Lalabadie

It's been good for one-off cases in my limited experience. I would describe its behaviour as Sonnet-shaped, if that makes sense to you. Good answers but it often decides to reason a lot about simple things before getting to an output.

At the speed Flash has on most providers, it doesn't really turn into a latency concern.

Anecdotal data from my own tests: allow only one provider if you want good cache usage on OR. 80 cents is probably 4x the price you should have paid.

Providers' cache hit stats are available to consult, and only 1-2 of them behave properly if I remember correctly, zero if you request providers that don't store and train on sessions.

Opencode also have a ZDR (zero data retention) deal with them – if I recall correctly, that's not something you can enable as an individual DeepSeek subscriber.

"I completed the Synacor challenge" is still a good way to get experienced CTOs' attention.

Lutke's own post was about pension = no votes, which I identify as the thing he decided to actually spent time on. His other tweet is a Musk-like "Good idea" reply to someone else's proposal regarding tax brackets, no?

I don't think memory fails me on the main points. If you think ranking the same read facts differently from you means faulty reading, I don't really owe you the labor of explaining where that fails.

That would be true if he advocated for votes according to effective taxation rate or something similar. He caps his example at 800k reported income, if I remember, which is a rounding error to his income class.

Instead, he makes sure to include the idea of stripping low-income people of all voting rights.

I really appreciate that KOReader is an option. On my older Kindle (first-gen Paperwhite), the default reading experience had declined until page turns take several seconds. Firmware update and factory reset wouldn't fix it. Using the default reader just wasn't a viable scenario.

I jailbroke it, launched KOReader, and it's like I have a new e-reader. The amount of customizability gets kind of in the way, but once it's set up I don't really think about it from one month to the next. I also didn't feel the need to upgrade my original install either.

I don't think you need to go to China to find that people think the average north American eats terribly for their health.

Yes, however if a tool like this commoditizes a way to stay "higher" in the cache hierarchy to avoid getting dropped, then for the same total capacity, Anthropic have to become stricter about retention.

"You are trying to kidnap what I have rightfully stolen!"

But then that becomes an article about the performance of your draft router. This one is about the fact that there's this level of optimization potential.

Yes, hah. Historically, the US got its moment to define for everyone what the rules of capitalism and globalization are, and China then played to win.

It's also a capture made from seeded test data, all messages are at 5:42. Or a mockup based on said capture.

I've said a similar sentence on here in the past, but there's a large anti-tax sentiment in the US because their government is one of the most inefficient per dollar.

People are against paying for nothing, not against paying for services. There's just no confidence that the amount of services rendered can change away from "as little as they can get away with", so there's no enthusiasm for paying at all.

If you want to direct blame, depression robbed all three of them of their future together, and their insurance allowed depression to progress until fatal.

Hoping for someone to get ahold of themselves in the middle of mental illness is not an expectation grounded in reality.

Same in Québec, where the legal minimum for guaranteeing against defects is 1 year (enforced at both the manufacturer and seller).

Usually the terms will use wording like "30 days or the minimum applicable by law".

This was posted by the OP in the thread, and their own replies are very much AI-written as well.

I looked for the same, and no, not currently. It's only local models or LM Studio's own API.

Work Louder is a different company, the LoveFrom hardware is still unknown at this point.

A functioning democratic leadership listens when told they've done good, and listens when told they messed up.

Arguing that they should receive no support or positive reactions because they also deserve blame is how the center and left break down their own power: Believing that disengagement from one another is a stronger moral obligation than working together and fixing shit with people who are willing to listen and work.

I think you'll need to give people an example of where you think the center is.

(And be prepared for a lesson in political literacy from outside of the US)

iA Writer has Mono, Duo and Quattro fonts, with the latter two being almost monospaced. They concede some size variations for specific characters (Duo has 150% width characters, Quattro also uses 50% and 75% for narrow characters).

It's a fun subtle adaptation to keep close to the typewriter-like experience of the app.

https://ia.net/topics/a-typographic-christmas

I'd lean the same way as you (just a hypothesis from me too). A .app on MacOS is just a special kind of directory, so the compression covers the normal file types inside of it.

Localization files compress well, compiled code compresses well, repeating assets (@1x, 2x, 3x) and the pair of binaries in a universal app (x86_64 + arm64) do too, etc.

Ah, and dmg compression is just LZFSE, zlib or bzip2, so pretty standard stuff as far as I understand it.

Some (lots of) people will trade a lot of money for general life freedom. If it's well-booked, a service like this can come to around 105k/year for each dev.

A salary like this is only a big compromise if you live in a very high cost of life area.

A lot of Mac apps compress like this. Not so long ago, it was pretty common to download a 3-400mb dmg file that decompresses to a 1.5gb app package, for example.

You're applying US cultural logic to Chinese bureaucracy

They won't? They're profiting from every subscription to AI providers, because all companies are state-owned.

They're also well-positioned to control how groundbreaking US AI companies are allowed to appear to their investors, by offering open models that match what market leaders claim can only be attained with trillion-level spend. That's a strong control on US economy, considering how few stocks are propping up the US stock market, and how these stocks are all dependent on the same beliefs and factors.

Also, regarding China's tech investment in general, the five-year plan is just... available to read. It's not a secret. It explains the strategy, and you can draw a direct line from the plan to their now abundant solar infrastructures and tech achievements, including in AI, which is specifically named.

https://cset.georgetown.edu/publication/china-14th-five-year...

> "Ring" means you are navigating linearly and circularly. Isn't it better to provide a list of links so users can choose where they want to go "next"?

That depends! If your goal is some optimal way of navigating, sure. But the shape of a webring means your site has two "neighbours", which is fun and uses spatial notions we don't see often anymore. It also means someone can remember Site Y as "the one next from Site X", which is also fun and creates more ways sites relate to each other. This relating to one another is usually the point of a webring.

"It really whips the llama's ass"

Archived soundbite: https://archive.org/details/youtube-LWKxtWEkS1c

It's also a financial optimization: Even with the very expensive RAM upgrade prices, it's way better to use the same memory chips to manufacture and sell four Mac Minis or two Mac Studios, rather than a single high-RAM unit.