Forget Vibe Coding, we're all about Vine Coding nowadays It's been a while since our last Flame Of The Week, but it appears that AI is generating some strong feelings among our beloved readership. Last week, we published a short piece noting the Raspberry Pi Foundation's position paper on why it is still important to teach kids how to code in the age of AI and the advent of Vibe …
This post has been deleted by its author
When you code, use AI like you use a spelling checker, especially autocorrect.
Whomever blindly trusts the autocorrect while texting or emailing will have a short but "interesting" social life.
Anyone who trusts AI generated code will have a short but "interesting" career in programming.
"short but interesting social life."
Given the number of words that rhyme with luck I can easily see how autoconfuse might sabotage your social life.
Texting "After work I am going to muck about with a few girls from the typing pool" while excusing the anachronism and meaning a few drinks and gossip with Mrs Slocombe et al., autoconfuse and muck isn't likely to always end well.
... like you use a spelling checker, especially autocorrect.
In another life, I occasionally used spelling checkers, albeit to (very) limited success.
It was not long after that I realised that the device could only be as good as the dictionary it was using.
And then only if I thoroughly read everything over before finally accepting the corrections.
As most if not all of the time said dictionaries were utter crap, inexorably leading me to search for the correct data in a good unabridged dictionary, I gave up.
Aa a result I polished the knowledge of the English language that Mrs. Fowler tried to drill into my head, gods of literacy bless her.
Automatic correction?
Never dreamed of using it.
.
In my experience ( your mileage may vary) vine coding is when one partakes of the fruit of the vine prior to coding.
Personally as as young programmer, and a good part of a bottle of jack daniel’s ( technically not of the vine, but related) I did have great results.
Not part of my general m.o. but…. could this be what the letter writer was referring to? Defo sounds like it was vine writing at least…
Systemd is a prime example of vine coding. It wraps its tendrils round everything.
Redhat, which was also the leading culprit in the systemd felony, until RHEL5 had a device discovery service called kudzu which apart from causing some peculiar problems,† I assumed was a neologism from some dude programmer.
I discovered it is actually an extremely invasive vine particularly in the US (imported originally from Japan) that rapid gets its tendrils into everything.
There is some similarity between those two services and the way AI is infiltrating just about everything. Can't wait for AI enhanced systemd... actually I very much can...
† on servers kudzu was usually best disabled. That option is unavailable for systemd, alas.
... RHEL5 had a device discovery service called kudzu ...
Interesting ...
These days we have zeitgeist.
It seems to be as kudzu as it can possibly be.
----------------------------
The Zeitgeist Project
Desktop Activity Logging
Providing Desktop Activity Awareness
Zeitgeist is a service which logs the users’s activities and events, anywhere from files opened to websites visited and conversations.
It makes this information readily available for other applications to use.
It is able to establish relationships between items based on similarity and usage patterns.
Features
Zeitgeist logs files opened, websites visited, conversations, and emails and provides all this information over a D-Bus API
Track events by time to figure out exactly when and how often a user accesses something.
Extensions (or as we call them Smack-ins) to provide more information on desktop usage, such as geolocation, focus timer, relevancy, and much more.
----------------------------
Smack-ins indeed .... 8^ /
.
Brought to mind the (I thought pseudo) Latin aphorism "in vino veritas" but apparently from the elder Pliny.
Perhaps vibe (< vibrancy < fervor) we might have the effrontery to recast the Pliny as "in fervore falsitas."
vino appears to be the ablative after "in" hence fervore rather than its accusative. Please pardon my ignorance if I err.
Short of a litany of expletives, AI in its current form is a horrid fetid mess:
* its bad for business that already provide crap customer support, often farmed out overseas, now they use chat agents to frustrate consumers to just hang-up.
* its bad for tech. businesses, more likely to create more work for developers to verify the correctness of code, instead of letting developers do what they do best - create.
* is bad for the environment, given AI monstrosities demand for power will negatively impact climate change further; power that would better serve the public for AC or heating as weather becomes extremely variable.
* is bad for education by creating a generation of youth incapable of critical thinking for themselves or understand how to do anything for themselves; a massive power cut and its not just mobile phones they loose, but their ability to think and deal with emergencies.
Either from the editor's in-tray or from the forum itself.
Thanks for Adam Buxton, I hadn't previously encountered him but undoubtedly a national treasure. I had imagined the missive was from the goose quill of a Reform footsoldier if not Faredge himself.
Also the link to the previous Flame column (2018) which was brilliant.
How the regonomized Norman managed to work Wittgenstein into a comment on the coverage of some tedious typically British administrative bun fight completely eludes me but the University of Woolloomooloo's philosophy department must have been chuffed by the "sloshed as Schlegel" reference.
Thanks for Adam Buxton
I got him confused with Adam Henson. But I guess he practises a different type of AI altogether.
Their trajectory is likely to be one of improvement as time goes by.
I find fault with this. People are always saying that AI will chart a never-ending path of improvement but if anything recent movement in LLMs has been more of a tug-o-war, improving some things and degrading others as new models and new sampling methods push and pull weights to fix specific problems without being able to understand how it will affect the whole.
They've already done that though, they've run out of mass data to train on, aside from the relative dribble that gets produced from now on (going from being trained on all previously discussed philosophical, mathematical, scientific and literary concepts to tweets and tiktoks is a harsh comedown). That's why they've moved on to "synthetic data", training on data created by other models.
Most of the advancement in machine learning in the last 7-8 years (the techniques of which have existed long before the current boom) was made because GPUs are now capable of processing vast quantities of data, which just wasn't possible before. They've gorged themselves on all the data they could get their hands on now though, so the most of the huge advances are done already.
"Their trajectory is likely to be one of improvement as time goes by."
This is what bothers me most in the debate. This very assumption. Just because you pour billions into something doesn't guarantee it's going to get "better". There's a fallacy for that.
What if this tech is garbage for "the dream" of barking technobabble at the Enterprise computer and receiving applications and solutions. What if they're realizing they've promised too much and so now they're just pushing it as hard as they can because there are BILLIONS invested in these companies and the investors want their ROI and "slightly better Google and advanced plagiarism tool" isn't exactly the world changing product they were going for?
I find that the quality and usefulness of AI generated code is directly proportional to the volume of useful examples out in the world (where 'world' == wherever the AI model was plagiarised from. like Stack Overflow and Github, rather than usefully inferred from the technical documentation). Since most of my job involves fairly obscure and/or new things with vanishingly few working examples (I couldn't look them up on Stack Overflow either), I tend to get wildly hallucinogenic material that takes as long to figure out whats wrong and why I need to discard it and start again as it would have to do it myself in the first place. In the business scenarios I usually field, it needs a pretty dumb requirement to get a 'quality' response (something usable with little or no re-writing). That would be of the order of 1-in-20 cases in my experience, but your miileage will vary.
The Register 
Biting the hand that feeds IT
Copyright. All rights reserved © 1998–2025
