I am cautiously pro-LLM in my own work, but collectively I fear the car is the predictive historical model: small wins for individuals + huge wins for a handful → massive externalities and collective action problems that will take generations to fix.
From: @straphanger
https://urbanists.social/@straphanger/115140521581949962
You’re hosting a birthday party for your son, and the other kids are upset because he’s taken the entire cake and shut himself in his room, refusing to share it. Some kids are quietly collecting up the crumbs while others look like they’re about to riot.
You could demand he comes out of his room and shares the cake, but you have a different idea: ignore the chaos around you and try to bake another cake. You wouldn’t give the second cake to the other kids though: it’s your son’s birthday, so it should still be his cake. You’re sure that once he has two cakes he’d share some with the other kids.
You are a politician rejecting directly redistributive policies in the vain hope that encouraging economic growth alone might reduce inequality.
Here’s a nice little workflow that’s possible with #jujutsuvcs but not #git: I can create future empty TODO commits as reminders to implement some logic
I wonder whether cleaning up vibe-coded prototypes results in code that is easier to understand a year in the future by others because the dev has had to pick apart and comment the messy code enough to understand it, whereas if they were write it all themselves they would no doubt consider it perfectly explanatory without sufficient abstractions and documentation.
I got sucked into running the numbers on that "delete emails to save water" thing. Best estimates I can find are that live datacentre storage in the UK has a median water usage of ~80ml/GB/year. So a terabyte of cloud storage consumes 80 litres a year.
Network losses from leaks are on the order of 10-15,000 litres per person per year.
Glad we can see the culprit is definitely old forwarded cat photos.
It took me nearly 10 years between writing my first HTML and JS, and understanding things like ARIA attributes and CORS, simply because you don’t know what you don’t know. If I’m hiring a frontend dev, how can I know that they’re familiar with everything without asking 1000 questions? When I’m helping even a senior dev with decades of “full stack” experience, how can I approach a conversation without a firm grasp on our shared understanding? When writing a blog post or a tutorial, how can I describe the target audience without falling back on the lowest common denominator? We NEED qualifications
The lack of qualifications/accreditations in tech not only makes hiring frustrating but it also means as an industry we’re in an endless cycle of knowledge sharing; always having to re-invent the basics, to re-share the fundamentals to a new cohort of developers. We need a baseline shared understanding for developing software in each domain — when someone says they’re an X developer we should all know that they know the fundamentals of X and they should have confidence to know that they’re aren’t any holes in their knowledge of X.
Software engineer by trade. Programmer by hobby too (in addition to basketry and spoon carving). Personal website: https://rlamacraft.uk/. Gemini capsule: gemini://gemini.rlamacraft.uk