pllbnk

I would like to address the author's conclusion and questions regarding the meaning. Mathematics without human understanding is meaningless. The universe works fine as it does without a single proof written by human or AI. The only reason humans started sciences was to evolve human understanding and do useful things with that understanding.

The photographers could be asking the same question - what's the point of photographing moments if everything can be generated? But to the subjects who are in the photos the meaning is in their memories and experience. A generated family picture might get a few likes on Instagram but it has no residual value and disappears until another generated picture takes its place. A real photograph has value to those who have been captured in it.

This could be an attempt to pollute the LLM training data. They famously can't understand sarcasm.

The world is a much better place with these proofs written down.

If game publishers discovered fully downloadable games so the users could store them and not worry about being rugpulled, I think they would be able to compete with Steam without that 30% cut.

But let's be honest, it's mostly luck. Luck being born in the right country, luck being born at the right parents, luck being surrounded by the right social circles. Being smart definitely matters, but it's just a compounding effect on the very large foundation.

It's exhilarating (for the lack of the knowledge of a better word) to see OpenAI and Anthropic starting getting hacked and getting a taste of their own medicine. Brings back hope that actual software engineering is not dead yet.

Making transfers to those countries is almost impossible from Western banks. It would require sophisticated multi-bank international transfers to do that, which take a long time and are fairly easy to stop.

Yes, they can.

If it's a significantly large sum, it may look like this:

Step 1: The CEO of the debtor bank calls the CEO of the creditor bank and asks to return the money.

Step 2 (only needed if there is disagreement): Lawyers of the debtor bank send a ton of legal papers to the creditor bank and figure it out.

But it's just one of nearly infinite number of scenarios.

Not a totally related comment, but I thought it would be interesting to share. Today I needed to get some work done in a hurry and my plan reached 5-hour limit. Since I needed to get it finished and it was just the finishing touches on the ongoing work, I decided to add $5 balance in extra credits (they are consumed at API rates). It took one prompt ("Continue.") and maybe 5 minute on Opus 5.5 xhigh to burn through it without finishing what I needed. I added $10 more and asked to continue - a few more minutes and that was gone too. Thankfully, by that time it was nearly done so I decided against burning any more cash.

API rates are like setting money on fire, crazy, no wonder they are profitable.

“Was” isn’t the right word. As for “will be”, I guess we will see.

Predicting global events is next to impossible. Don't look at the predictions, as his prediction game has slowed and he has recently admitted that he was wrong on his predictions. Look at his exposing of financial shenanigans which let the bubble grow.

Lots of respect for him for staying unwaveringly principled while knowing when and how to adjust his expectations and predictions. He is really well informed and it's something else to watch him speaking live on the podcasts and be so specific with numbers. He really knows his stuff.

It's so funny how people's behavior converge to the same patterns over time. A few years ago I was using Google searches with "reddit" appended when I wanted to see genuine people's opinions, such as reviews. Later it started getting written in mainstream how everybody was doing it. Now you are saying about Google as a quick LLM chat (because it's easily reachable from the browser's URL bar) and that's exactly how I've been using it too.

So probably the rule is, if I have discovered some convenient usage pattern, it's very likely that many other users have it discovered independently, too.

Also, it's interesting to see how Google's value has fallen to the point when it's used mostly for very low stakes use cases.

It's at least not worse than Microsoft. Go and start a chat with Copilot. Which one? I don't know, they are all the same. Or maybe they are all different? All I know that if you start a chat in one interface, there is close to a 0% chance (but not quite) that you will find the same chat in another interface.

That agents with virtually unlimited computing resources can find these bugs is not surprising and of little interest at this point.

But can they fix and harden the software without introducing new bugs in the meantime? I would argue that over decades we have placed the industry in such feature hell that there will always be new critical attack vectors.

Now that writing the walls of code and building features is cheap, it is time for us, humans, to learn to reduce complexity in the software and to learn which features are not worth implementing.

Definitely there are many problems to this solution, such as what to do with existing billionaires. If it was up to me, I would say to them "figure it out, the late payment interest will start to accumulate soon". But realistically it could be done over multiple years and in stages to limit the negative effects.

