yesco

It's funny you mention this, because I've been thinking lately that all the focus on code quality benchmarks for LLMs seems to miss something. In my experience, LLMs aren't actually all that good at prompting other LLMs, and that's exactly the skill that matters for proper subagent usage.

The corpus of code available to train LLMs on is rich in both volume and quality. The same can't be said for LLM prompts. Prompts only came into existence alongside LLMs themselves, so most of the prompts out there are actually LLM generated, and training on them causes every model to overfit to the same patterns.

LLMs can give clear and concise directives to other LLMs, but only for basic tasks. Beyond that, they frequently overstuff their prompts with too much prose and too much direction, and worse, they often fill them with subtle contradictions.

Of course, there's a spectrum here. Some models turn out to be better at this than others, and I suspect this is what differentiates Fable-type models from the rest, and why vendors have started keeping their models' reasoning confidential.

I think this is one of those areas where the translation from research to the public ends up too precise on the details, beyond the cognitive load most people are willing to commit to it day to day.

In the same spirit of upāya from Buddhism, I like to compress this type of topic into a simpler model, even if imperfect.

Here's how I model cancer, and how I judge whether something is likely to give me cancer, without trying (and failing) to memorize a million risk factors individually:

You are a collective of cooperating cells. When cells rebel, the order that holds your body together collapses, and you get cancer. Cells generally don't want to rebel. They're content as long as they get their energy, their paycheck.

But what about activities that cause frequent cell death? A sunburn, smoke in your lungs, alcohol in your bloodstream, radiation (essentially atomic shrapnel flying through your tissue). Anything that causes inflammation, especially chronically. When cells die before their expiration date, the survivors have to keep dividing to replace them, and they start to get pretty mad about it. Maybe even rebellious.

With every round of division, they are in essence "evolving" away from their cooperative nature to cope with a hostile environment. A cell doesn't care whether the collective lives or dies, and it doesn't have the capacity to reason about that in the first place.

So the question becomes: are you keeping your cells safe? Are your activities hurting them? If they're not happy, they'll rebel, and you'll die.

This is where the model earns its keep. Take obesity. It's on every risk list, but excess fat isn't what makes cells rebel. The real culprits are the conditions obesity creates: chronic inflammation, more cells that need to be maintained and replaced, a body that is generally harder to live in.

Weight loss is on the list because it's the easiest way to fix those conditions, not because thinness is protective in itself. Once you see that, you don't need the list. Anything that makes your cells' lives harder, day after day, is raising your odds, whether or not anyone has published a paper on it yet.

The math is real, but a model like this tells you which variables you actually control.

Tech leaders demand nations of the world give them a moat or else

Because they aren't and never have been! This has been scientifically proven from the astrological data according to my sources.

Somewhat tangential, but this reminds me of when I took a certain Microbiology midterm in college. I hadn't put much effort into that class leading up to it, as I had some tough CS classes that consumed my focus that semester (also was admittedly not a great student in general). So when I finally went to take the midterm, after cramming like mad before the exam, I quickly realized I basically didn't know the answer to most of the questions, even what I crammed was only a small portion of the material.

The exam was huge, at least over 10 pages, and even when we technically ran out of time, the professor was kind enough to move the remaining exam takers to the neighboring lecture hall to continue taking it. I recall I spent a total of 2 hours on that exam.

Now mind you it was mostly short answer or multiple choice questions. The multiple choice questions were pretty sharp too, lots of traps and false but sounds right answers mixed in. But if it had been purely essay questions, I would have been screwed.

However with such a huge corpus of information in front of me, I ended up basically learning all the material on the spot. I just kept doing multiple passes through it, each time I noticed one of my answers contradicted one of the others, I would make adjustments to harmonize, which indirectly refined my understanding.

In the end I got B+ in the exam (which was curved to an A), and walked out understanding the material better than I did walking in.

Reflecting in the experience years later, I've wondered if a hypothetical LLM which was ignorant of microbiology could do the same thing if fed that exam. In some respects the traps they placed in the multiple choice questions actually were what helped me refine my understanding the most. Made me appreciate information theory more.

Yeah I walked into this with my pitchfork ready but isn't this stuff pretty... reasonable?

McDonalds is just a restaurant, they predict what menu items you like and how likely you will come back, is that really a problem? What is the concern here? I don't use their app but presumably for those that do, isn't this exactly the type of data that would simply influence whether the app offers a coupon or something? Or how well a franchise is doing? Maybe also what to keep on the menu or how much to stock?

What exactly is the bar here I wonder? McDonalds has security cameras, it also has menu kiosks with cameras (I think?). The alternate reality evil tech-like McDonalds could have theoretically used facial recognition to categorize all footage of any given customer, then used this footage to estimate their friends, spouse, BMI, age, gender, ethnic group, attractivness, disability, etc. They could then sell it to third party data brokers for extra cash.

