DonsDiscountGas

Even if you're "stuck" at 1, a person with a reasonably high salary can buy equity in the public stock market and start capturing exponential growth. And if they can outperform the market by even ~5% that can really add up[0]. Not to billionaire status but maybe it's okay to not be a billionaire.

[0] Big "if" obviously but so is a 15% monthly growth rate in revenue.

Wouldn't it make more sense to remove the supply constraint?

Engineering always involves trade-offs. The right architecture for the military in Iraq might not be the right architecture for sharing cookie recipes with your family

True but supply chain attacks are real. A good security plan needs multiple layers, this seems like a good one to include.

In my experience the accuracy was fine but actually reading the output was so annoying I removed it.

What's your definition of "required". I usually don't tip ride share drivers

America in a nutshell. We won't do anything about mass shooting events besides hold up deployment of a life saving technology.

Engineering should be exactly as rigorous as the task requires. Anything more is wasteful.

A conversation with your boss about a COLA raise really shouldn't include your own personal finances. "I just bought a house" is not a good reason for a raise; "prices in our area have increased" is a much better one

Oh yeah tons. I'm not looking to invent just tinker

According to Claude it requires active control. So complicated and only used in very niche applications.

You sure can, by not using Claude code.

You can set billing/usage limits pretty easily. And the default settings on subscription (IIRC) are no extra usage. So I have no idea how they could bankrupt someone like that.

You'll notice by the crappy output assuming you're paying any attention at all

Not the OP but my Fable says something to the effect of "I'm afraid I can't do that Dave" and very obviously (not secretly) downgrades to Opus. I work in life sciences.

Only slightly related but does anyone know anything about motors with magnetic bearings? As in, no contact or friction. I'm looking for a hardware project

Westlaw is very expensive

You could try actually reading the code that it wrote

I am shocked at the low scores from previous models. Maybe I just have low code standards but I've generally been vibe coding since 4.6

I have no idea how you came to that conclusion. Unless your training pipeline involves actively querying one of Anthropic models, no they can't. And if it does you're distilling their model.

I don't think offering a product under a certain set of terms obligates a company to maintain that offering forever. The bait and switch is certainly annoying but seeing as they're very upfront about it you can't say you weren't warned. Don't like it? Don't use it.

I expect that depends on demand, feedback, and whether GPT-6.0 gets released and is competitive

So you placed limit orders at the same time but sometimes only one of them filled? Do you think it would be possible to reduce the frequency of that happening by adjusting the limits? Less profit per arb but reduce risk of missing a leg

For the purposes of content creation they don't even have to be right once

It's pretty likely that inference will get substantially cheaper. His argument is that for these companies to be profitable some very major and (pre 2022) unprecedented things have to happen. Which I tend to agree with, except I think they will happen, seeing as how they've been happening for a few years.

"fixed" is definitely incorrect but there's probably a ceiling on how fast the demand can grow, just because other bottlenecks will take over at some point.

AI/LLMs have been dramatically improving for 7+ years. There's now a lot more funding to support continued improvement. You're correct this is an "assumption", but continued improvement at the same pace (or faster) for the next 3+ years is just extrapolating a trend. Believing we've hit the top today is based on nothing at all. Continued improvement is much more likely.

Concerning but not really surprising. They offer about hundred thousand antibodies, a few hundred frauds is likely the tip of the iceberg.

> “Similar image” searches using Google Lens, Bing Images or DuckDuckGo betray hundreds more that we have yet to document

In my experience these would return any image of an antibody (edit) Western blot, not just the exactly matching background. Would be curious to hear others thoughts.

Seemed to work when it comes to selling ads. I'm thinking training LLMs is harder than anthropic and openai make it look

> My main motivator has always been to understand how things work

Which things, specifically? I got started programming in order to control scientific instruments and analyze the data I collected. I care about the systems I'm studying. I don't really care much about the software I'm programming. As OP said it's a means to an end.

I think you're right that people on HN are generally more curious but it doesn't have to be curiosity about every single element of software. We don't write assembly code anymore either, half of software is about making software easier and faster to write.