Jorge1o1

While I agree with your sentiment, I just want to say that if your approach is to have the LLM read every file into context, or you're working in some gigantic thread (using the million token capacity most frontier models have) that's really not the best way to do it.

Not even a human would work that way... you wouldn't open 300 different python files and then try to memorize the contents of every single file before writing your first code-change.

Additionally, you're going to have worse performance on longer context sizes anyways, so you should be doing it for reasons other than cost [1].

Things that have helped me manage context sizes (working in both Python and kdb+/q):

- Keep your AGENTS.md small but useful, in it you can give rules like "every time you work on a file in the `combobulator` module, you MUST read the `combobulator/README.md`. And in those README's you point to the other files that are relevant etc. And of course you have Claude write the READMEs for you...

- Don't let logs and other output fill up your context. Tell the agent to redirect logs and then grep over them, or run your scripts with a different loglevel.

- Use tools rather than letting it go wild with `python3 -c`. These little scripts eat context like there's no tomorrow. I've seen the bots write little python scripts that send hundreds of lines of JSON into the context.

- This last tip is more subjective but I think there's value in reviewing and cleaning up the LLM-generated code once it starts looking sloppy (for example seeing lots of repetitive if-then-elses, etc.). In my opinion when you let it start building patches & duct-tape on top of sloppy original code it's like a combinatorial explosion of tokens. I guess this isn't really "vibe" coding per se.

[1] https://arxiv.org/html/2602.06319v1

Well, pretty much all of the LLMs are based on the decode-only version of the Transformer architecture (in fact it’s the T in GPT).

And in the Transformer architecture you’re working with embeddings, which are exactly what this article is about, the vector representation of words.

That’s actually a really good point! I wonder if one day there will be analogues for music and writing?

I suppose you have gumroad / serialized novels or webcomics but I’m not sure if there’s any albums where the musician is putting out one song at a time

Well, to me it seems like he just shared the original so that others could benefit from the work he had already done, but that since his main priority is to continue making new videos, he may not have the time resources to:

- Avoid breaking changes

- Keep APIs stable

- Test and document everything, etc.

I personally think there's nothing wrong with that. We wouldn't say that a musician is *obligated* to put out a second album or a remaster. We wouldn't say that an author *must* make a sequel to their popular book. But when it comes to code sometimes we feel like the original author has an obligation to keep working on it just because it would convenience us.

(edited for formatting)

Monkey see, monkey do?

Andrew Yang launched a presidential campaign based on this idea, he wrote a book:

“Smart People Should Build Things”

Sibling comment mentioned the connection between lox and Pisces, but the connection to Cancer is that the mascot for Rust is a crab.

I think it’s health related, as the article mentions.

>One executive in the entertainment industry said younger people were less inclined to go out raving until 6am as they were more health conscious and less frivolous with money than previous generations

This is the same generation that has 12 step skincare routines, eats only organic food, chooses to vape or zyn rather than smoke because of secondhand smoke, everyone has an Apple watch on their wrist tracking calories, etc.

If anything I’m surprised that binge drinking and going out late as survived as long as it has.

And as far as the money comment, this generation is not less frivolous there’s just less money to go around haha.

thank you!

A classic example would be that if you have a test (say for cancer) with a false positive rate of “””only””” 5%, and your disease has an incidence of say 1 in 1000.

Let’s say that you get a positive diagnosis for the disease, and you ask someone the question:

What is the probability you actually have the disease?

Most people will say 95% or 99%, but your actual probability of having the disease in this example is <2%

Is this Q? or k?

Well, theoretically you could make your current "Add Link" button and your current "Remove" button just trigger a server-side request and then refresh the page.

maybe some combination of the <noscript> tag and then if so wrapping the buttons in <form> and making the buttons submit those forms?

As a kdb+/Q programmer I would say it depends on the type of problem.

For example, when working with arrays of data it certainly is easier to think and write “avg a+b” to add two arrays together and then take the average.

In a non-array programming language you would probably first need to do some bounds checking, then a big for loop, a temporary variable to hold the sum and the count as you loop over the two arrays, etc.

Probably the difference between like 6ish lines of code in some language like C versus the 6 characters above in Q.

But every language has features that help you reason about certain types of problems better. Functional languages with algebraic data types and pattern matching (think OCaml or F#) are nicer than switch statements or big if-else-if statements. Languages with built-in syntactic sugar like async/await are better at dealing with concurrency, etc.

Ebay is full of $1 timeshares. There’s even a website with 17,000 timeshares for a dollar called sellmytimesharenow.com

Relying on an economic system to make moral decisions is like driving in screws with a hammer.

Capitalism is not perfect, but at least to me this is more of a moral issue than an economic one.

I think your CO2 point is a great example, why would a Communist or Socialist economic system be inherently more eco-friendly than a Capitalist system? You now have a system where there is a government willing to "plan" or "intervene" in certain aspects of the economy, but unless the government actually cares about pollution nothing will happen. You could even theoretically have a centrally-planned economy that WANTS to increase emissions.

