17 chunks · 15 episodes
... Quist: OpenClaw is What Apple Intelligence Should Have Been. Though OpenClaw's security model also demonstrates why Apple could never have done it. It's just too catastrophically reckless to be viable as a mass-market product. OpenClaw shows t...
...ideally scale with the size of the community, not the size of the company. Many security models require moderation of some form. The default approaches scale their capability with the number of employees of the company. But those can quickly ge...
.... Like, say, working with fissile materials, or money transmission, or software security models. There is significant downside if you get it wrong. That means you need to be trained, and it might even need to be regulated. But imagine someone c...
Networks used to have a permitered-based security model. Everything within the castle wall was trusted fully; everything outside it was fully untrusted. The most important thing was to verify that the cast...
...t was self-catalyzing. But Greasemonkey ultimately had a low ceiling due to its security model. The laws of physics of a system, its security model, sets both the speed of distribution and also the ceiling of max penetration. Imagine a Greasemo...
...y rethought in a local first world. When it's local first you can also cut some security model corners. If it's local and you have only a handful of collaborators, then it's OK to run arbitrary JS from any collaborator on the data. But that obv...
...when was the last time you thought about the same origin model, the fundamental security model that undergirds the web and apps? In practice what you're looking for is not "will everyone understand it and therefore accept it." You're looking in...
...e), you need a permission dialog: a border crossing. It's possible to imagine a security model that allows safe composition more granularly, where the boundaries between things could get fractally more nuanced, and overall feel less like a high...
...As Gordon Brander has put it, aggregators aren't open-ended. This is not just a security model limitation. It's also against the aggregator's interest. Turing-completeness makes an open-ended system; one that can generate its own requisite vari...
...ter looking through many, there are a number of recurring patterns I see: 1) No security model For composing untrusted components, they'll "figure it out later". But this is the core dynamic that has to be figured out, you can't retcon it onto ...
...pps' laws of physics? The inductive willingness of less-savvy people to trust a security model is about iterative trust in people who are more knowledgeable than them. An example inductive chain: "My tech-savvy niece says this is secure, and I ...
...t also to distribute it. Laws of physics here cover things like the privacy and security model. The laws of physics set how strong gravity is, and also the speed of light. Even if it's possible to create a privacy-preserving, hyper-bespoke appl...
...entally gullible. You cannot rely on them for the structure of any part of your security model. A GPT could configure a solid box to have certain privacy properties but it can't be the box. The structure is what makes it safe. The LLMs are what...
...rt folks building frameworks for 3P agents to cooperate. I asked them how their security model worked to allow untrusted 3Ps agents to participate. They told me "oh we'll figure that out later". To which I replied, "...no you won't!" The only w...
...stant, carting along your cookie jar of state. One of the key invariants in the security model is "tapping a link should never be dangerous" This means that instead of each site being an island that has to be viable on its own, experiences are ...
...e wouldn't have been discovered. The web did this with a deliberately curtailed security model. Critically, the web did this with a bare minimum of centralization--quite a feat! This means that the web is a proper platform with no ceiling. Appl...
Post-hoc tightening the security model of a widely used software system is extremely challenging. If you want to support as many existing "good" uses as possible, you'll have to design a c...