple data structures? 30 years ago gamers were tweaking their config.sys to get a couple of KB more so their sound would work at the same time as the rest of the game.
I suppose your worst nightmare then is quantum computing. :-)
Between the lines, you're describing the fear of change, our world is crumbling, hours and hours spent learning things that will soon be worthless.
And the security of the predictable, knowing what's going to happen, is reassuring for everything, both when facing the keyboard and when facing a future that's changing faster than we might be able to comprehend.
It's not a condition exclusive to neurodivergent people; everyone is at that point, I think.
I love this! My Neovim has been my sanctuary for many years, just like you described. I'd never attributed this to how it required zero Theory of Mind. Great read!
I liked while reading this blog and learned new things