aenis

Their pace makes me wonder how effectively do frontier labs protect their weights against motivated, capable adversaries. By now the value of something like Mythos is well in double digit billions, and more so in strategic advantage.

Considering the hosting happens across many providers, and many geographies, as does training, I wonder how airtight the CC tech really is to prevent, I don't know, key extraction from the secure enclave of the GPUs. This requires physical access and cutting edge techniques, but this is also the absolute cutting edge of secrets-worth-stealing.

I'd not be surprised if the distillation attacks via API were a smokescreen to theft of actual weights. Long shot, but given the motivations, and the general weird state of US AI labs. I'd not surprise me.

The models are already brilliant at that. My own todo app generates 128x128 pixel art icons for my todo items. They are mind blowingly creative and funny.

Even boring, slow moving companies are well on their way to eradicate the last traces of oracle java. We kicked the last of that 2 years ago, despite having >3000 different systems across the globe. (We started circa 5 years ago).

I rebuilt a couple of games I used to play as a kid (jet set willy, mario, thrust, now i am working on Mercenary) - and for each I am also asking LLM to build an autopilot "AI" (which of course is really entirely deterministic). I am doing those things for fun while I am waiting for Claude to finish something I am actually working on. Not sure if it counts.

For sure. What it gave me was I think similar to the Amstrad version of Thrust and since I originally played it on a Commodore PLUS/4 I had to tweak it a bit. The fist version also had wrong gravity and rotation speed constants. But those were easy tweaks. Whenever I am bored or waiting for something more serious to complete I recreate games I used to play as a kid. I am currently working on Mercenary.

To be fair, designing an autopilot was an ordeal. Opus stuck to the principle of a mix of waypoint based routing plus a PD regulator - which made playing perfect levels hard. I wanted it to use the fact that the game mechanics is perfectly deterministic so theoretically it should be able to compute a perfect walkthrough before starting a level. It very reluctantly tried a monte carlo approach which ultimately kept failing. I have given up and went back to a smarter version of the waypoints + PD + softening the transition between waypoints. It is producing good, fast paced playthroughs but I've seen better speedruns by human players. Need to find time to see if Fable does a better job.

Funny. I had claude recreate Thrust based on wiki article alone - and it nailed it. I then spent a week trying to implement a perfect autopilot for it. And then a solveable level generator. I have this running as a screen saver in one of my apps.

I am not arguing with that. Its a fact that companies are staffed with salaried people incentivised with externalising costs and avoiding personal responsibility. People legislating stuff should know this well and design taking this reality into account. But they themselves are incentivised largely the same way, and some of them are also reaping material benefits from establishing regulatory regimes like this one. So here we are. For the 0.0001% of the population actually concerned about privacy billions of people have to click on the stupid banners 20 times a day. I hope hell exists, and that its where the people behind it will end up.

That ignores the main issue, though. Companies moved on, cookies are no longer needed to associate your behaviour on various websites. I'd argue that because cookies were regulated like this, we now have state of the art tech that profiles people across vpn's, not to mention incognito browsing. Just ask anyone working with big data aggregators or having access to telco data and other logins. Its not just browsing behavior, some companies can get your extremely intimate, offline-only facts by combining those data sources. But hey, a bunch of technically illiterate people in Brussels feel like they did us all a favor.

The populists are the only one raising this issue. So, yeah.

Opus 4.8 xhigh is my daily driver for everything. I'd say Fable's edge is visible when designing for a complex problem with no obvious, idiomatic solutions. It is good at greenfield designs, and good at pointing out the pros and cons of hard design choices. I now do designs with Fable, and do implementations off those designs with Opus. Pretty happy. For a while I was using Fable for everything, but burned through a lot of real money for not much value, I think for coding its slow and not at all better than Opus.

Ya, and it works. I now default to Fable-for-the-design, Opus-for-the-build mode, which is sort of cost effective, but I am hooked. Opus does not compare when it comes to architecture, designs and algorithms, esp. for gnarly problems without an obvious compromise-free solution.

Yes. Absolutely yes. And I think this is why people vote for things like brexit, or really-obviously-bad-politicians. This is the only way in which they can pay back the many things the reasonable, progressive governments "gave" them. Stupid beeps included. I swear to god, I hate right wing populists, but I'd at least be inclined to vote for a party that promised to fight back those stupid systems.

This is a bit more nuanced than that.

Under the GDPR provisions, strictly necessary cookies are exempt from the consent screen. But the way the exemptions read prompts all the lawyers to "well, maybe it's sort of necessary for you to have a session_id, but its so, so much safer to have the damn consent screen, so please...".

