viraptor

It's an interesting review, but it seems to mix some things together. Gruber's claim is that we shouldn't sacrifice quality, but none of those sources show that's what's happening. Having less diversity in answers could mean you're more consistently getting a better/worse/same quality answer. There are lots of people complaining about not being able to get a real temperature=0 answer - they'd actually welcome the consistency too.

The practical example of not liking the sentence is not great in my opinion. Instead of just re-rolling the answer, why not continue in that thread and say what you don't like? Or request 20 alternatives, say which go in the right direction and request 20 more? Retrying from scratch with one sample is a waste of time and doesn't indicate anything about quality.

Basically, it's all theoretical - show us the actual benchmarks/samples where the quality is consistently degraded.

Yeah, it shouldn't. But at that point it's not "Why didn't Claude detect the hack" but rather "Why isn't Claude running all processes at your company autonomously", which is a bit silly. We know it isn't and it's not at that level, because anthropic employs engineers.

Some laws are there not to guarantee compliance, but to enforce rules, fines, and punishment. For example the laws forcing you to declare bribes in various situations are not there to make you actually declare them.

I'm not sure where you see the flex. If someone gets hold of the token allowing the upload of your app to the store, it's completely irrelevant how good your internal code/review systems are. They've been completely bypassed and never triggered.

This is a cheap troll, but just in case someone is wondering: This extension likely never hit the standard workflow and/or repository. Most likely someone got the credentials and uploaded a locally modified version. No review system at anthropic has seen this version.

> modern Qwen and DeepSeek

It's worth checking different behaviours across the available models from time to time. What you end up preferring may be very dependent on what you do. For example while being diligent is nice in a lot of situations, sometimes I ask a personal agent a question and mimo spends 10min validating the answer in a few different ways. Sometimes I want that big research, sometimes I just want to know what that transformers toy is called.

There's also a huge difference is writing style. I'll take GPTs version of a commit summary over Claude and its copies any time.

Serious answer:

- they provide a service, not software

- you'd also need a maintained /24 of IPs with good email reputation

- unless you want to only use an email client, you need to do lots of testing on the web editor part for how the formatting gets interpreted in other clients, how to render weird emails correctly, how to protect the privacy in external resources, and lots of other things

Even with an LLM that's both a multi-week implementation and an infra management project for life. Their users in general can't do that. (There's a handful that could)

Can you link to a specific ruling that supports this?

> Batteries are traditionally grid-following, so they'll just follow whatever frequency the grid is already running on.

Is there in theory an amount of batteries with inverters where enough energy is created to "not notice" the spinning generators going away? I.e. everyone is trying to follow the frequency that's already flowing therefore everyone sees some frequency flowing? Or would it be like talking with echo/delay where they start slowing down by always being behind until everyone stops?

So it was a Hungarian $ - got to know :)

There's likely some spare solar capacity that will help. Currently the peak of solar is almost counterproductive, but converting that to storage is good for both the consumers and the power plants. (Less feed in during the day, lower evening peak)

Technical detail question: if all of those batteries were on connecting allowing feeding back to the grid, would they be able to synchronise the frequency/phase? What's the threshold/requirement for that to happen?

It doesn't have to be specific attributes. It's just map(string, derivation) for the package list, however you language expresses that.

Similar to how JS has x["a"]==x.a, they're attributes.

Some prices. It's not like billionaire spending will influence the local carrot prices. They're unlikely to overlap at all with most day-to-day spending for normal people.