Claude Sonnet 4.5 • Hacker News Board

Re: Claude Sonnet 4.5

#581

Post by fullstackchris » Tue, Sep 30, 2025, 1:26 AM UTC

qafy wrote: post #555

hsn915 wrote: post #537

Earlier quoted context omitted.

If anything is depressing, it's the amount of people who think these models actually "can code". Because they can't. Also, Clean Code is a really bad ideology, you should regret wasting time on it regardless of LLM code generation.

Optimization hinders evolution. - Alan Perlis Write that garbage code as long as it works. PMF doesn't give a shit about your code quality.

In general, a good rule of thumb is only code "clean" enough so that you / your team / someone else can figure out what the hell you were doing at that particular area of the source code

Re: Claude Sonnet 4.5

#582

Post by bopbopbop7 » Tue, Sep 30, 2025, 1:29 AM UTC

gabriel-uribe wrote: post #548

hsn915 wrote: post #542

It is time to acknowledge that AI coding does not actually work. ok, you think it's a promising field and you want to explore it, fine. Go for it. Just stop pretending that what these models are currently doing is good enough to replace programmers. I use LLMs a lot, even for explaining documentation. I used to use them for writing _some_ code, but I have never ever gotten a code sample over 10 lines that was not in…

Very interesting observation. I haven’t written a function by hand in 18 months.

Have you built anything in 18 months? I keep asking to see these apps that people supposedly vibe coded in a weekend but when I ask them to share it, nothing.

Re: Claude Sonnet 4.5

#583

Post by martin-t » Tue, Sep 30, 2025, 1:30 AM UTC

int_19h wrote: post #541

martin-t wrote: post #518

Earlier quoted context omitted.

And now wait till you realize it's all built on stolen code written by people like you and me. GOFAI failed because paying intelligent/competent/capable people enough for their time to implement intelligence by writing all the necessary rules and algorithms was uneconomical. GenAI solved it by repurposing already performed work, deriving the rules ("weights") from it automatically, thus massively increasing the value…

It's not that it was uneconomical, it's that 1) we literally don't know all the rules, a lot of it is learned intuition that humans acquire by doing , and 2) as task complexity rises, the number of rules rises faster, so it doesn't scale. The real advantage that genAI brings to the table is that it "learns" in a way that can replicate this intuition and that it keeps scaling so long as you can shovel more compute and…

In a way, yes, you'd be paying the people not just to write down the rules but to discover them first. And there's the accuracy/correctness/interpretability tradeoff.

But also, have there been any attempts on the scale of the Manhattan project attempting to create a GOFAI?

Because one idea I ran into is that we might be able to use genAI to create a GOFAI soon. And it would be as hard as using genAI for any kind of large project. But I also can't convincingly claim that it's somehow provably impossible.

Re: Claude Sonnet 4.5

#584

Post by bopbopbop7 » Tue, Sep 30, 2025, 1:33 AM UTC

qafy wrote: post #549

hsn915 wrote: post #542

It is time to acknowledge that AI coding does not actually work. ok, you think it's a promising field and you want to explore it, fine. Go for it. Just stop pretending that what these models are currently doing is good enough to replace programmers. I use LLMs a lot, even for explaining documentation. I used to use them for writing _some_ code, but I have never ever gotten a code sample over 10 lines that was not in…

sounds like poor prompt engineering. Devin and Claude can both do better work than many interns I have mentored in my career, and faster too. We likely have many many years until it will be even close to replacing an experienced developer but we are already at the point it IS replacing junior engineers. Whether you agree or not, the market has spoken. New grad hiring is WAY down. Fresh CS grads are having an hell of…

Got proof that hiring is down due to AI?

Re: Claude Sonnet 4.5

#585

Post by technocrat8080 » Tue, Sep 30, 2025, 1:34 AM UTC

ChadMoran wrote: post #578

technocrat8080 wrote: post #318

Earlier quoted context omitted.

Curious about this too – does it use the standard context management tools that ship with Claude Code? At 200K context size (or 1M for the beta version), I'm really interested in the techniques used to run it for 30 hours.

Sub-agents. I've had Claude Code run a prompt for hours on end.

