anonyfox

theyre not. this just funnels money into "insurance" constructs that makes everything more expensive for everyone.

At some point its making a butterfly responsible of the butterfly effect. I get the sentimentm but reasonably somewhere the line should be drawn where accidents can happen, especially noncritical stuff like these "hacks" that don't even deserve the word.

artists sell live experiences and original merchandise to actually make money these days. only a rare subset of megastars are living off royalties to begin with, and I am totally fine if some tailor swift has less millions to begin with.

people support/pay real artists if they want more of it, else there is no way I see a right to get paid a wage for just being an artist in itself. if youre not original enough so you get copycats in a day well thats also a signal of artistic value. you cannot separate art from the artist, even if some people are okay with knockoffs after all. there is a difference between art and just content production, the latter is already worthless today with AI speedrunning the decline for mass producing mediocre slop people think is good enough.

Independend of AI I am fully supporting the idea to abolish copyright in general, full stop. Allowing cheap duplication and tweaking is what leads to fast innovation and culture changes. Explicit (costly) patents can remain in place to secure industrial-grade research processes, but the general everyday stuff, incluing music and literature and code and so on is something to strip default rights away to allow faster progress. and oh the poor artists: next to none percentage wise gets much royalties anymore, streaming income is close to zero usually even if popular, the only real world industry to defend are the parasites that do not create art to begin with but siphon production/distribution channels. you might disagree with me, just wanna make clear that some people like me are against this copyright theater before AI blew up in popularity.

unironically sama is probably one of the most honest players here after all, he is burning money and if/when they have achieved AGI they will ask it in how to make money. its a different investor incentive story than trying to monetize into profitability right now that is indeed doomed to fail against china. Having no idea but a vision achieved is rrquired to be met to have those returns is actually the honest part here.

districts are diverse in hamburg, its a huge city-state. I live in the south, in harburg, which has a historically bad reputation but a lot of subdistricts here are in fact pretty nice, and commute to the centre via train is like 20 minutes. some days per years its problematic though, when somthing breaks the train/sbahn lines a lot of people also switch to car and then you just sit in traffic. it just boils down to cross the river which most of the year just works fine with like every 5 minutes somthing drives. harburg itself has been not a district but a distinct city in itself historically, so all "services" upto university) is there too naturally, reduing the need to cross the river to begin with.

Also living in Hamburg right now, AMA if you want to

this in fact is excellent articulated what I also strongly feel.

went from decades of oldschool crafting by hand and got good at it, then went into vibecoding and the experience of decades made me leverage agents significantly better than other peers, but it is also addicting and getting lazy, fast. to the point I type "git push" into claude regularly, not only because I am too lazy to switch my UI tab, but also because occasionally there is this bit of git friction with some local/remote states and claude/codex just resolve it instead of me getting distracted with it.

and now I essentially force my squirrel brain the opposite way again. everything new and exciting and MVP grade ideation happens with vibecoding in hours/days, shippable in isolation somehow so strictly limited blast radius, and then only the <10% of stuff that turn out valuable enough I go back to and rewrite/extract by hand, literally typing it out manually, _without_ autocomplete beyond what a lsp in zed gives me (no cursor magic, ...). AI is only used as a assistant to discuss solution options quickly like a coworker/mentor on the side, but even then I type the result manually.

And it feels like an alien now to do this but its getting better slowly and boy, even my handcrafted code gets better than ever, slow and deliberate, and also I switch to OCaml now for everything that matters, having learned it through LLMs for weeks. No more Node/Go/Rust for me, finally settled after like 20 years. So a quick slop of vibed experiments (react, node, go services, whatever fits best quickly), and the important stuff then slow careful crafted rock solid and lightning fast ocaml.

feels like this is the way. even if it hurts still, kinda like DOMS after going back to gym after long breaks.

just used mistral for a database/scraping creation tool and ended at <10k€ in token costs (via openrouter), beating gpt5.4-mini in output quality and speed and costs after actual testing A/B fairly. so its a super scoped task to be performed hundreds of thousands of time for some automation and mistral just did it better across all dimensions that gpt-5.4-mini. of course thats not a headline in terms of frontier model competitiveness, but for "the boring parts" it just was flat out better than anything else consistently. bonus points it handles mixed-language-content with nuances surprisingly well to turn web content in the wild into structured data really good and fast.

the very point is that theft means you no longer have something since someone else has. copying is you still have it and someone else now too. there is no harm done by copying, except you actually believe that exclusivity as a separate concept is important to you. (debatable, I don't).

