I looked at the trails (lists of passages) and wondered whether it would be cheaper to build these lists by clustering passage embeddings, rather than having an LLM construct them.
It's liquid propellant being vented, the fuel is under extreme pressure so when its released it immediately expands to a gas. I don't know that Honda has said what their propellant is, but it's probably liquid hydrogen and liquid oxygen.
A combination that works nicely to solve bugs is: 1) have Gemini analyze the code and the problem, 2) ask it to create a prompt for Claude to fix the problem, 3) give Claude the markdown prompt and the code, 4) give Gemini the output from Claude to review, 5) repeat if necessary
I like Gemini for "architect" roles, it has very good code recall (almost no hallucinations, or none lately), so it can successfully review code edits by Claude. I also find it useful to ground it with Google Search.
Gemini's context is very long, so I can feed it full files. I do the same with Claude, but I may need to start from scratch various times, so Gemini serves as memory (and is also good that Gemini has almost no hallucinations, so it's great as a code reviewer for Claude's edits).
Yeah right now I just dump entire files at it too. After 100-200k tokens I just restart. Not because the LLM is hallucinating (which it is not) but I like the feeling of treating each restart as a new feature checkpoint.
Many apps won't work in the Android emulator. Banking apps, etc. will detect it and ban it. Far more apps will ban the emulator than will ban GrapheneOS. We aren't aware of an app which would work in the Android emulator with a standard emulator image containing Google Play but wouldn't work with GrapheneOS. Nearly all Android apps work on GrapheneOS. It's nearly entirely just the subset which ban using any non-stock OS with the Play Integrity API which can't be used. We convinced a few banking apps to start allowing GrapheneOS via hardware attestation but the pace of banking apps integrating the Play Integrity API is unfortunately quite a bit faster than the pace we're convincing apps to support it after they do that.
You can build it for the emulator. It's straightforward to do, but it requires a lot of disk space and it will take around 40-60 minutes on a high end desktop CPU like a Ryzen 9950X. We don't publish official releases for the emulator at the moment because it's not intended for production use and isn't really a good experience to use. We could start doing it, but it'd add some extra work and we'd be concerned about people misinterpreting what it's meant to provide. Emulator builds don't have the regular security model intact or OS updates, etc.
Please read https://grapheneos.org/articles/attestation-compatibility-gu.... It's possible to support GrapheneOS by using the Android hardware attestation API either as an alternative to the Play Integrity API or instead of it. By using the hardware attestation API, you can make a list of allowed key fingerprints for the SelfSigned boot state for non-Google-certified operating systems. We list all our current keys for non-end-of-life devices on that page. Recently, Swissquote used this approach to add support for GrapheneOS to their Yuh app and may be adding it to their main Swissquote app soon.
You can test hardware attestation on any modern Android device but you'd need GrapheneOS on a real device to fully check that you have the SelfSigned fingerprint allowlist working properly. It wouldn't be hard to do it without testing it though, and our users can test if app developers ask our community on https://discuss.grapheneos.org/.
What problem do you think Play Integrity solve other than keeping the user's under Google's walled garden? Play Integrity is a fake marketing term for DRM fromGoogle . It does not guarantee security of the device in any way. My 6~ year old and unpatched Android 10 passes Play Integrity and can run banking apps. That explains everything about Play Integrity. I don't use apps from developers who think they know better than their users.
>What problem do you think Play Integrity solve other than keeping the user's under Google's walled garden?
It ensures requests to your backend are vaguely from actual devices, rather than a bunch of emulators. There's many reasons why developers might want this. It significantly raises the bar for credential stuffing attacks, for instance.
Yes, developers are of course going to opt for APIs that allow them to be as lazy as possible, forego proper backend security, and reduce costs. But is allowing developers to be lazy worth the cost of destroying user freedom? If so, maybe we should work on bringing back the Web Environment Integrity API. That would really help out web developers and would make web apps a lot more secure. Nobody uses Linux on the desktop anyways besides weird nerds so it wouldn't have any real impact.
Can't wait until we finally kill hacker culture for good. Everyone and everything will be fully secured. It's going to be beautiful. The nerds can cry about it all day long, but they're powerless to stop it.
Please don't add play integrity to your app. There are many of us using custom ROMs, and it can relatively easily be worked around, but very much is often a giant screw you to technical users...
I've noticed 4o uses a lot of emojis, and, in general, is very enthusiastic. I find it funny. If I want a more formal bot, I switch to one of the o3 family.
I use a very simple custom system prompt (not on my work machine at the moment, but essentially something along the lines of "for technical questions, please be concise and to the point, and when asked for code, omit explanations and emit just the code itself unless I ask for explanations"), and it does wonders.
It’s interesting that my default prompt is exactly the opposite one: “do not write the code unless I ask for it specifically”. I like to use LLMs as a discussion partner, but writing code is trivial after a good discussion and I can do that myself
I guess it depends on use-cases. I use ChatGPT a lot for "trivial" questions a la "how do I uncommit a specific file in my last git commit" or "how do I paste from one PIL.Image into another one". In the past I would have to search google, click on the StackOverflow link, and then parse that whole page. Asking ChatGPT to give me just the snippet is faster, so doesn't get me out of my flow as much.
