RE: https://mastodon.social/@colincornaby/117069185435215102
Holy shit, so the "Claude plagiarised an app" story is actually "Claude produced the same app twice when prompted to do so by two separate developers" 😅
Remote
Nick Lockwood
@nicklockwood@mastodon.social
mastodon 4.8.0-nightly.2026-10-06iOS. 3D graphics. Retro games. He/Him.
3403 Followers
732 Following
50 Posts
Joined April 03, 2017
RE: https://sauropods.win/@futurebird/117291573233666795
People sometimes like to say that the state of AI right now is "the worst it will ever be" (meaning that the models will get stronger and problems like hallucinations will be reduced, etc) but this fails to take into account enshittification
AI _technology_ might be the worst it will ever be, but at any point now the AI companies are going to decide it's time to start profit-maxxing, and at that point the _products_ are gonna get real shitty, real fast
Quoting
The Untouchables
This pitch calls them the "untouched audience." These are people who avoid ads. Sometimes I wonder just how large we are? I think it is likely that a significant portion of people with disposable income, the exact people advertisers want to reach, are also doing their best to avoid ever seeing any of those ads.
OpenAI has been planning to introduce ads to chatGPT for about a year. Sounds like they have done some trials already.
Were the ads even clearly identified as ads?
Replying to
Incidentally, the crash was due to a missing dylib - something that could easily have been detected by the automated scan process, if only that scan were designed to protect users' and developers' interests instead of Apple's
"Trust this computer?" my phone asks me, when connecting to my personal laptop for the 67 millionth time
I keep seeing this claim that software has suddenly become more buggy and unreliable since the advent of agentic coding, and I'm like… have folks just not been paying attention for the last decade?
Software has been getting progressively buggier and more unreliable every year for as long as I can remember
I'm not saying AI isn't increasing the rate of garbage code, but I do wonder how you can tell the difference between vibe coded shit and regular organic shit, because it smells the same to me
🤣
Replying to
I guess rather than vagueposting I could actually explain how they are broken since it's not documented anywhere:
List - best all-round solution, but resizing items is janky and sometimes leads to random scroll offset
VStack - horrible performance for more than a few items
LazyVStack - smooth scrolling and resizing of items, but programmatic scrollTo() doesn't work reliably because items are lazy loaded
Replying to
The handwringing about the implications for open source codebases is also especially hilarious
I'm sorry to be the one to break this to you, but the answer to the question "does this mean people can now take my MIT-licensed code without attribution if I used a coding assistant?" is that they were already doing that, long before LLMs
If you're lucky, that will continue to be the case. If you're unlucky, then everyone will just vibe code their own solutions and your project will die in obscurity
Sharing this again because it's so fucking good: https://pluralistic.net/2025/12/05/pop-that-bubble/#u-washington
It's long, but only because it makes so many brilliantly insightful points. I thought I remembered these points being spread across at least three different articles, but they were all in this one
Replying to
But the idea that natural language makes it easier for non-programmers to program is a misunderstanding
Programming is not about transcribing English to code, it's the art of turning vague requirements into concrete ones - identifying and filling in the blanks so that an imprecise spec becomes precise
When you "vibe code" you are asking an LLM to do that work - and they are remarkably good at it - but that is not *programming*, because you are not identifying and eliminating those ambiguities
tfw you're signing up to a website and the email verification link takes so long to arrive that you dissociate and forget what you were doing until 3 hours later when the session has long since expired
Replying to
@krzyzanowskim if I were a human, I would simply become a rich human, then all my problems would be solved
Replying to
@AshtonDev I guess the in-store apps would be pre-subscribed, so they would just not mention it at all in the session and let the kids be disappointed when they tried it at home later
Replying to
The job of a programmer is to identify ambiguities and make decisions about the correct thing to do
High level DSLs make programming easier because they eliminate ambiguity by baking in sensible, deterministic, domain-appropriate decisions, so the programmer doesn't have to make them
LLMs make programming easier by YOLOing all those decisions in a semi-random fashion so the programmer isn't even aware that a decision had to be made at all
Replying to
@carnage4life@mas.to I actually think this was true for many features even before LLMs. I once implemented a feature while I was bored during the zoom call about the feature
In SwiftUI, if I'm using ScrollViewProxy to scroll to a particular anchor in a List, is there any way to have it take the top contentMargin into account?
