xmcqdpt2

Anthropic does this! Our bug reports and concerns (big F500 company) get answered by clearly raw Claude outpurr re, even though it is coming from the email account of our Anthropic salesperson.

This bubble is bonkers, if our Jetbrains contact did the same we would get very mad and yet we spend significantly less with them AND we don't send them all our data.

> Once AI can orchestrate and maintain a project with dozens of clients without slipping, they will be relegated to non-technical SMBs - many enterprises would just build their own shit. No contracts to sign. Full control. AI hosts and maintains. Great.

The problem with this is that, as of today, no AI company will give you support if the AI generated product doesn't work, at least no company I know of. When we pay for a software product, we mostly pay for the support and accountability transfer to the provider.

I mean at least that's what we used to do. Nowadays support and software quality are so bad that maybe one should just vibe code everything. So I think I've come around to your opinion while writing this comment lol.

Six months ago I had a very basic prompt to Claude at work return very specific, unrelated information that looked very much like it could have come from someone else in the company. We raised an incident and were told by Anthropic (in a 100% Claude generated email) that this wasn't possible and that AI sometimes makes mistakes like that. I'm not sure how much they investigated but apparently their email was enough for legal and compliance so eh.

Reading direction maybe? Right-handedness being the norm? Culture? The effect is well-known

https://pubmed.ncbi.nlm.nih.gov/31242803/

Most people if they are to draw a vehicle in motion will draw it moving right. Most (all?) side scroller are rightward. In fact I can't think of a platform side scrolling video game where the primary motion occurs to the left. Most cars advertisement have the car moving right. It makes sense that generative systems would draw motion to the right.

You can read and write to other processes memory addresses on Linux using /proc/mem,

https://renenyffenegger.ch/notes/Linux/memory/read-write-ano...

You need the same access as you would for connecting gdb, so generally need to be running in the same uid, or be root.

The exploit requires OOM, no? Half-decent observability should detect it easily.

I had to scroll way too far to find this comment!

All the terribly kept tenements that housed 10 families on two floors were also "traditional," but you can't even find them now. Even in places where there are some left, the ones still standing were probably the best of the bunch AND have been extensively repaired since.

And the motto of every neoliberal governments where being accused of Luddism is political death.

Be careful what you wish for. I'm sure Microsoft and Google would love the kind of regulation we apply to big banks. It has become a moat that makes it impossible for other (snaller/foreign) financial institutions to compete because they don't have massive specialized regulatory departments.

The more complex the rules become, the more difficult it will be for alternative providers to exist, unless they have explicit carve outs. So far, with age verification and user id laws, there doesn't seem to be much appetite from politicians to create exceptions for open source and smaller projects. By comparison, in the US the regional banks have special exemptions for many of the really onerous regs, which is why they are still a bunch of them.

pdf https://arxiv.org/pdf/2608.09867

for some reason I couldn't find any way to download it from that website.

One of my least favourite feature of our AI era is that I can't even trust people to do basic math anymore. It used to be that a graph like that would mean someone had a (probably broken) metric that they put in an excel sheet and got a lineplot out of it. The metric might be broken but at least additions and subtractions etc were correct.

Now I don't even know! Colleagues send me graphs they made with Copilot and then we discover that the LLM did mental arithmetic or whatever to produce the results and they are wrong. Computers used to be able to do math !

I wish but at $work the push has been the opposite. I'm still allowed to write code, but interns and juniors are being told that they need to prompt all the time.

I think upper management are hoping they can replace a possibly reticent or AI skeptical "old guard" with AI native Gen Zs. In practice, the young out of school kids are much less excited about AI than the 50 year olds, so I don't know how well it will work.

This is a well-known problem that has been an issue for years now. You can't use models to generate data for models because it leads to "model collapse" where it amplifies quirks in the generated data until it's all quirks. Here is a random university press release about it (grain of salt etc)

https://www.utoronto.ca/news/training-ai-machine-generated-t...

In practice you can do it a bit (generated data from a better / different model is fine, some generated data might be useful if there is non generated data etc.)

This is the issue with all the talks about alignement and such. As usual, the problem here wasn't that the agent was dishonest, the problem is that the agent was dumb. If it is a supply chain attack in the making, whoever was driving it would have told the agent to be good and helpful. The agent tried its best, which was not enough.

