surround

btw Flock was funded by Y Combinator https://www.ycombinator.com/companies/flock-safety

In the first paragraph Karpathy says he used Opus 5.5. The parent comment is asking if he used a different harness besides claude code.

This article is missing the point. $GOOG didn't drop because search revenue is down. Search revenue, and revenue overall, actually grew. It dropped because the rate of growth is lower than expected, and the price of a stock is determined by how investors think the company will be worth a decade from now.

If you're brave enough to want to see it yourself, visit https://archive.ph/jXv3A#:~:text=filed%20on%20Tuesday then click on the "Filed on Tuesday" link.

It redirected me to https://stolidityseashellshotter[.]com/ezrsb5dd1268e6f6ebaad... which then redirected me to a fake clone of RT news: https://rurtnews[.]com/

Unlike the Reddit mods, I'm not getting redirected to the malware site, just the fake RT site.

Edit: I clicked it a few more times and its redirecting me to either the fake RT or https://gullible-thanks[.]com/w5FDGi, which redirects to either https://sportshard[.]com/ or https://eatcells[.]com/land/?token=[a random string]

I had to disable uBlock Origin, since it does something to block many of these redirects.

I'd rather not have to explain to people why I opened a website with a huge, animated banner of a cat lady twerking in a thong.

NSFW warning please. I clicked on this link while my screen was visible to to others and freaked out.

I highly recommend using the open source Sodium and Lithium mods to improve performance. And I hope you're not still using OptiFine!

https://github.com/CaffeineMC/sodium

https://github.com/CaffeineMC/lithium

The title is misleading. This is model routing, not distillation.

Poles. The video shows a tent they set up just by zipping up the y zipper.

Netflix never had everything. I seem to remember watching a skit 10 ish years ago that joked that Netflix only had bad movies.

The ads are what get me.

Minecraft gained popularity super fast in its early days, probably because it was a free-to-play game that worked in the browser (thanks to Java). Notch had no reputation or marketing at the time.

It seems that Deno made the right decision by choosing Rust from the get-go.

How does the voice model delegate requests to GPT-5.5? Can the voice model generate text?

If you look at a satellite image of the Great Salt Lake [1], it looks like there's a digital seam/glitch between the north and the south half of the lake. In reality, a railroad was built through the middle of the lake in 1904, separating the water. The salinity of the north half has since become toxic to all organisms except some algae and cyanobacteria.

[1] https://www.google.com/maps/@41.1985997,-112.4903027,201762m...

I'm not sure what you're referring to. If corn is replaced with solar panels that doesn't solve the fact that there will still be a ton of cars that demand ethanol-infused gasoline.

Its more efficient energy-wise, but isn't the issue that energy from solar panels can't power gas cars?

Fair enough. Though they certainly could still break in if the laptop isn't encrypted, so this tool is only useful when combined with disk encryption.

How do you define "state-level actor?" Police departments certainly have access to state and federal forensic resources to access unencrypted data in memory.

> in sensitive situations, law enforcement and border agents in many countries can compel a biometric unlock in ways they cannot with a password.

If the threat model includes state-level actors, then disabling biometrics won't prevent data from being retrieved from physical memory. It would probably be wiser to enable disk encryption and have a panic button that powers down/hibernates the computer so that no unencrypted data remains on RAM.

The website says shutdown "takes time" and "kills your session" but a hibernation button would take effect just as fast and would preserve the session.

It seems like Google's policy is unconcerned with the intent of the practice. If a website JS redirect ruins the user experience by breaking the back button, it will be demoted in search results. It doesn't matter whether or not the redirect was meant to be deceptive or malicious, websites shouldn't be ruining the user experience.

It seems Sam Altman has the same suspicion, based upon his response:

> There was an incendiary article about me a few days ago. Someone said to me yesterday they thought it was coming at a time of great anxiety about AI and that it made things more dangerous for me.

https://blog.samaltman.com/2279512

https://news.ycombinator.com/item?id=47724921

> There was an incendiary article about me a few days ago. Someone said to me yesterday they thought it was coming at a time of great anxiety about AI and that it made things more dangerous for me.

For context his blog post seems to be a response to this deep-dive New Yorker article:

"Sam Altman May Control Our Future—Can He Be Trusted?"

https://www.newyorker.com/magazine/2026/04/13/sam-altman-may...

https://news.ycombinator.com/item?id=47659135

Yes, DMCA made the mere act of breaking DRM illegal, even if what you do with the media is legal.

That's what the comment I was originally replying to was saying.

Try asking an LLM a question like "H o w T o P r o g r a m I n R u s t ?" - each letter, separated by spaces, will be its own token, and the model will understand just fine. The issue is that computational cost scales quadratically with the number of tokens, so processing "h e l l o" is much more expensive than "hello". "hello" has meaning, "h" has no meaning by itself. The model has to waste a lot of computation forming words from the letters.

Our brains also process text entire words at a time, not letter-by-letter. The difference is that our brains are much more flexible than a tokenizer, and we can easily switch to letter-by-letter reading when needed, such as when we encounter an unfamiliar word.

But for lisp, a more complex solution is needed. It's easy for a human lisp programmer to keep track of which closing parentheses corresponds to which opening parentheses because the editor highlights parentheses pairs as they are typed. How can we give an LLM that kind of feedback as it generates code?

I think you're right. Try asking GPT-5 this:

> Are the parentheses in ((((()))))) balanced?

There was a thread about this the other day [1]. It's the same issue as "count the r's in strawberry." Tokenization makes it hard to count characters. If you put that string into OpenAI's tokenizer, [2] this is how they are grouped:

Token 1: ((((

Token 2: ()))

Token 3: )))

Which of course isn't at all how our minds would group them together in order to keep track of them.

[1] https://news.ycombinator.com/item?id=47615876 [2] https://platform.openai.com/tokenizer

Wow I'm surprised, you're right, and it has happened before:

> the attacker issued and registered a free temporary 3-month certificate for the developers[.]kakao.com domain through SSL certificate issuer called ZeroSSL. Because the routing policy was already manipulated by the BGP Hijacking, the attacker was able to register the certificate.

https://medium.com/s2wblog/post-mortem-of-klayswap-incident-...