Every time I scroll to an item, it ends up behind the top toolbar 😞
Replying to
In programming we tend to think of high-level languages like Python or JavaScript as being "easier" than low-level languages like C, and that's true, but it's not because they are closer to English
It's because they eliminate (or otherwise smooth over) more ambiguities at the syntax level, instead of allowing the programmer to write ambiguous or erroneous code but then have to deal with the resultant errors at runtime
Replying to
But now we have LLMs, and they actually are (mostly) capable of taking unstructured English text and divining, if not the *correct* meaning, then at least a plausible meaning
But therein lies the problem. Traditional parsers are very good at detecting ambiguous or invalid input, but very bad at guessing what might have been meant
LLMs on the other hand are great at guessing what you might have meant, but not at detecting ambiguity, so as a consequence they tend to mask any bugs in the input
Replying to
You could be forgiven for thinking that English is the ultimate high-level language, but in many ways it's more like a *low-level* language because it doesn't impose any structure to guide you, and provides infinite ways to make mistakes or miscommunicate your intentions
*Everything* in English is ambiguous and open to interpretation. Almost no errors or ambiguities are prevented by the "syntax", and so they must be dealt with by the "interpreter" (which traditionally has always been a human)
Replying to
This is generally fine if the programmer is capable of reviewing the (DSL) output of the LLM to verify that its decisions made sense, but this requires that the DSL still exists and that the programmer is capable of reading and understanding it
Which is why I currently don't see a future in which LLMs will eliminate the need for DSLs, or for programmers to need to learn them.
Replying to
@nuthatch@infosec.exchange @carnage4life@mas.to the best case I ever saw of that was once on a project when, having run way over budget, the PM decided to get super petty about scope, and spent multiple hours in calls with the client explaining that we couldn't possibly make the IAP price label dynamic without charging another $20K, only to discover that the cheapo offshore subcontractor we were using had already implemented the feature days earlier at no extra charge
Replying to
The former is true, because no existing DSL has the scope to describe software at such a high level
English can describe systems that span multiple programming languages and platforms, but DSLs are, by their nature, *domain-specific*
LLMs are general-purpose. If I want an instruction interface for an LLM then clearly no existing DSL is going to cut it
(meta-languages like JSON, XML or Markdown don't count - these only formalize the structure, they don't specify how to interpret the content)
Why does the Coding Assistant in Xcode keep asking permission to read or search files inside my project no matter how many times I tell it to always allow?
It means I have to basically sit there babysitting it constantly for every task even when no other human input is required 🙄
Replying to
They really said the quiet part out loud huh
Replying to
DSLs are good for programming because they help you to eliminate ambiguities in two ways:
1) they use a restricted grammar that simply eliminates many possible ambiguities entirely. The higher level the language is, the fewer ambiguities it allows for (at a cost of flexibility)
2) they come with a parser or interpreter that identifies any remaining ambiguities and turns them into errors that are reported back to you to be fixed
Replying to on mastodon.social
Replying to
@matthew Spotify just announced that their developers haven't written a line of code in months. How big of a company are we taking about?
Replying to
This is why, up until the advent of LLMs, natural language programming - or natural language interfaces in general - were a fairly unsuccessful niche, because it was impractical to write an interpreter program that could handle and resolve all the possible ambiguities of arbitrary English input to produce any sort of reliable output without human supervision
Replying to
@isaiah@mastodon.social I sometimes save these things as xattributes in the document itself. If you want them to be use-specific though, you'd probably be better off saving a separate per-document file in ApplicationSupport or something
Replying to
@godofbiscuits non-generative AI doesn't pose any concerns for either energy use or IP theft, but since you seem to want to be angry rather than discussing in good faith I don't see any point in continuing this
Replying to
@kkolakowski@mastodon.social websites can open new windows since the '90s (though browsers tend to block them since they're mostly used for popup ads)
Replying to
@beno I'd be prepared to do that, but I'm not even clear how it would help at this point?
Replying to
@cstross@wandering.shop @christianselig@mastodon.social I mean, you mainly never want to use a linked list because on modern CPUs the cost of out-of-order memory accesses and cache misses typically dwarfs the cost of naive O(n) insert/delete ops on a simple array. But trees are cool too
Replying to on mastodon.social
@dahukanna@mastodon.social I may be missing some context, but natural language seems like it fits with definition 2?
My understanding was that it's derived from expressions like "speak naturally" or "act naturally" - i.e. it refers to the nature of the person speaking, and not the machine doing the processing of that speech
Replying to
@NeoNacho@mastodon.social it had a rushed final season, written by different authors with a noticeably different tone and style than the rest of the show. I think that annoyed many people who otherwise liked the premise of the ending (which I'll admit I didn't)
Replying to
@krzyzanowskim@mastodon.social @helge@mastodon.social @kkolakowski@mastodon.social Slack is an electron app though - so that's one
Replying to
@NeoNacho@mastodon.social is your thesis here that the negative reception to GoT's final season was because they had the temerity to go for an "interesting" conclusion? Because that seems ahistorical
Replying to
@dahukanna@mastodon.social I see your point, but I think "natural" works quite well to convey the idea that you can communicate with a software tool as you normally would with another human, rather than in a special syntax like COBOL or legalese
What term would you suggest as an alternative?