> One does often hear that where LLMs shine is with greenfield code generation but they all start to struggle working with pre-existing code. It could be that this wasn't a like for like comparison.
In my experience working in a large codebase with a good set of standards that's not the case, I can supply examples already existing in the codebase for Claude to use as a guidance and it generates quite decent code.
I think it's because there's already a lot of decent code for it to slurp and derive from, good quality tests at the functional level (so regressions are caught quickly).
I do understand though that on codebases with a hodge podge of styles, varying quality of tests, etc. it probably doesn't work as well as in my experience but I'm quite impressed about how I can do the thinking, add relevant sections of the code to the context (including protocols, APIs, etc.), describe what I need to be done, and get a plan back that most times is correct or very close to correct, which I can then iterate over to fix gaps/mistakes it made, and get it implemented.
Of course, there are still tasks it fails and I don't like doing multiple iterations to correct course, for those I do them manually with the odd usage here and there to refactor bits and pieces.
Overall I believe if your codebase was already healthy you can have LLMs work quite well with pre-existing code.
Realistically many can't. It's a nice illusion of choice to be maintained though, gives the system more power for self-exploitation if you are a true believer.
300 million users in 2007 is mighty impressive, the internet was not absolutely ubiquitous like now, mobile access to it was in its infancy. Relatively speaking it is as impressive as 1 billion users in 2026.
> 1. The tax could cause him to sell equity he doesn't want to. It's not like he has $270B in a checking account.
If there is $270B in equity invested, making those 5% back should be rather straightforward for someone with that much wealth, a decent wealth manager would recoup that easily. Money makes money.
> 2. If they do it once then why not again next year? Maybe next time it's for only $100 million or $10 million or $1 million. Eventually everyone is paying 5% of their wealth every year. Why not? That's how we got the current income tax.
This is just a slippery slope fallacy.
> 3. It's the principle. Resisting these efforts sends a signal that they aren't a good idea.
Exactly, it's blackmailing, sending a message "look what you've made me do" when the government attempts to reign in the ultrawealthy.
> 4. Do we really think the money is better off in control of the incompetent CA government than invested in private enterprise or donated to charity? I don't see how it's better for it to line Newsom's Swiss bank account.
With this argument you can defend never paying taxes to CA then, do you think it would be better as a complete anarco-capitalist state? It makes me sad that USA's public governance is so bad that this argument is always used to defend rich people not paying taxes; the political system is so absolutely broken that people prefer to allow ultrawealthy folks to keep hoarding even in the face of very real issues fracturing society stemming from that instead of thinking about how that money could fix many public issues.
> If there is $270B in equity invested, making those 5% back should be rather straightforward for someone with that much wealth, a decent wealth manager would recoup that easily. Money makes money.
Its his voter shares, he wont get them back. Larry and Sergei currently control 51% of Google votes, if they sell any more they lose control of Google so they can't afford to sell.
> Exactly, it's blackmailing, sending a message "look what you've made me do" when the government attempts to reign in the ultrawealthy.
If it’s blackmailing, it’s blackmailing that is legal and extremely common. Similar events have occurred in multiple jurisdictions around the world whenever wealth taxes are attempted. The only way to prevent this in a country like the US that doesn’t allow retroactive taxation is to institute an exit tax, but this is also not legal at the state level in the US. Oh, and you also likely need capital controls, another no-no inside the US.
Capital flight is so historically common that there’s a common phrase to describe it. I have no idea why CA thought they were any different.
If the EU announced that non-EU entities aren't subject to GDPR, I think that would substantially defuse and perhaps entirely eliminate the conflict. Their current guidance is precisely the opposite (https://gdpr.eu/what-is-gdpr/): "the GDPR applies to you even if you’re not in the EU". They even have a details page to make sure it's 100% clear (https://gdpr.eu/companies-outside-of-europe/): if you're a Colorado company with more than 250 employees, selling mainly to other Colorado businesses, the GDPR applies to you in full and the EU claims the authority to levy fines against you for violations.
I don't understand your response. As I said, the EU's position is that it doesn't matter whether you "just leave", because the GDPR still applies to companies who are not located in the EU and do not do business in the EU.
