jjani

Something not copy-pasted from an LLM would be more acceptable.

Hey CodeRabbit employees

> The researchers identified that Rubocop, one of our tools, was running outside our secure sandbox environment — a configuration that deviated from our standard security protocols.

This is still ultra-LLM-speak (and no, not just because of the em-dash).

Would love a source on them not paying. They've appealed the latest one rather than refusing to pay [0].

The 500 million one is also for anti-trust rather than GDPR, which is the one that includes % global revenue fines.

[0] https://www.reuters.com/sustainability/boards-policy-regulat...

I think image/video generation models tend to get released later? And some subscription products?

That would really require an absolutely dramatic escalation of the fines, as the current ones - even the Meta €1.2 billion fine they got in 2023 - are absolute drops in the ocean compared to even just their yearly EU profit.

And the reality is that the US government would start blackmailing the EU long before that dramatic escalation is reached.

Because Google never made such profits in China even when they did operate there, neither do they really have the opportunity to do so, even if they'd comply with everything they'd be asked of.

Entirely different from their EU operations to just give one example.

"letter" here wasn't intended to mean "letter of the law", rather "letter of whatever we tell you to do".

For GDPR they already are, it should indeed be made to be the same for anti-competitiveness laws.

https://gdpr.eu/fines/

> The less severe infringements could result in a fine of up to €10 million, or 2% of the firm’s worldwide annual revenue from the preceding financial year, whichever amount is higher.

> These types of infringements could result in a fine of up to €20 million, or 4% of the firm’s worldwide annual revenue from the preceding financial year, whichever amount is higher.

And then there's places like China, where the effective fines are "you either comply to the letter or you won't get to operate in this country".

No, that's not the real answer at all, it's anything but.

You have no idea just how much revenue Google et. al make from e.g. the EU. The shareholders would absolutely eat Google alive for just walking away from many billions of dollars rather than just complying. I've said this here before:

> A point we're still lightyears away from. The lengths they go to in order to operate in China are magnitudes greater than to operate in the EU, yet EU makes them $10+ billion more profit than China.

> What would actually happen is that the US would start seriously threatening (blackmailing) the EU to a degree where it's forced to relent long before Apple would pull out.

> Apple's estimated operating profit from the EU is around $40 billion dollars. If the US government wouldn't get involved, they could force Tim Apple himself to live on top of the Alps and he'd happily do it rather than lose that $40 billion, or shareholders would vote him out ASAP.

You can substitute Apple for Google or any SV big tech.

>We got a little peek into this when the GDPR was rolled out and many small and medium companies simply blocked GDPR countries rather than risk the massive fines spelled out in the GDPR.

So you do "% of global revenue", "gatekeeper/minimum size applicability" and so on. Absolutely trivial stuff, this has been figured out ages ago.

> go to any non-western country. height still matters but personality, integrity and intelligence is enough to get you ahead in the dating market.

Any non-western country? Bold statement. I can tell you that there are non-western countries where height matters more than any Western country.

They're swapping out hardware, which is why they're asking money for this to compensate the labor costs. Not saying this justifies it, but the title is misleading.

I really don't get why you'd use an ethnic slur as your HN username. Does it make your secret place all tingly when you imagine those who understand, spot it?

> This is actually pretty smart on their part.

It's the literal concept of a tax haven.

Late reply, but this isn't really so much about being "lazy", more about unrealistic standards. It's like if all IT fresh grads in Sweden (to name the same random country again) would only want to work at Spotify or DICE or bust. And they'd rather live with their parents for two years after uni in the hope they might get a job at either of those 2 then go work at a smaller company outside of Stockholm that would instantly hire them.

This permeates through in every facet of society. For example, the exact same psychology is the reason behind having the lowest birthrate in the world. Don't have the money to raise a child in a nice new apartment in a great neighborhood including tens of thousands dollars spent on private education? Ok, then no child it is.

I've been to Starbucks in Seoul 100+ times (not proud of it, more the opposite) and have never seen anything like this.

> the current level of exploiting each other doesn't come naturally for people.

Agreed, it doesn't. Nothing about that fact seems reason for optimism, however. It's a gigantic leap to go from "the current level of exploiting each other doesn't come naturally for people" to "that level is going to decrease, or stop rising, any time soon".

"Wealthy startup steals EU trademark" still works as most will fill in the blank with US.

For life? How can you be bound by this? Unless you sold yourself out for an extra month pay.

That can all be trivially fixed by style of seating and tables, removing all power outlets and so on. People who don't go there to work won't care.

