svachalek

Often it's taken out of context by pop sci channels to propose the universe "could be" or even "is" in a black hole. But from my understanding, as a non-physicist, it's not surprising or interesting to most physicists that the math works out like this, and there's nothing else to suggest that we're in a black hole.

It's kind of crazy to think about. But silk is probably even crazier.

Same. Usually found next to the wax soda bottles. Even as a kid I thought they were objectively terrible as candy went (like a pretzel stick coated in sugar) but there was something cool and grown up about puffing a fake cigarette at age six.

Maybe you shouldn't take the "you" of this article personally, but as a reference to your 100 mouth breathing neighbors. Do you think the average person who spends their evening reposting troll memes and watching Tiktok videos is now using AI to enhance their understanding of the world?

To be fair we've already had decades of advancements that allow society to more than ever avoid expending any effort on actual thought, but I can only see AI continuing and expanding the trend, not reversing it.

Yeah there's even an "Agent SDK" from Anthropic that manages Claude CLI sessions for you, and a separate token budget for apps that use it, on top of the subscription tokens.

Thinking about a scary moment can make your pulse race, your eyes dilate, and cause you to sweat. Thinking about a lover can do all that and other things. So clearly the imagination can affect the body. Not saying this study will hold up necessarily, but it's not that far fetched.

I've been thinking recently, there's hardly any reason not to use a native language like Rust anymore. The classic reasoning for languages like Python is that it prioritizes developer time and resources over compute resources, but now developer time is rapidly trending to zero and everything is compute dominated.

It's not really all that fast, for most things it's full seconds or at least a significant portion of a second. As computing goes that's quite slow and there are many things that require much faster speeds than this. However compared to the typical LLM it's quite speedy and opens some new doors for real-time-ish scenarios.

Oh /loop was the crowning achievement. Basically guarantees you'll run a prompt at max context without caching over and over.

My problem isn't that it can identify such conditions, the record on cybersecurity lately shows these things are quite good at finding real faults. But the annoyance is that they always want to fix it by adding more things that open cracks. It's possible, sort of, to drive it to fix by removing things instead but it's a battle.

The sad thing is I think it's like Elara, it's just a favorite LLM name for things and there's likely not a human behind it.

I think these are cheap and easy toy models to toss out for the big labs, to see what they can do with limited inference resources. I'm not sure what the plan is for a company like LFM that seems to be doing this as an entire business plan.

I see it as a niche now but it will hopefully grow in the future. The amount of information these servers are collecting about you for potential use against you is staggering and makes adware spy tech look like the good old days of privacy.

Strongest and easiest use case now would be just to hook one of these up to a search engine tool and ask your random questions to it rather than adding to OpenAI or Gemini's database about you.

I think the use case for this sort of model would be building something like Google or Apple's agents on device. Find an email for me, write a message for me, add an appointment to my calendar etc. Simple agentic stuff, not writing apps.

And when you do find people out, they're wearing headphones and/or their nose is in a screen, and if you try to talk to them they look startled and annoyed like how dare you speak to me in a public place. The culture is just gone.

I agree. The latest models are getting quite good at validation, as long as they are clear on what the requirements are. Recently I got frustrated that while we have a huge test suite I noticed the LLM was hardly ever breaking a test with its changes (PRs only added tests, never changed old ones) which seemed suspicious.

I had it do a full audit. There were about 100k lines of tests, of which it determined about 60k were near useless. Many were overly mocked, some mocked to the point that they really weren't testing anything at all. It deleted those and wrote more expansive scenario tests to replace them, caught a lot of bugs on the way and fixed those too. Total time was about 2 hours.

I think humans need to review requirements, specifications, and test scenarios. And for some industries and uses they're still going to have to review every detail. But for a typical CRUD app, I think we're at a point where we can already set the process and quality bar higher than ever, in addition to moving faster.

4xx errors are for when the client did something wrong and needs to know about it. 5xx errors are for when the server is having problems and the only people who can do anything about it are the operators, who should be looking for errors in the logs rather than the HTTP client.

Truly "application level" issues are therefore correct as 200, although some application communication issues lead to 4xx errors that should be handled as such. But the case of an application trying to communicate via 5xx errors should never have been the case anyway.

You're a victim of "sane washing". The media has been suppressed to the point of trying to justify every move of this administration as serving some kind of brilliant purpose. Surely the usual special interests are getting their gifts, but a lot of our national policies can strictly be explained by narcissism and dementia. Big Gulf and Big Lake aren't making billions off his other name changes.

Where does (2) happen? I couldn't find that. I see his followup speculating it was just a bug.

Opus often picks it when it's doing a "find me something" subagent. But largely it's been held back by being fully a year old at this point, and priced at a much higher price than models that are far more capable.

> If social media were truly “rewiring” the adolescent brain, we would expect the damage to be more consistent than a selective worsening on some measures and improvement on others.

Would we? A consistent trend is a consistent trend and "rewiring" doesn't necessarily mean all metrics need to go downward.

This is what worries me. The services are deliberately designed to be addictive and the proven technique for that is making users angry and/or depressed, these two emotions are found to be the most "sticky" and revenue enhancing.

Wow that game makes my head hurt. Cool idea!

Wow. Context management is one of the big remaining hassles with modern LLMs so this could be big. The obvious complication is cache busting so it's also exciting they investigated solutions for that.

Those are competitive advantages but a "moat" usually refers to being on top of the hill and able to keep anyone else from climbing. I think it's arguable whether Google is on top of the hill at all, and if they are they're sharing it with at least two other companies.

Absolutely. Objectively speaking it was far less consistent and capable than even small local models today.

I've been working on a game that has dice rolling and even knowing about this effect, I started going crazy yesterday when I had a long streak of numbers, like 1-20, it was 15-16 like 9 times out of 10. I was sure there was some kind of bug in how it was initializing random, or saving the number, etc etc. Just could not find it. Streak continued to another roll, another roll... still couldn't find it. Then the streak just broke. Apparently just random being random.

So you're living in the opening of Neuromancer: "The sky above the port was the color of television, tuned to a dead channel."

Really this is the entire economy now, giant companies that have billions invested in mass surveillance and heavily min-maxed dark patterns targeting the wallet of the average unsuspecting consumer. Nonsense subscriptions that are impossible to cancel, replacing ownership with "license", customized pricing, endless advertising, zero customer service, shrinkwrap licenses that change daily, etc etc etc.

"In the United States, to be considered an accredited investor, a natural person must have a net worth of at least $1,000,000, excluding the value of one's primary residence, or have income at least $200,000 each year for the last two years (or $300,000 combined income if married) and have the expectation to make the same amount this year, or must otherwise be a holder of a specific license in good standing."

Of course you can still lie but legally there is a line.

This isn't eating Jev's lunch. This is someone who doesn't understand the entire use case of Jev replacing it with something that doesn't handle it at all.