I worked in technical roles at large companies that had completely honest-to-god private, strictly necessary cookies - where compliance people forced the cookie consent screens, there was no space for a discusion. At my latest company, despite being a CTO, I could not convince our legal team that going without the consent screen as we are really only using "strictly necessary" cookies was fine. Its "mate, you may even be right, but really, nobody is skipping the consent screen, and we won't risk a compliance issue on your say-so.".

Sure, lots of companies - maybe even most - do shit things, and of course the real tagging of people and fingerprinting happens in sophisticated ways that actually defy GDPR, require no cookies whatsoever and so on - and thats another problem. Those moronic consent screens server exactly no real benefit for a while now.

Those are the same insane morons who came up with the cookie consent. Cottage industry of lawyers that push for those regulations and then collect lucrative retainers from companies wishing to not be fined. One of the reasons EU is so hopelessly behind on any innovation.

I once made the mistake of renting one of those cars, putting my backpack and some groceries behind me, and driving straight onto a freeway. This absolute sh*tbox was beeping so loud I was afraid for my hearing. I drove some 10 miles like this before I could pull over and move the things. I'd not set foot in one of those cars ever again (same goes for Lexus). I was dreaming of really bad things happening to people who thought it was a good idea to emit series of very loud beeps while the car is driving.

Impossible to measure, many other uncontrolled variables - esp. significant improvements to infrastructure in Europe, and regulations. Take NL, where a crash involving a pedestrian or a cyclist effectively forces the driver to prove their innocence. I can walk across a Dutch town blindfolded with the biggest risk to my wellbeing being cyclists (well, and the canals). I'd guess the impact of those intervention dwarfs the "i will beep at you until i make you deaf if you don't put your seatbelt over your grocery bag" innovations.

If I hate anything about the EU, its the morons writing regulations for cars. My car constantly distracts me with some beeps, sometimes loud enough to be dangerous. Its surely one of the reasons far right is on the rise -- with things like 'drivers party' in some European countries winning serious votes. I spend 1-2hrs in the car each day, and I hate what those regulations did to driving.

(Worst offenders: Japanese cars since they seem to take the regulations most seriously. Least annoying: generally BMW, Volvo, though they are both getting worse each year).

Some text editors replace the -- (two separate dashes) with a proper em-dash. Literate people - who understand why em dash exists - have been using it all the time. Thats, after all, how the models learned to use it.

I would be doing the same thing with opus or gpt-5.5 -- project in crunch phase. But I really did enjoy Fable. Not sure if it was worth the premium, probably, but it offered great design help.

Indeed when they are coming out with a model that is, say, gpt 5.5 or opus 4.8 grade. Gemini currently does not compete in the same class. I wonder what do they use internall at google. Surely not 3.5.

I assume what they saw during the first coming of Fable gave them pause. Lots of people I know were using it non stop to the point of serious sleep deprivation. I managed to burn 6k in 3 days on it, and was myself using it from the minute it landed till the minute it was shut off. (Yes yes, I know, I am just saying there were stupid people like me swarming the available capacity)

How can you spend more on a still early stage tech, well in the depths of R&D?

or, well, scrolling claude code. That, too!

From what I recall they were big in the extortion business back in the day, and these days I think they are into construction or waste disposal, maybe both.

Closed, proprietary code is way, way worse.

Good programmers are ashamed to push anything less than good (at least in their own opinion) to popular public repos. Some of those same pedantic programmers have no problem pushing crap in enterprise repos, and feel absolved because they are pushed to focus on deadlines, new features, and refactoring is very rarely planned for. I did and managed a lot of corporate software development in companies big and small, and did my fair bit of M&As and looked at codebases of successful companies. I dont ever recall feeling impressed. And I am regularly impressed by the aesthetic qualities of popular open source packages. I think commercial code is mostly shit, with the exception of regulated, serious industries (power, space, flight, etc.).

I am waiting till apple copies the "allocation" concept from high end car manufacturers. "Sure, buy the 25 iphones ans we will gladly put you on the waitlist."

Fully loaded costs for an engineer where I live are anout 15-25 a month, plus a maybe 2-3k in token consumption. Whether the laptop costs 9k or 6k makes no difference and apple knows this.

Nope.

1: most of them dont use their own products

2: someone else pays for their laptops

What absolute horrible formatting of the blog. Small font, 60pct of the screen area unused. Who designs this shit?

I am wondering - if in those circles, questions such as 'is NBA intentionally discriminating against asians - or is the fact that long distance running is dominated by, say, Ethiopians an example of discrimination' are ever discussed - or declared taboo and racist? I don't doubt that the assumption is just plain, demonstrably wrong - we all evolved under different types of environmental pressures - I am just wondering if the proponents of the all-the-races-are-same-on-average are ever discussing those obvious facts, and what answers do they come up with to explain the, say, unfair underrepresentation of Japanese in the NBA.