She meant impossible in that one doesn't earn a billion dollars through work alone. The only way to get there is to set up a structure that extracts a billion dollars from a market (usually by building a structure that's more efficient but also generates externalities that are not borne by the person getting the billion dollars).
pg's reading of it is so blunt and misrepresentative that I'm nervous about what kind of content he's consuming.
The replies here arguing we should publish it all are wild in the worst kind of first-order thinking way.
It’s a census: it just asks questions.
If you start publishing and weaponizing the data against people with various attributes, they’ll just lie or not answer. And then you are left with worse than nothing: bad data people try to act on.
As is often the case, the title is hyperbolic. The discovery applies to 20% of tumors, and "one of cancer's significant defenses" or "a key weakness of cancer" would be more accurate.
That said, I'll happily take "we discovered a key weakness in 20% of cancers," please and thank you.
This is ignoring the fact that the main reason retired phones are e-waste is proprietary firmware blobs and locked-down systems preventing users from maintaining their phone with security updates, and very limited support length from OEM's leads to VERY insecure devices after they drop out of support.
You should not be connecting these old devices to an internet accessible network.
Google notably does well here with 7 years of support, but others such as Sony are 4 years, and Xiaomi on non-flagship devices are similar, or Samsung on their lowest budget models...
I still am struggling to understand why they informed the government about something that is known to be an issue in every LLM. There is no LLM that cannot be jailbroken, so unless this means that we have reached the absolute maximum publicly accessible US made LLMs are allowed to operate at with GPT 5.5, this is not grounded in any sane regulation attempt.
Does anyone know what limits Fable 5 has overstepped in the eyes of the government? Parameter count? Certain benchmark results? Training computer?
Cause if it’s just the ability to assist with cyberattacks and being jailbreakable, there is no model previously released that isn’t equally guilty.
Remember that for GPT 5.5 and 5.4, OpenAI also restricted the cybersecurity focused use under designated models, otherwise rerouting to 5.3-codex like Fable did with Opus 4.8. And both OpenAI models can also be jailbroken all the same.
Basically, what was the reason to tell the government now and not with Opus 4.5 or GPT 5.4? sama has been doing the rounds with apocalyptic predictions…
To update 10th-gen Honda Civics, Honda ships updates on specially-formatted USB drives. They're essentially Android 4.2.2rc1-era recovery packages with some Honda-added version checks (which can be spoofed). The packages are signed with the publicly-known AOSP test key, so with physical access to the front USB port you can sign and flash your own package for arbitrary code execution on the headunit. This doesn't require root/su. I've run it end-to-end on my own 2021 Civic and separately confirmed an official EU update file carries the AOSP test-key signature. Tooling and writeup in the post.
I "enumerated" for the last census. Trust in my community was already not high* and I had lots of interesting encounters. I really believed the rather invasive data I was collecting with a friendly face would be used and handled responsibly. I feel for the poor souls that'll sign up to go door to door for 2030 now that the firewalls against weaponizing and monetizing all of our sensitive government data has been torn down, and even more for those that will volunteer information that can hurt them.
The comments that this rather expensive endeavour should just be about getting a head count are also amusing to me. The data collected was such an important baseline of common understanding, and this will not be a good thing for its future quality. I've grown very jaded now seeing all the things taken for granted in this country and lost or degraded recently with a whimper.
*: To be fair, they sent me specifically to places that didn't respond, so I was naturally led to believe that everyone in my region hated the government, ignored bizzarrely threatening fliers, or had recently moved and had no knowledge of the inhabitants (if any) during the census period.
> Researchers at Amazon had used a series of prompts to get Anthropic’s Fable 5 model to provide them with information that could be used to aid cyberattacks...
Are there going to be bans on things that could be used to aid in school shootings next?
I agree that some of the examples the author provided are instances of bad animation. But I don't agree with the premise of the article.
Computer graphics is all about exploiting features of the human visual system. We perceive things differently when they're moving vs. when they're standing still. It's very possible that a "wrong" frame in isolation is the best looking one in a real-time context. We can also pick apart screenshots but these don't capture everything about how the user perceives a display in real-world lighting conditions.
I would draw an analogy to film. A fast tracking shot might look bad on individual frames because of motion blur. A wide-angle shot might make some objects look "wrong" because of optical distortion. But these are still the right choice if they have the intended artistic effect in the theater.
Cancer is not one thing, it's a huge zoo of many many many ways that cells start to break the social contract and divide in an uncontrolled manner.
One of the most commonly observed broken mechanisms is mutation in the gene KRAS that turns this on/off growth switch into the permanently on position.