I get why people find this hard to believe, because it is kind of a crazy rule, but I repeat once again that this does not matter. Even if you have never sold a single product to an EU resident, and never plan to do so, the EU says as my original comment detailed that you are subject to the GDPR the instant an EU resident provides you with personal data.
(And of course, it's also the case that "selling to an EU resident" is substantially broader than "doing business in the EU" - EU residents do often travel to foreign countries and provide personal data to stores they transact with while there.)
American laws also have universal jurisdiction (for example, the Bill of Rights doesn't say, "unless you are located outside the US"). Most countries do not explicitly recognize that their laws do not have universal jurisdiction.
In practice, it is easy to pick out the situations in which there is "practical" universal jurisdiction, vs "theoretical" universal jurisdiction.
A Colorado company selling locally in Colorado falls in the "theoretical" bucket.
> (for example, the Bill of Rights doesn't say, "unless you are located outside the US").
The Bill of Rights is a set of constraints on the US government, so even to the extent it applies to the government when acting outside of its borders [0], it isn’t an imposition of US law on the territory of other countries, but a limit on such imposition.
[0] And it doesn't fully, see, e.g., Johnson v. Eisentrager, 339 U.S. 763 (1950), subsequently limited somewhat with the core holding retained in Boumediene v. Bush, 553 U.S. 723 (2008).
1. GDPR applies to EU residents in the EU. The protection does not apply to EU residents going on trips to the US.
2. Based on the examples they've presented, there is a SUPER clean solution to your concerns. Geo-blocking. Problem solved, bye bye GDPR. But don't go crying for EU citizen money, can't have it both ways.
Just read the examples they present, they're fairly well written.
When you talk about liability, where is that relative to? Liability means that a court will order you to pay money. Which court are you worried about? US courts won't order you to pay anything, and European courts can't take away your money if your money's in the US. Businesses break laws in other countries all the time, and nothing happens to them. Remember when Russia charged Google with a quadrillion dollar fine and nothing happened?
> But the GDPR does not apply to occasional instances. Rather, regulators look for other clues to determine whether the organization set out to offer goods and services to people in the EU. To do so, they’ll look for things like whether, for example, a Canadian company created ads in German or included pricing in euros on its website. In other words, if your company is not in the EU but you cater to EU customers, then you should strive to be GDPR compliant.
2. As a general rule Europeans are MUCH less lawsuit addicted than Americans. Plus the way the GDPR works is that generally complaints are filed with a government agency that investigates.
Not very dissimilar to Mussolini's tactics back in the 1920s/1930s, it's actually quite impressive how many similarities there are between Trump's and Mussolini's ways to find political power.
They will open up to be undercut by players like Gemini that can provide a less-shitty free tier, and capture their market share.
I seriously doubt, at this moment, that OpenAI can come up with a offering good enough to entice people to pay for them when there will be other free to use services around. Google seems to be well positioned to eat their lunch.
Google does indeed. But that ignores that Microsoft is playing the same game.
For example in the UK's NHS, the worlds sixth largest employer is now fully committed to Microsoft 365.
That's a lot of Copilot money if Microsoft sees it that way.
And OpenAI is funded via Microsoft, I also have a Microsoft 2TB subscription. And so do many people have both work based and personal home subscriptions.
It's a complete mess of a situation. If Microsoft moves away from GPT (it can since it's advertised under the copilot brand) OpenAI is dead in the water of course.
In my experience working in a large codebase with a good set of standards that's not the case, I can supply examples already existing in the codebase for Claude to use as a guidance and it generates quite decent code.
I think it's because there's already a lot of decent code for it to slurp and derive from, good quality tests at the functional level (so regressions are caught quickly).
I do understand though that on codebases with a hodge podge of styles, varying quality of tests, etc. it probably doesn't work as well as in my experience but I'm quite impressed about how I can do the thinking, add relevant sections of the code to the context (including protocols, APIs, etc.), describe what I need to be done, and get a plan back that most times is correct or very close to correct, which I can then iterate over to fix gaps/mistakes it made, and get it implemented.
Of course, there are still tasks it fails and I don't like doing multiple iterations to correct course, for those I do them manually with the odd usage here and there to refactor bits and pieces.
Overall I believe if your codebase was already healthy you can have LLMs work quite well with pre-existing code.
reply