The largest polluter in the world is a Communist country, and in the last 30 years their CO2 emissions have sextupled, while the US's emissions per year have remained flat.

Morality != Economics

That’s accurate enough. I think the workflow was more built for a q dev occasionally dipping into python rather than the other way around.

I think you touch on something really interesting which is the kink in the kdb+ learning curve when you go from really simple functions,tables, etc. to actually building a performant kdb architecture.

Their pykx integration is going a long way to fix some of the gaps in:

- charting

- machine learning/statsmodels

- html processing/webscrapes

Because for example you can just open a Jupyter Notebook and do:

  import pykx as kx
  df = kx.q(“select from foo where bar”)
  plt.plot(df[“x”], df[“y”])

It’s truly an incredibly seamless and powerful integration. You get the best of both worlds and it may be the saving feature of the product in the next 10 years

“Slave labor” is not a hyperbole or commentary on global wage differentials, or underpaid workers.

It’s legitimately 2 million enslaved Muslims imprisoned, working for $0.

https://selectcommitteeontheccp.house.gov/sites/evo-subsites...

I think the bigger issue than the environment, not to undervalue its importance, is that Shein and Temu literally use slave labor.

Archive.is times out if you're behind a VPN or on a VDI.

For example, my work virtual machine on Azure can't ever use it.

You can try https://replayweb.page/ as a test for viewing a WARC file. I do think you'll run into problems though with wanting to browse interconnected links in a forum format, but try this as a first step.

One potential option but definitely a bit more work would be, once you have all the warc files downloaded, you can open them all in python using the warctools module and maybe beautifulsoup and potentially parse/extract all of the data embedded in the WARC archives into your own "fresh" HTML webserver.

https://github.com/internetarchive/warctools

Funny that you mention perfect-pitch because in a similar vein it seems like one of those things that can be taught but the window closes very early in childhood.

https://journals.sagepub.com/doi/abs/10.1177/030573561246394...

So the source code of the bot is here: https://github.com/Kharacternyk/paper-tactics/blob/master/pa...

It seems like there's an evaluation function but there's no state search / recursive minimax tree for the bot.

I think if we added a 10-ply lookahead minimax algo for the bot here it would be pretty good. [1]

I will probably do it myself if I end up having a bit of free time this weekend.

[1] https://en.wikipedia.org/wiki/Minimax#Minimax_algorithm_with...

"For a Linux user, you can already build such a system yourself quite trivially by getting an FTP account, mounting it locally with curlftpfs, and then using SVN or CVS on the mounted filesystem."

RIP AST SpaceMobile

If you're an investor in a company, theoretically you can think of the company as a stream of future dividend payments (cash flows) discounted over time.

So maybe the company has $100 in profit in 2024, and they have 5 shares, and they pay out all of their profit as $20 dividend per share. And in 2025 you expect them to make $200 in profit, with the same 5 shares, which is $40 per share.

Although that's a theoretical model, and there's many companies that don't pay dividends at all and still command mighty stock prices (e.g. TSLA) this is how real companies are actually valued -- on the expectation of future cash flows, maybe times some multiple.

The reason why revenue is more important than profits for the vast majority of technology / growth-stage companies is that most of these companies are focused on creating double-digit percent revenue growth every single year. Their profits will be 0 or negative because all of the money that they make from selling widgets is getting spent on marketing or hiring new engineers to build more widgets, or building a factory.

But after 5-10 years, this company will have "scaled over" its fixed expenses like rent, they will have huge revenues, they'll have serious market share, and then you can simply slow down / turn off the marketing and other SG&A expenses, and then suddenly you go from being a 0 or negative profit company to a multi-billion $ profit company.

But if you're too focused on clipping coupons and scrounging pennies as a startup founder, you aren't focused on the right aspect: revenue growth.

Many buildings are deliberately designed with mirrored glass on the first floor and lots of trees so that they can "blend in" with the urban landscape and look more "natural", less "artificial".

What's interesting is that what we as humans consider more natural is actually antithetical to nature, because when the birds see the reflected trees, they can't parse that it's a reflection and fly straight into it -- leading to bird deaths.

If you bought a $17k Rolex Submariner back in 2015, and they told you it could no longer be serviced, I think many people would be outraged.

Rolex's own recommended service timeline is every 10 years, so you would actually be told it's obsolete before the first recommended service.

Wow. The use case that comes to mind for me is when you take a big family photo (or 20) and someone inevitably ends up cut-off by accident.

So then you just feed RealFill the 20 pictures you took and your uncle is magically painted in.

I’ve had to fill out forms for new algorithms / quant strategies with questions like:

- how many orders per minute do you expect to create?

- how many orders per minute do you expect to cancel/amend?

- what’s your max per-ticker position?

- what’s your max strategy-level GMV/NMV?

Etc.

Any one of those questions can be used to set up killswitches.

[edited for formatting]