usernamed7

this is "too little, too late" for me. I'll stick with opus 4.8.

I was surprised to see .cern as a TLD - i would not expect it to need a family of websites in the way .gov or .mil do.

surprised at the responses to this post. While I thought the title was dumb, the underlying thesis is not the ragebait I was expecting, and I actually agree with the author.

LLM's work best when they can call a tool and observe the success/failure of a change. If you're HITL then you're the tool, but the result is the same. only slower.

I'm working on a 2D game (pixi.js) with claudecode, and after I moved some logic into a webworker the LLM created a headless simulation exercise of it and would run this to test performance changes against (or in exploration of an issue), which I was surprised by.

I also created some robust graphs & metrics which were easy to screenshot and upload to claude. this was a HITL but it gave claude a lot more insight into what's actually happening instead of guessing when the browser plays the game and has FPS drop.

LLM's do best when they can see what their code is doing. If you can't remove yourself from that cycle of testing you should at least optimize it so you can give rich errors.

it started out alright, but then it just became obvious.

Wall Street calls it securitization. The merchants of Assur called it Tuesday.

was the give away for me.

I haven't told you to do anything, only highlighted that you can choose how to live your life, including not using LLM's.

Believe it or not, some people actually do derive a great deal of value from LLM's and it's also ok if you don't or can't.

That's a choice you are free to make, just like you're free to shrug and not use the internet or computers.

Let us hope this only accelerates the proliferation of local models

I am telling you the count because the count is the point.

The honest caveat, because it matters:

This one I got right in the original, and now I have the number to back it.

Thanks Claude.

You ask for a date picker

    <input type="date">

Slashing prices is only going to go so far...you couldn't pay me to use chatGPT or codex. I used chatGPT for a long time but once i switched to anthropic i could sense a higher quality and a lot less frustration and correction on my part.

I agree there is a "sameness" that you get from tailwind.

Tailwind has a unique benefit in that you can change the CSS of the page and just that page. There is no chance that you make a change that breaks the rest of the site because you wrote a rule wrong. In some environments/applications this is a big deal.

But honestly that's the only credit i'll give it. The class names are still confusing to me and you do get more flexibility with CSS. And i'd rather be writing classless css and targeting custom HTML elements anyway.

this reads to me as a false equivalence. Choosing a Mac is opting into a specific alternative. Choosing Windows is just taking the default.

These are long synthwave mixes that were uploaded before Suno became popular (it was released December 2023)

I was thinking the same - makes the article feel very amateur and unprofessional. And I know for a fact that AI can do a better job at writing than this, I doubt they read it and had any sense of how poor the writing actually is.

I've been reflecting on this lately. I bought a window air conditioner which has higher BTU but also has an "app" you can download on your phone. Which I have no interest in. The remote has basic controls and that's...fine.

My old air conditioner would remember its last mode and temperature settings but this one has amnesia and starts up in the same state blank state every time. It also automatically raises the room temperature over night which is dumb.

I haven't bothered downloading the app, because I don't need my air conditioner on the wifi and i can't fathom what would be so great about an app when i basically just need on/off functionality.

I see the same thing happening with cars. I heard that dave (of dave's garage on YT) has a car from 2018 that can no longer start remotely because the car company shut down the servers. It's not even 10 years old!

And good luck finding a TV that isn't "smart" that shows you their own ads, sends screenshots of what you watch, and pauses ads when you press mute. Ridiculous.

I confirmed searching for "Iran" with a before set ends up showing me all recent results. BUT if i use this search form to generate the query (and not the youtube search bar) it works.

via the search form: https://www.youtube.com/results?search_query=+%22iran%22+bef...

via the YT search: https://www.youtube.com/results?search_query=+%22iran%22+bef...

as expected, this is slop

I won't be cancelling by subscription because I don't have one :) I have found I get way more value out of subscribing for 1-2 months to a service and then switching to another or going a few months without any.

At this point, netflix will keep raising prices. Because they can, and because they have to as a public company beholden to shareholders. I'm not sure there are other markets or other products/services they could expand into, i think they've already reached a point of saturation.

This argument doesn't make sense to me. did airplanes kill road trips? AI lets you go faster, but you don't have to use it at all, or can use it in select ways to collab with. Unless you're somehow bothered by how other people code, you've only got more options now.

this is pretty cool! what I do for my project (3d voxel web page platform) is treat the scene as a document and push updates to the server when things change. this works well for distribution and simplicity, but does mean multi-player is "difficult". using spacetime to manage and track that info and have threejs render would be cool and enable a lot of things.

I may explore spacetimedb as a project sometime, although i am reluctant to be tied into a platform that my project cannot work without, that i cannot host myself, that can change their pricing/limits/terms of at any time and for which there is no alternative.

aaaaaaand this is why i prefer anthropic. There is just too much sneaky/misleading/deceptive things with chatGPT. Even if benchmarks show codex to be slightly better, the developer experience with claude code is much better.

glue work is real work and a lot of projects get stalled or blocked because there was no glue; especially in SOA where you have different teams with differing roadmaps integrating with each other. It's not just about communication/socialization, but also how code interacts and how the contract is defined.

I wish youtube would also take action. So much of my recommendations for music is just AI slop now. I'd be OK if they flagged videos as AI and let me block them. At first I was open to the idea of AI synthwave as a thing but now it's just gotten out of hand and every day i am flagging new channels to not be recommended.

using tailwind docs is awful. I'd MUCH rather use an LLM than try to grok their documentation. That it was their only way to promote commercial offerings is not my problem, there are many other ways to approach this than encouraging a worse experience for devs.

while AI does lower the barrier to who can do software development it does not nullify their need only moves them into more complicated domains. Yes, if you're job as a SWE was building landing pages, you're pretty much cooked. But if you're working in complicated domains, or domains that require a level of technical awareness or social skills to create success, AI is just an amplifier and makes the boring/frustrating parts easier.

I am using claude to build a pretty complicated project. Technically, a lot of what i am prompting are things that other people could prompt. But I also do find myself leveraging a lot of knowledge in shaping what the code should do and how it should do it, and also needing to step in when claude reaches limits of it's training. I am confident that the number of people who could build what I am building is pretty small.

So I think the author is creating a narrative that is unfounded. There will always be software engineers. There will always be engineering challenges that it takes a human to resolve. Yes, always; no matter how "smart" the AI gets. For sure, AI will be taking some development jobs. But calling for a collapse is simply hyperbole, shortsighted and naive.

hey this is pretty cool! very interesting concept to transpile to rbs.

in most of my coding career (rails) i have not needed types. But when i was working on a large codebase with many teams, the lack of types was a recurring issue. now i work in a SOA and face different coupling challenges.

But i much prefer this syntax to what RBS required.

"Your scientists were so preoccupied with whether or not they could, they didn't stop to think if they should"

-Dr. Ian Malcolm

We're developing a responsible access framework that makes models available to researchers for scholarly purposes while preventing misuse.

oh COME ON... "AI safety" is getting out of hand.

the AI hallucinated and somehow it's rails fault?

GID's are great - i think the issue is with how they leveraged rubyLLM for something they should inherently not be using LLMs for.

Remember that GIDs were made for facilitating ActiveJob serialization - they are a system-level facility, not a product-level facility.

I think this is somewhat obvious given the signature like gid://awesome-app/Post/32; there is no scoping to the user or account so it should be treated like a global lookup. If you need scoping to a user/account you can build that.

Honestly I think this is a matter of the author using poor design decisions and over leveraging LLMs. But this is not the fault of Rails, it is working as expected.

Be careful with LLMs!

They just described servant leadership and called it something else. It's not about parenting or treating people like children.