jfim

That's kind of how it feels though. I get the impression I'm micro managing various Claude code instances in multiple terminals.

In Claude code I believe it's /context and it'll give you a graphical representation of what's taking context space

You can also tell it the optimization to implement.

I asked Claude to find all the valid words on a Boggle board given a dictionary and it wrote a simple implementation that basically tried to search for every single word on the board. Telling it to prune the dictionary first by building a bit mask of the letters in each word and on the board and then checking if the word is even possible to have on the board gave something like a 600x speedup with just a simple prompt of what to do.

That does assume that one has an idea of how to optimize though and what are the bottlenecks.

What are the axes on the chart? Rpm?

I agree, it'd actually be great if they did give maybe $5 or $10 worth of API tokens per month to max subscribers, since they're likely to be the most likely to actually build stuff that uses the Claude APIs.

I built a quick thing to download YouTube videos and transcribe them using with whisper, but it kind of feels clunky to summarize them using the claude CLI, even though that works.

> I'm not sure what the best use of $1k is from a health standpoint is, just noting that it's good to have a comparator.

Spending 1k on a gym membership and more fresh vegetables would be a pretty high return on investment, if one isn't in shape and eating healthy already.

Keep in mind there are two Americas, a wealthy one and a not wealthy one; someone posting on HN is likely in the former bucket, and not juggling a retail job and doing Uber on the side while being unable to afford healthcare.

It depends what you include in cost of living. Healthcare and education tend to be more expensive than other first world countries, and the taxes are high given the fact that healthcare and education aren't part of what one gets from paying taxes.

I'm a bit disconnected from Canadian politics these days, but wasn't Pierre Poilièvre polling quite well compared to the liberals until the US election?

And while it's true that there are no confidence votes, Harper has prorogued the parliament before in order to prevent such a vote.

Canadian politics is healthier in general than in the US but it's not a panacea.

The biggest problem with Canadian capital markets is that investors aren't really that comfortable with taking risk, because the alternative for them is to park their money in real estate, which has done quite well with little risk in Canada.

If Carney really wants to make this happen, he'd have to thread the needle on deflating the property bubble in order to make venture capital and productivity investments more attractive compared to real estate, all while managing the rest of the highly US dependent economy for the upcoming years.

Switzerland has fewer sociocultural problems than the US, but it's also a smaller relatively homogenous society with far less immigration, and a highly educated and politically involved population.

You'd be surprised how shoddy human translations can be, and it's not necessarily because of the translators themselves.

Typically what happens is that translators are given an Excel sheet with the original text in a column, and the translated text must be put into the next column. Because there's no context, it's not necessarily clear to the translator whether the translation for plane should be avion (airplane) or plan (geometric plane). The translator might not ever see the actual software with their translated text.

Optimizing for capital returns is a simplification of the real world, where it allows for comparing whether it makes more sense to put one's money into opportunity A or B.

There's a lot that's not captured by solely looking at dollars, like the examples that you bring up, such as quality of life, human welfare, and so on.

That's what's meant by efficiency, it's allocating it to the place that has the highest return on investment.

As you point out, in practice what's efficient is what can capture the highest return, not necessarily the highest return per se. If say investing in education had high returns society wide but those returns couldn't be captured, that's not an efficient use of private capital.

It's been a while since I touched Scala but wasn't that a thing in previous versions, minus the braces not being present?

They could use the json.org license: https://www.tldrlegal.com/license/the-json-license

It's literally the MIT license with an added clause of only using the software for good, not evil.

Obviously, corporate attorneys will advise not to use the software since good and evil aren't really well defined legal terms. It's also not open source using the osi definition.

The smell from those though is quite something

Because people aren't going on AWS for EC2, they go on it to have access to RDS, S3, EKS, ELB, SNS, Cognito, etc. Large enough customers also don't pay list price for AWS.

A training corpus that includes the images from Google image search probably helps a lot.

Counting letters is tricky for LLMs because they operate on tokens, not letters. From the perspective of a LLM, if you ask it "this is a sentence, count the letters in it" it doesn't see a stream of characters like we do, it sees [851, 382, 261, 21872, 11, 3605, 290, 18151, 306, 480].

The automatic wipers on non Tesla cars use infrared sensors that have existed for decades, so they're a well known quantity. The Tesla wipers use the front cameras to detect rain, and those cameras are focused at a distance that's far enough to be able to see other vehicles, so they're not focused on the windshield, which is why they're unreliable.

Hadn't heard of jj, it's a source control tool that advertises that it's fast and compatible with the git on disk format. https://github.com/jj-vcs/jj

It is a scalding hazard though. At 70°C skin burns occur in about a second or so.

The thermostatic valve makes it so that the water that comes out of the water heater is at a more reasonable temperature.

It's easier to get from zero to something that works on divided highways, since there's only lanes, other vehicles, and a few signs to care about. No cross traffic, cyclists, pedestrians, parked cars, etc.

One thing that's hard with highways is the fact that vehicles move faster, so in a tenth of a second at 65 mph, a car has moved 9.5 feet. So if say a big rock fell off a truck onto the highway, to detect it early and proactively brake or change lanes to avoid it, it would need to be detected at quite a long distance, which demands a lot from sensors (eg. how many pixels/LIDAR returns do you get at say 300+ feet on an object that's smaller than a car, and how much do you need to detect it as an obstruction).

But those also happen quite infrequently, so a vehicle that doesn't handle road debris (or deer or rare obstructions) can work with supervision and appear to work autonomously, but one that's fully autonomous can't skip those scenarios.

Companies can be successful despite bad practices though. Google is successful and they developed multiple competing chat apps, but that doesn't mean that developing multiple chat apps will make one successful.

If you want to build a quick one, it's yt-dlp to download the video, whisper to transcribe the audio, and Claude code have it summarize the transcript.

I'm not at my computer RN but I'll share it later.

> This sets a dangerous precedent for selective enforcement of ToS really

Companies can enforce their terms of service as they see fit, including enforcing them selectively or not at all, with very few limitations. They're not bound by precedent as courts would be, nor do they need to be fair.

I'm skeptical too, but the fact that it works slower isn't too much of a problem if it doesn't require human attention and finishes before one is back home. It's just like how the Roomba can take as much time as it needs to to vacuum the living room when I'm gone for the day, as long as it's done by the time I get back.

> I guess the big question is if experienced product management types can pick up enough coding technical literacy to work like this without programmers

I'd argue that they can't, at least on a short timeframe. Not because LLMs can't generate a program or product that works, but that there needs to be enough understanding of how the implementation works to fix any complex issues that come up.

One experience I had is that I had tried to generate a MITM HTTPS proxy that uses Netty using Claude, and while it generated a pile of code that looked good on the surface, it didn't actually work. Not knowing enough about Netty, I wasn't able to debug why it didn't work and trying to fix it with the LLM didn't help either.

Maybe PMs can pick up enough knowledge over time to be able to implement products that can scale, but by that time they'd effectively be a software engineer, minus the writing code part.

It's not clear if their design allows for it from the details in the article, but you'd want a chain in case you run out of battery so you can pedal manually.

Maybe the future of human connection is chatting with a large language model, at least according to Meta. Haven't they added chatbots to messenger?