Wake me up when it does OCaml fine.

cursor feels so 2025 to me guys. these days zed is just way better for my macbook battery and with acp can talk literally to my installed claude code and codex CLI tools, plus their own and custom providers ontop. I was kind of a decade of a vscode user and always just stayed through the evolutions until cursor, but at some point I just need a lean fast editor+lsp combo, git included and a chat pane next to it that uses my real subs underneath easily. (also: codex-cli can spawn and manage subagents and _resume_ them, acting like a real manager).

could be only me though, but longer interactions over days makes my codex gui app grind to a crawl and cursor was not only expensive with opus via api costs but also heating my room a lot. now I have a dozen zed instances open all crunching along with LLMs barely noticeable on system load (except the occasianal testuite runs but thats expected).

Okay so if this model is half a year behind, so let’s say January opus pre-nerf, this is it.

Inference is actually quite cheap for token costs, the frontier labs burn most of their money on training new models, priced into their token costs ontop of some margins and paying record salaries. So if this goes open, distills are tried out, independent providers around the world host it with actual price competition, the house of cards for anthropic collapses pre-ipo. The floor is opus (open models caught up), the current ceiling is Mythos (self inflicted ban due to the safety bullshit theater), and no way out.

It’s really comical I think it’s even the same guy that warned about gpt2 being too dangerous to release, well that mindset seems to now doing existential harm to anthropic, while the rest of the world essentially laughs and progresses anyway.

wholeheartedly agree. even as a techie with decades of experience, the meta business world is a complete clusterfuck its unbelievable bad. yes they bought and assimilated things like insta or whatsapp, but with their sheer amounts of engineers plus AI coding now even its not acceptable they cannot even get the basic UX of their money printing machine merged into one coherent thing? The god damn creators of react give us us this messed up slow garbage of a maze of config pages weakly connected with complete design breaks and menus I have to navigate with codex/claude to finally find the thing I need? I man accidential complexity can happen, but what are the thousands of engineers and trillions of tokens used for? its just sad

Between professional Elixir, Go Rust and Node over decades now I am arriving actually at OCaml now. Using LLMs to actually teach it to me.

Andd boy, a REAL type system is just something i won't ever again compromise upon. I mean yeah I did many years of Ruby/Rails and loved it back then, and Elixir in that regards at least on surface felt strictly better (sweet pattern matching, pipes, ...) but just SO MUCH CODE is written either at runtime or in loads of tests that essentially make up for the lack of a compiler guarantee about type errors i cannot unsee it anymore. Rust is way better here for example for sure, Trait system and all, but here the compile time tax is very real even after fiddling with optimal crate splits. Plus _sometimes_ a bit of simple mutable code just hits home in a few lines instead of often slower pure FP equivalents.

Happy to see that Elixir finally after years in the making is arriving somewhere, but I essentially left the ecosystem now since I really do either TDD (Type driven Development) now or quick solutions with node/go when quality isn't the concern... and now I discover OCaml (with Effects based multicore now) and yes the syntax is _a bit_ alien but damn it checks all boxes of all techstacks I ever wanted. I can write nearly Elixir style code, pattern match pipes and all, I can write (nobody does but I could) failry powerful OOP stuff, compile instantly, in a statically linked binary, with true parallelism, and a type system that is amazing (don't get me started about module functors). Beam is a impressive feat of engineering, but its also moving like molasses and deployment is nontrivial and quite cumbersome to operate (at least people need quite a lot of learning curves until theyre comfortable with this powerful beast). And then there is OCaml. And the tradeoff here is on the human side, nearly no one knows it, learning curve is high, so statistically no team would pick it in most businesses or has experience with it, and that specific situation is personally for me irrelevant now as a solo builder in an LLM age.

Lets see how good this becomes at some point, I am watching and would have loved to have this at least gradual typing available years ago!