That would certainly be the kind of dystopian surveillance to get me swinging my pitchfork around, but this? Having assumed the worst I'm now walking away realizing McDonalds might not be so bad after all.

I reject your characterization of my comment. I explicitly called myself an outlier and said I didn't want to get too specific, so I'm not sure where you're getting "clever" from. I wasn't claiming credit for my circumstances. I was identifying the actual variable that separated my outcome from peers in similar circumstances, because that's useful information for anyone who does have that option available.

And plenty of people do. I know people who could live at home rent-free and chose apartments anyway because they wanted independence sooner. That's a legitimate preference, but it's still a tradeoff, and pretending the tradeoff doesn't exist because some people don't get the choice at all helps nobody. "Not everyone can do X" is true of literally every piece of financial advice ever given. It doesn't make X bad advice for the people who can.

I purchased a house within commuting distance of Boston in my mid 20s, back in 2022. Didn't need PMI, the mortgage terms weren't unreasonable to me (fixed 5% interest), and the monthly payments for utilities plus the loan are still well within my means. It's even got an attached garage.

I don't want to get too specific here for privacy reasons, but I believe the main difference between myself and my peers is that I chose to live with my parents and save for a house instead of moving out and getting an apartment right away during/after college. With almost zero real expenses outside my student loans, and the median software engineer salary for the region, I was able to save enough for a down payment after just a couple years. I did this because I disliked the idea of having no equity and paying rent to offset someone else's mortgage instead of saving for my own. My parents really liked that reasoning.

In many respects I'm fundamentally an outlier, yet once I bought mine, my peers with much lower incomes started doing the same after I explained the financials. When they actually sat down and calculated the mortgage payments, they realized it wasn't that bad, as long as they got something on the smaller side, which is still bigger than an apartment anyway.

To be clear, housing costs were crazy and still are. But I sense "crazy" is pretty relative when talking to people online from other countries. Housing in the US is expensive, especially compared to various convenient periods within the last 100 years. But even now, it's not the unrealistic back-breaking dream it seems to be when I talk to Europeans about it, and I suspect that's why so many Americans on social media overestimate the costs.

My perspective until now was that the Wikimedia foundation was already supposed to be a union-like organization. Would it make sense for Linux maintainers to form a union within the Linux foundation? The vibes feel similar to me.

Yeah I was expecting some giant text dump based on the complaints in the comments here, yet all I see is a pretty normal README? It's not even that large.

If we are splitting hairs some sections could have been left out or broken up into a dedicated doc, but this is still vastly better than a README that explains too little. At minimum I expect a README to give me guidance on what this is and how to get started with it and having stuff like a file layout helps with exactly that.

Yes it's truly a shame when the people responsible for critical infrastructure only hire the cheap inexperienced software engineers to build it.

Much like how relying on a resident nurse instead of a cardiologist when you are experiencing a heart attack can end your life pretty badly as well.

I see no reason not to be empathetic. The frustration is fair, but it's aimed at the wrong layer. These people were guided into this spot by bootcamps and curricula that start at React and never go down the stack.

My experience was the reverse. I learned HTML and CSS first, then Rails in college to serve templated pages. I understood the client/server boundary fine as a concept, what I couldn't see was where it actually sat in a web context. I sort of knew JavaScript ran in the browser, but then I'd see ERB templates stamping values directly into script tags, so the server was writing the JavaScript that ran on the client, and my mental model fell apart. Where does my code actually execute? Why does this variable exist here but not there? Why does the page have data the network tab never fetched? Nobody ever sat me down and explained the request/response lifecycle as its own thing. I had to assemble it from fragments over years. This was around 2017 for context.

How you learn something shapes how you keep learning. If your mental model is misaligned, everything downstream is friction. The thing that finally made it click for me was reading the actual HTTP RFCs, which is apparently a weird thing to do, because HTTP itself is absent from nearly every guide and curriculum. Tutorials teach you the framework, maybe the language, and just assume the protocol underneath. These days I make newbies read the MDN docs like a book and skim the HTTP wiki page, learn the history of the protocol. It's short! It's not even a book! That gives you a firm foundation. But if your foundation starts at React, drilling down is like digging past bedrock. People don't know where to start, and Googling only shows them wrong answers because they don't yet know how to ask the question.

You can agree or disagree with the consequences, but the voting rights act never had any explicit provisions about districting, this was something conjured entirely by the courts. It was even framed as a temporary measure at the time of the original ruling.

So not exactly bewildering, I personally saw it as closer to inevitable. The Supreme Court never had the power to legislate, it can only interpret, and a shaky interpretation always has an expiration date no matter how popular it is.