News for hackers. Stuff that matters.
Re:Anthropic hits $2T in Nasdaq debut by akmiller on Thursday November 19, 2026 @02:29PM (#50565014)
> actual hard problems of software development: making sure the software does what it's supposed to, isolating nontrivial bugs, finding ways to make it perform better.

These are exactly the things LLM's outperform all of us on.

Re:Anthropic hits $2T in Nasdaq debut by chrisco255 on Thursday November 19, 2026 @02:36PM (#50565121)
> making sure the software does what it's supposed to

If you've tried to build anything nontrivial with LLM you know this just isn't true.

Re:Anthropic hits $2T in Nasdaq debut by akkad33 on Thursday November 19, 2026 @03:02PM (#50565501)
I don't know most times claude picks up edge cases I never thought of. I spend a non trivial amount of time telling it this edge case doesn't happen
Re:Anthropic hits $2T in Nasdaq debut by chrisco255 on Thursday November 19, 2026 @04:14PM (#50566464)
Well, correcting LLMs from overanalyzing is part of your value in the loop.
Re:Anthropic hits $2T in Nasdaq debut by akmiller on Thursday November 19, 2026 @03:10PM (#50565630)
I've been a developer for 27 years and have built many non-trivial things with LLM's (that's all I do anymore). I'm sorry, but anyone down-voting my comment can't be serious.

Why is much of our legacy software suddenly failing us if we are so much better at finding intricate bugs than the LLM's? Why are all of us easily capable of speeding up our software by orders of magnitude in small amounts of time utilizing them. Yes, they are better at verification as well.

Re:Anthropic hits $2T in Nasdaq debut by platinumrad on Thursday November 19, 2026 @03:38PM (#50566020)
LLMs are better than people at finding intricate bugs, in part because they do not tire or get bored, but people are still much better at designing coherent systems.
Re:Anthropic hits $2T in Nasdaq debut by Archer6621 on Thursday November 19, 2026 @04:32PM (#50566708)
That's because LLMs are incoherent by design, and need extensive guardrails to be useful for tasks that demand coherence.

I share a similar experience with bug finding; I think it also helps that LLMs can simply hold much more context at once than a human, as seeing a much larger portion of the codebase simultaneously makes it much easier to spot discrepancies.

Re:Anthropic hits $2T in Nasdaq debut by lallysingh on Thursday November 19, 2026 @03:39PM (#50566031)
Because we have a new tool that finds certain classes of bugs way better than our old tools. That's it. Sanitizers and linters had similar impact - they just happened early enough that there was less computer-dependent infra that was impacted.
Re:Anthropic hits $2T in Nasdaq debut by reisse on Thursday November 19, 2026 @08:08PM (#50568877)
I've yet to see a sanitizer able to generate a full RCE chain from an out of bounds write it found.
Re:Anthropic hits $2T in Nasdaq debut by CamperBob2 on Thursday November 19, 2026 @09:36PM (#50569482)
A couple of days ago I had DeepSeek 4.1F 1017 do just that, rooting my Synology box when they announced a CVE without any details.
Re:Anthropic hits $2T in Nasdaq debut by chrisco255 on Thursday November 19, 2026 @04:12PM (#50566446)
You bring 27 years of experience and context to every single LLM session you open, so every session you have with an LLM is already going to be far more productive than someone starting from scratch with zero experience in application development. You're able to understand how to map business requirements to technical outcomes and you can competently direct the LLMs to the desired outcome. That is the high-value part of your job that has always existed. It never was about coding the solution in whatever esoteric Turing-complete language you used.

As for legacy software suddenly failing us, what you're really talking about are active systems like Microsoft 365 (it's not legacy code). Those systems have always been complex. That they are "suddenly" failing could be an indication of a number of things, but I'd pin a large part of the responsibility on organizational debt (even beyond technical debt). Maintaining complex software systems is hard, software mostly isn't a static thing it has to constantly evolve. It could also be an indication of over-reliance on LLM competency, for all we know about those system's internals. We just don't know enough to make an assessment.

There have been some zero days discovered by LLMs. There have also been many socially engineered, regular old exploits that have been impressed on us as indications of AI breakouts.

As for speedups, most software is comically unoptimized and has been for some time. We made an explicit trade-off in software development as CPUs and machines got more capable: we started coding in simpler languages with higher levels of abstraction (Python, Ruby, JS, Java, etc) sometimes costing us 10x or more in resource utilization, in exchange for the iteration speed that bought us. Converting those to more efficient languages like Go, Rust, C, etc alone buys you so much. On top of that, there are a number of algorithmic optimizations most developers never implemented that map to well known access patterns in hardware. Lots of low hanging fruit.