This has been known for decades, of course. And there have been huge amounts of effort to try to develop drugs that target KRAS in cancer, but for decades it's always been thought of as 'undruggable' because of the difficulty of finding any molecules that would affect it.
This new drug, that finally treats KRAS mutated cancers, goes about it in a new way. Instead of trying to gum up the works of a single protein by sticking a small chemical in it, it effectively "glues" the KRAS protein to another protein, CypA, which keeps the switch away from reaching the normal areas where it's "on switch" activity works.
So this new drug means two things: 1) a lot of the most difficult to treat cancers are now far more treatable, and in the next 1-5 years clinical trials will tell us which cancers this particular drug works well for, 2) there's an entire new class of drug activity that everybody is chasing at this very moment, so in 5-25 years we'll likely have a huge number more of these sorts of treatments.
“ GLM-5.2 is Fully Open, Frontier Intelligence Belongs to Everyone
Today, the sudden restriction of certain frontier models is deeply regrettable. At a time when access to frontier models is abruptly cut off for non-technical reasons, we are even more convinced of one thing: science should be global.
The path to AGI (Artificial General Intelligence) must never be enclosed by high walls. We have always believed that AGI should be the cornerstone for all of humanity to collaboratively explore the boundaries of intelligence and solve complex challenges, rather than a privilege monopolized by a few rules and subject to revocation at any moment. In the face of external blockades and restrictions, our attitude is one of radical openness. Frontier intelligence must remain open-source, accessible, and buildable, serving every dedicated developer.
GLM-5.2 is Zhipu's most capable open-source model to date. It not only supports a truly usable 1M context window but also maintains a continuous lead in the independent completion of long-horizon tasks, providing solid foundational support for building complex agent applications. It also continues to be our main engine for creating the strongest domestic coding model.
Tonight at 5:21—at this special moment—GLM-5.2 will officially be available to all GLM Coding Plan users (including Lite / Pro / Max). The API will also go live next week.
A step closer to frontier intelligence for everyone.
The future of AI is open, and it is for the people.
ModelKey: GLM-5.2”
As a New Yorker this doesn’t shock me too much. The level of “Mamdani is an anti-Semite” sentiment I saw online (Reddit particularly) felt truly hysterical. And wasn’t matched by any equivalent in the offline world.
The headline buried the lede -- this is a way to get some summer vacation (niiice) AND encourage enterprise support contracts, which will still have availability. I don't think I've heard of this particular open source / support / summer vacation business model before but I like it!
We started the company two and a half years ago, and raised $7.3m in 2024 (announced only almost a year later). We've spent less than half of this amount.
Earlier this week we came to the difficult decision to wind down the project. The open-source repository remains available on GitHub (Apache 2.0) but won't be actively maintained by the team moving forward.
Seems like there's no official blog post with benchmark results yet. But I'm once again thankful for the Chinese AI labs for being open with their work and contributing it to the world under permissive licenses like this. The Fable 5 fiasco is just another reminder of how valuable these things are to have.
Adobe has always been like this, too. They squandered an enormous marketshare with Flash because the alternative would've been spending a couple million on QA and they managed to unite all of the browser manufacturers in agreement that the web was better off without such an unreliable partner.
I shipped a couple of things on Flash back in the day but it was staggeringly bad software — random crashes, various heisenbugs where changes in one area would affect unrelated functionality in other modules, etc. — and while it cost something like $800, it was completely unsupported: I filed a number of trivially reproducible bugs with reduced test cases but never heard anything back until the next release came out and they sent automated suggestions that the bug might be fixed so I should buy a full-price license and find out.
The exponential growth model that pg presents is an obvious lie - you could use his exact explanation to claim that good founders will soon become trillionaires, all it takes is 9 months of 93% growth after your first few billions; quadrillionaires will soon follow. The obvious reality is that exponential growth is just a small phase in any company's history, and it is anyway limited by the size of the market(s) it is operating in. A company can quickly capture a market, but not quickly grow forever.
Then, he completely hand waves away the externalities, lawlessness, self dealing and similar issues that are the supporting power of all such growth events. He claims this is all powered by "making your users happy", which makes them tell others about your business, as if every restaurant that people fawn over becomes a billion dollar business in 1-5 years. Even the examples he can think of of companies that did this are infamously bad companies that have become rightly hated and accused of spreading various forms of social ills - e.g. social media addiction induced by Meta, spy ads by Google, excessive tourism and house price increases by Airbnb, or the erosion of worker protections by Uber. There isn't a single billion dollar company that doesn't at least get accused of similar problems, which represent t the meat of the argument, and pg ignores this completely in favor of the "build something useful" model.