What kind of agents do you have setup?

Re: Claude Sonnet 4.5

#586

Post by fullstackchris » Tue, Sep 30, 2025, 1:34 AM UTC

Forgeties79 wrote: post #221

rudedogg wrote: post #70

I just ran this through a simple change I’ve asked Sonnet 4 and Opus 4.1, and it fails too. It’s a simple substitution request where I provide a Lint error that suggests the correct change. All the models fail. I could ask someone with no development experience to do this change and they could. I worry everyone is chasing benchmarks to the detriment of general performance. Or the next token weight for the incorrect c…

> I worry everyone is chasing benchmarks to the detriment of general performance. I’m not sure this is entirely what you’re driving at, but the example I always think of in my head is “I want an AI agent that will scan through my 20 to 30,000 photos, remove all the duplicates, then organize them all in some coherent fashion.” that’s the kind of service I need right now, and it feels like something AI should be able t…

azure vision / "cognitive services" can do this for literally a few bucks

am i even on hacker news? how do people not know there are optimized models for specific use cases? not everything (nor should it) has to run through an LLM

https://azure.microsoft.com/en-us/pricing/details/cognitive-...

Re: Claude Sonnet 4.5

#587

Post by vinhnx » Tue, Sep 30, 2025, 1:36 AM UTC

Claude Sonnet 4.5 has landed support in my CLI coding agent VT Code, combining SOTA language model and agentic semantic code understanding github.com/vinhnx/vtcode

Re: Claude Sonnet 4.5

#588

Post by simonw » Tue, Sep 30, 2025, 1:38 AM UTC

michaelt wrote: post #510

Redster wrote: post #268

Earlier quoted context omitted.

Simonw is a cheerful and straightforward AI journalist who likes to show and not just tell. He has done a good job aggregating and documenting the progress of LLM tools and models. As I understand it, OpenAI and Anthropic have both wisely decided to make sure he has up to date info because they know he'll write about it. Thanks for all your work, Simon! You're my favorite journalist in this space and I really appreci…

> As I understand it, OpenAI and Anthropic have both wisely decided to make sure he has up to date info because they know he'll write about it. And the wisest part is if he writes something they don't like, they can cut off that advanced access. As is the longstanding tradition in games journalism, travel journalism, and suchlike.

If they do that I'll go back to writing about them after they ship. Not a big loss for me at all.

Re: Claude Sonnet 4.5

#589

Post by milkshakes » Tue, Sep 30, 2025, 1:43 AM UTC

inopinatus wrote: post #391

siva7 wrote: post #380

Does 4.5 still answer everything with "You're absolutely right!" or is it now able to communicate like a real programmer?

I won’t be satisfied until I get a Linus Torvalds mode. “Your idea is shit because you are so fucking stupid” “Please stop talking, it hurts my GPUs thinking down to your level” “I may seem evil but at least I’m not incompetent”

ChatGPT Monday has you covered

Re: Claude Sonnet 4.5

#590

Post by FridgeSeal » Tue, Sep 30, 2025, 1:44 AM UTC

jaggederest wrote: post #477

FridgeSeal wrote: post #441

Earlier quoted context omitted.

“Do t worry about the fit and finish in your craftsmanship anymore, just bolt everything together and move on to other woodworking” Is how that argument comes across.

And contrariwise, the argument against tools like these sounds like: "I never use power tools or CNC, I only use hand tools. Even if they would save me an incredible amount of time and let me work on other things, I prefer to do it the slow and painstaking way, even if the results are ultimately almost identical." Sure, you can absolutely true up stock using a jointer plane, but using a power jointer and planer will…

I’d actually argue that we have some absolutely fantastic tools already that are the equivalent of the things like CBC and power tools.

Dev tooling has gotten pretty solid these days, LSP’s and debug protocols, massively improved type-system UX, libs and frameworks with massively improved DX, deployment tools that are basically zero touch, fantastic observability tooling, super powerful IDE’s.

The CNC machine doesn’t wander off and start lathing watermelons when you’re not looking and your planar doesn’t turn into a spaghetti monster and eat your dog if you accidentally plane some wood on the wrong day of the week.