Re:Anthropic hits $2T in Nasdaq debut by javier2 on Thursday November 19, 2026 @07:04PM (#50568371)
The thing I still see as the biggest issue with the current LLMs is that they never say "NO". They will try very hard to fulfill your every wish, and nobody stops to ask if they should even be building it. Up until about april this year I was certain AI and LLMs was just gonna be another tool in my toolkit, but now I am not so sure anymore. Relatively non-technical people can get some pretty neat stuff built if they are determined enough. I am not sure what my future or if it will involve coding, but I hope so. The thing about coding, is that many non-technical people still seem to believe it was writing code that was the difficult part. That hasn't been hard for 15-20 years. Modern tooling was already great. The difficult part was reading over enough code to make write the changes in the right places to achieve the requirements you need.
Re:Anthropic hits $2T in Nasdaq debut by bodeadly on Thursday November 19, 2026 @08:03PM (#50568841)
LLMs are good at conventional things and much less so at unconventional things. While this goes far beyond for loops, http calls and concurrency checks, there are nuanced unconventional things that LLMs are not good at. I spend 8-10 hours a day in claude and I am constantly steering it away from turning the codebase into a incomprehensible mess.
Re:Anthropic hits $2T in Nasdaq debut by rvz on Thursday November 19, 2026 @02:56PM (#50565394)
If you are building software that has no risk, then it's not important enough for anyone willing to pay for it and they will believe it is safe enough to build it themselves.

Examples are: Note taking apps, voice recording apps, calendar apps, video games, etc.

Re:Anthropic hits $2T in Nasdaq debut by lallysingh on Thursday November 19, 2026 @03:37PM (#50566009)
My experience is exactly the opposite. 30 years writing software for my job, been using claude/grok for most of this year. At first they were really amazing. A real lever for higher output. But, I spend most of my time now telling them how to design high-performance systems and how to manage high concurrency and low contention.

They'll shove out a pattern quickly, but no matter how new and magical the latest model is, they will just put out some tweaked variant of existing designs. I find myself treating it like a junior-mid engineer. "Add rdtsc-based timers around major parts of the code to whittle down what's taking too long." "Use that to figure out where the transfer amount is way too small to justify the time taken without copying the payload." etc.

But concurrency bugs from large LLM-developed codebases are a major issue. They constantly find bugs they wrote. I eventually told it to start writing formal models of the concurrent uses and to (a) validate the models; (b) verify the code implements the models; to make it work.

So yeah, I stand by my words. For simple commands and tiny codebases, it /can/ do those things, but even small projects (50 KLOC) are a mess of making it fix its own mistakes, reconcile basic logic in requirements analysis, and everything has to be taken over with a fine-tooth comb to make it any good.

Re:Anthropic hits $2T in Nasdaq debut by CamperBob2 on Thursday November 19, 2026 @09:31PM (#50569445)
My experience is exactly the opposite. 30 years writing software for my job, been using claude/grok for most of this year.

If you don't specify the model you are talking about, you have written a sentence with near-zero information content. If that's your habit, well... that's likely part of the problem.

I promise you, Fable and Neon are better than you are at pretty much whatever. Sonnet 4.X and MechaHitler, well, maybe not.

Re:Anthropic hits $2T in Nasdaq debut by Tanjreeve on Friday November 20, 2026 @01:50AM (#50571034)
I just got sent around In a loop by fable for several hours on solving some more complicated rust lifetimes. It's not magic. It's a more powerful reasoning model. But to use them in an automated process it's still an exercise for the user. Again it's fine.

In a similar vein public clouds contain the components to built any software you can think of. But that doesn't mean that all the solutions possible spring into existence just because AWS exists.

Re:Anthropic hits $2T in Nasdaq debut by lallysingh on Friday November 20, 2026 @10:25AM (#50576041)
Opus 4.6/grok 4.7 at work, Fable at home for personal (50 KLOC) projects. I'm still waiting to be impressed.
Re:Anthropic hits $2T in Nasdaq debut by saguntum on Friday November 20, 2026 @12:20AM (#50570473)
The async code issue resonates with my experience. I spent the better part of a day trying to get one of the Claude models (high 4 version or Opus 5, I don't recall) to try to implement a bidirectional gRPC stream of some complexity correctly in Go. I ended up just writing it myself. I am not anti-AI by any means, but I was surprised it struggled with something there are ample examples and documentation for online.

The tools are amazing, but I think they are most amazing when you can have deterministic verification loops. Async bugs aren't always encountered, so it's a little more difficult for the LLM to prove its work and iterate. At least, that's my current hypothesis for why it tends to degrade so much for this type of problem.

Re:Anthropic hits $2T in Nasdaq debut by LtWorf on Thursday November 19, 2026 @03:55PM (#50566240)
> These are exactly the things LLM's outperform all of us on.

If you throw an excessive amount of money at the problem yes.