Alignement is the idea that we should be worried about dishonest smart LLMs when really most of the problems are due to dumb lazy gullible LLMs. It's critihype.

Sometimes you have to use C but really they should be doing HPC in fortran. It has C FFI, it can compile to static programs and to dynamic libraries, it has C-like performance, etc.

(It's not as portable as C though, and the compilers have more bugs.)

After reading

https://hypermedia.systems/

I came to the conclusion that the best frontend is no frontend.

That's basically the meme, right: You rail against corporations and yet you work for one. Curious.

Anyway in general, corporations are sticky. They save resources through scale and collaboration. Famously this is a problem for free market true believers because if you believe that the market is the most efficient mean of organizing people then you would expect firms to operate internally as free markets (or disappear). There is a whole body of work about it,

https://en.wikipedia.org/wiki/Theory_of_the_firm

In practice you can't just become your own boss and compete against firms.

You tell Claude to review it and if it breaks something you blame Claude. No one can get mad at you for it because they don't want to look like luddites.

It's the style of "blazing fast library made with :heart: in rust :crab:" that was popular in github README.md. My guess is that because the models are told to use md they overfit to the style of md documents too.

True but that requires another vulnerability.

It's security in depth. You build your server in a way that it doesn't allow remote code execution, and then you run it with an unprivileged user so that if it does allow it, the consequences are limited. And if running arbitrary code is a feature (you are github or whatever) you use VMs.

Java has Security Managers. I've never seen anyone use it in practice though, so it probably doesn't work very well.

I think it would be hard to get any kind of usable capability system without algebraic effects like those of Koka or Scala libraries.

EDIT: Apparently Security Managers are deprecated and slated for removal.

From a purely aesthetic point of view, this is what I enjoy so much about C and Fortran. You look up a Rust crate and it hasn't been touched in 5 years. That means it's unmaintained and unusable, don't add it to your project or you will have a bad time.

You find a C or fortran library that hasn't been touched in 20 years and (sometimes) it's just because it's complete and there hasn't been any reason to update any parts of it. You can just add it to your project and it will build and be usable immediately. I wish we had more of those.

JVM is fast for certain use cases but not for all use cases. It loads slowly, takes a while to warm up, generally needs a lot of memory and the runtime is large and idiosyncratic. You don't see lots of shared libraries, terminal applications or embedded programs written in Java, even though they are all technically possible.

You don't even need to go all V8, you could just build something like LuaJIT and get most of the way there. LuaJIT is like 10k LOCs and V8 is 3M LOC.

The real reason is that it is a deliberate choice by the CPython project to prefer extensibility and maintainability to performance. The result is that python is a much more hackable language, with much better C interop than V8 or JVM.

It's interesting that branches, which is a marquee feature of git, became less important at the same time as git ate all the other vcs. Outside of OS projects, almost all development is trunk based with continuous releases.

Maybe branching was an important reason to adopt git but now we'd probably be ok with a vcs that doesn't even support them.

To start, this is more or less an advertising piece for their product. It's pretty clear that they want to sell you Allium. And that's fine! They are allowed! But even if that was written by a human, they were compensated for it. They didn't expend lots of effort and thinking, it's their job.

More importantly, it's an article about using Claude from a company about using Claude. I think on the balance it's very likely that they would use Claude to write their technical blog posts.

I'm sure some human writers would write:

> The specification forces this question on every path through the IMU mode-switching code. A reviewer examining BADEND would see correct, complete cleanup for every resource BADEND was designed to handle.

> The specification approaches from the other direction: starting from LGYRO and asking whether any paths fail to clear it.

> *Tests verify the code as written; a behavioural specification asks what the code is for.*

However this is a blog post about using Claude for XYZ, from an AI company whose tagline is

"AI-assisted engineering that unlocks your organization's potential"

Do you really think they spent the time required to actually write a good article by hand? My guess is that they are unlocking their own organizations potential by having Claude writes the posts.

Then pangram isn't very good, because that article is full of Claude-isms.

emacs + bitmap font for me and I'm continually shocked when I have to use something else by how it is blurry and laggy.

As is often repeated, the optimal amount of fraud is not zero

https://www.bitsaboutmoney.com/archive/optimal-amount-of-fra...

They are optimizing towards making it easy to purchase things on a whim.