Only by a weird definition of "scalable". The first sentence says:
> Counterintuitively, large DELETEs add work to the database.
There is nothing counterintuitive about this. It takes just as much work to delete a row as it takes to insert a row. Why wouldn't it? Obviously you have to do almost all the same operations: write a log, write the deletion, update indices, replicate it, etc.
And yes, it's a well-known trick for all major relational databases (not just Postgres) that if you want to delete 90% of rows from a large table, it's much faster to just copy the rows you want to keep to a new table, run DROP TABLE on the old table, and rename the new table to the old table. Since DROP TABLE is ~instantaneous, mainly involving table-level metadata.
DELETE scales just fine, in the sense that if you are constantly inserting and deleting individual rows, DELETE scales the same as INSERT.
Basic database functionality is designed around the assumption of lots of small transactions. Whenever you have to do something involving millions of rows at once, you generally need to investigate solutions that work well in "bulk". E.g. loading rows directly from a file rather than with SQL, adding indices only after the data has been loaded rather than before, disabling foreign key checks on large operations (if you know by design that the keys are valid)... and yes, taking advantage of DROP TABLE instead of DELETE. This doesn't mean small transactions aren't scalable, it just means bulk operations are qualitatively different and benefit from their own solutions. And DELETE is no different from INSERT in this regard.
> It's tough to answer because you want to hedge for both an AI enthused employer and an AI hesitant employer with limited information about who they are and how they personally use these products.
Have you considered just answering truthfully?
Would you even want to work somewhere where you need to play a role and where they flip out when you say the wrong word you should've correctly guessed through mind reading?
That sounds not like a job but a toxic relationship.
I continue to work on my city builder game Microlandia, launched here in HN ~6 months ago. I originally predicted a few dozen urbanism nerds would play it, but now almost 10,000 copies sold. I'm still a solo developer but now I collaborate with 2D, 3D and music artists. Which is good because the original art that I drew myself for the launch was horrible.
I'm currently working on modeling energy, climate and new policies like universal basic income
People here seem a little confused. This is a simulator in the same way Goat Simulator is a simulator. It’s from a collection called “screen toys” and it’s meant to be mindless fun.
That's because you're treating the problem as an engineer instead of an "influencer" or "10xer" or whatever. You're treating it as a problem to be solved with engineering and AI is merely a tool to do so. It is, in my experience, vanishingly rare for an engineer to have a problem that needs to be solved with multiple hours of unattended AI code generation.
I've only found one single application where it makes even the slightest amount of sense to have an AI grind away for hours on end. I'm reverse engineering a widget which contains five separate firmware images. I've dumped the binary from the widget and I set the AI to decompile and reverse engineer these interrelated firmware projects. It's a compelx task, but very well bounded. It's not complicated work, but it's a lot of work, and the end result is a C-shaped pile of text that is only informative, it never would be compilable on its own even if I did it by hand. The quality of the output is tightly bounded by the input assembly and the overall output artifact is documentation in the shape of code.
I don't have any qualms about letting an AI go ham on it unattended because the stakes are zero. But if the AI can beat the assembly into a recognizable C project, it's much easier for me to read and reason about. Easy win, I think.
Maybe this will be simpler for Anthropic to understand if they take their own high-minded philosophical nonsense and ego out of it and consider it the way a neutral party would.
Suppose a company calls themselves The Doomsday Device Company. They make and sell excellent-quality doomsday devices. They regularly go online to proclaim that their doomsday devices are the best and most powerful, and also that doomsday devices are dangerous and should be regulated.
The Doomsday Device Company then says they have the world's best doomsday device. (They don't, but they claim they do.)
The US Government hates the Doosmday Device Company for various political reasons, but also has a vested interest in there not being a massive proliferation of doomsday devices.
The Doosmday Device company spends a great deal of time and money telling everyone: "Our doomsday device is the most doomy of all time!" (though it probably isn't) and "Everyone can use it!" (for a lot of money)
It is completely logical, then, for the US Government to say: No, everyone cannot use your doomsday device, because doomsday is bad. (While also meaning: Only we should be able to use it, and you shouldn't be able to tell us how.)
If you do not want to be in the business of having your doomsday devices shut down by the government, well, it would help if you didn't so loudly and aggressively proclaim how doomy they are. It doesn't matter how trustworthy you claim to be, given that your business is making evil doosmday devices. You still won't be trusted!