I live in the country with probably the highest sit-down cafe density cities in the world (Korea), and this issue has been figured out ages ago. If you know any such cafe owners who don't understand how to deal with this, I'm happy to have a chat with them. Or they can come over here and I can show them a dozen cafes so they can see it with their own eyes.

You simply set up the cafe to accommodate the exact % of such laptop users as you're comfortable with, which can be 0%, 100%, or anywhere in between. If you do for some reason want to run a cafe where 100% of seating is usable for laptop workers, then the way to keep it all profitable is also straightforward: 1. you make your cheapest coffee (converted to Dutch CoL) 7+ euros a cup (use some single origin stuff that's still cheap when bought from wholesale). 2. As food, only offer small sweet bites and make those similarly overpriced. 3. Make the seating dense so you can fit a lot of these office workers. Bar seating is especially space-efficient for this.

The Netherlands even has an advantage; people can't just leave their setup on the table and leave for hours as it may well get stolen - this is not an issue in Korea so some people actually do this, the worst case scenario for cafe owners.

Love how the MCPs mimic the company they represent in that sense.

> If you remove the people born into privilege, from attending your college, all you succeed is in making your college irrelevant, not those people irrelevant.

Calling 99% of the world's universities irrelevant is certainly a take. And not a very intelligent one.

This was nothing but "temporary", it's still in place; the last time we ran the evals is 2 weeks ago and it's the exact same. It can't be a "capacity glitch" either, as it actually outputs those as proper tokens.

It's possible that it was an internal system prompt change despite the claims of "there is no system prompt on the API", but this is in effect the same as changing the model.

> There IS evidence that would satisfy me, but I'd need to see it.

Describe what this evidence would look like. It sure feels like an appeal to authority - if I'd be someone with a "name" I'm sure you'd believe it.

If you'd had had the same set of evals set up since then, you wouldn't have questioned this at all. You don't.

> I don't think you're making it up, but without a lot more details I can't be convinced that your methodology was robust enough to prove what you say it shows.

Go and poke holes at it then, go on. I've clearly explained the methodology.

That's a very roundabout way to phrase "you're completely making all of this up", which is quite disappointing tbh. Are you familiar with evals? As in automated testing using multiple runs? It's simple regression testing, just like for deterministic code. Doing multiple runs smooths out any stochastic differences, and the change I explained isn't explainable by stochasticity regardless.

There is no evidence that would satisfy you then, as it would be exactly what I showed. You'd need a time machine.

https://www.reddit.com/r/ClaudeAI/comments/1gxa76p/claude_ap...

Here's just one thread.

It doesn't seem clear to me what the result would be, what makes you think it is?

It's super unlikely that there's one answer across all use cases for "GPT5 vs Claude" anyway. Double so if you include Opus and the vs is pure quality instead of quality per $.

If they're meaningless why'd you post it besides getting views on your blog?

> Also regarding "Sonnet is faster" I did explicitly mention that I believe this is because GPT-5 is in preview and hours from the release.

I genuinely don't see this mentioned, where is it?

It appears to be overtuned on extremy strict instruction following, interpreting things in a very unhuman way, which may be a benefit to agentic tasks at the costs of everything else.

My limited API testing with gpt-5 also showed this. As an example, the instruction "don't use academic language" caused it to basically omit half of what it output without that instruction. The other frontier models, and even open source Chinese ones like Kimi and Deepseek, understand perfectly fine what we mean by it.

No, they do change deployed models.

How can I be so sure? Evals. There was a point where Sonnet 3.5 v2 happily output 40k+ tokens in one message if asked. And one day it started with 99% consistency, outputting "Would you like me to continue?" after a lot fewer tokens than that. We'd been running the same set of evals and so could definitively confirm this change. Googling will also reveal many reports of this.

Whatever they did, in practice they lied: API behavior of a deployed model changed.

Another one: Differing performance - not latency but output on the same prompt, over 100+ runs, statistically significant enough to be impossible by random chance - between AWS Bedrock hosted Sonnet and direct Anthropic API Sonnet, same model version.

Don't take at face value what model providers claim.

They're almost never straight upgrades for the exact same prompts across the board at the same latency and price. The last time that happened was already a year ago, with 3.5 Sonnet.

My chats so far have been similar to yours, across the board worse than o3, never better. I've had cases where it completely misinterpreted what I was asking for, a very strange experience which I'd never had with the other frontier models (o3, Sonnet, Gemini Pro). Those would of course get things wrong, make mistakes, but never completely misunderstand what I'm asking. I tried the same prompt on Sonnet and Gemini and both understood correctly.

It was related to software architecture, so supposedly something it should be good at. But for some reason it interpreted me as asking from an end-user perspective instead of a developer of the service, even though it was plenty clear to any human - and other models - that I meant the latter.