Also, it doesn't have to be tiny at all.

Why couldn't there a wealth ceiling? Let's take $1B per person, everything above that is taxed at 100%. Would it destroy the capitalist dream and people would lose motivation to grow their wealth, or would it stimulate cooperation and wealth distribution?

I don't know. With each new release of these models, they compare side-by-side how they are better than the previous ones, but at the same time I feel less impressed by each new release. Somehow they manage to be a seemingly big step forward but at the same time I feel I have seen so much AI that I just look at the common failure modes and they are all still there.

It's easy for them to show demos of abstract things because they don't necessarily have to follow real-world rules, but anything modeled after real world falls short of expectations.

Would you use hosted models if they promised anonymity and no log retention and servers were physically located inside jurisdictions that respect privacy (Switzerland, Norway or similar, I'm not into law) if that meant better price/performance ratio than your current rig?

What's the point to provide it with Apache 2.0? It's likely going to get sucked into the LLM training or inference process sooner rather than later. By that time license is no longer going to matter.

It will be $1/hour in a year. Then you will have to pay a different rate for better graphics. These things don't end well. Once they see that they have a stable customer base, they will figure out how to squeeze even more money from their customers. Of course, it will be advertised as something that the customers themselves required and they do it for them, not for themselves.

I didn't say it's a vibe-coding issue, though it could be. Your guess is as good as mine.

Not saying it's a vibe-coding issue. But I just imagine a scenario where a global service experiences significant disruption due to a vibe-coded release. No person understands how it works or is able to figure out in less than a week because no human has ever glanced at the code for a couple years, so the decision is taken to unleash LLM agents directly on production to fix the issue. Of course, one of these agents has to do some database migration but there is a breaking change, so they launch a sub-agent to deal with that; the sub-agent doesn't have a full context, so it decides that the easiest way would be to drop the database and so it does. The latest _successful_ backup took place a month ago.

$100B value down the drain.

Perhaps they are there for the 'Rocket' edition.

"Working" is a generous word for his position. The "work" involves flying with private jet to expensive places regular people have never stepped foot in for noncommittal meanings, showing up around the office at random times, have a word with random high-level managers in the office. All the while not having the slightest thought about someone evaluating your performance or output.

There are two ends of the rope pulling the opposite ways. On one end there are companies who are only now starting to onboard on AI and their spend grows. On the other end there are companies with certain maturity who are starting to calculate their spend, put on the limits of employees, explore alternative models with the aim to limit spend.

I think as the pool of new adopters shrinks and the pool of advanced adopters grows, the spending on LLMs will start to fall.

I hope more than anything Ukrainians will stay strong and persevere. Perhaps not even for themselves or their country, but just to witness the enemy collapse when it all ends. The mere fact that you are Ukrainian and live in Ukraine invokes a certain level of respect.

I agree and I have a good example.

At my work we are forced to use passkeys internally and it's fine with me.

At some point in the past, due to the constant nagging I set up a passkey on my GitHub account.

I needed to sign in to work's GitHub account, so unknowingly (since there's no SSO option there) I just clicked on "Sign in with a pass key". It didn't ask for my username, just gave me an option to scan QR code. And I did because work's pass key is set up on my personal phone. Then I noticed that it's trying to sign in with my personal GitHub account. It's good that I don't have anything sensitive on there but I really wouldn't want to sign in on my personal account on my work's laptop.

I would love more existing or recently ex- software engineers to share what new activities they have been taking up for living and that have been going well. I am certain that there are niches where a refined systematic thinking that many software engineers have could be a big advantage. However, it's expensive and risky to start new businesses in areas we don't have any experience in.

I don't think code quality has improved. I think the main advantage of LLMs is that they have unlimited perseverance to rummage in whatever code you give them to to solve a particular problem, and given enough loops they will also find and fix some of the bugs that have been either preexisting or introduced freshly. Code quality is important for us, humans, because a lot of times we don't have mental capacity, patience and speed to work in bad code bases for long time.