Anyone know if these are already powering all of Gemini services, some of them, or none yet? It's hard to tell if this will result in improvements in speed, lower costs, etc, or if those will be invisible, or have already happened.
I like this idea! I don't need the LLM bits, and want it to run on an old Android tablet I have lying around. Can anyone recommend similar software where I can get wikipedia / street maps / useful tutorial videos nicely packaged for offline use?
Kiwix has an Android app, that'll do Wikipedia and a bunch of other resources. You can get free offline maps from HERE maps or use something like Open Map from Fdroid that uses Open Street Map.
I'm very happy about this change. For long sessions with Claude it was always like a punch to the gut when a compaction came along. Codex/GPT-5.4 is better with compactions so I switched to that to avoid the pain of the model suddenly forgetting key aspects of the work and making the same dumb errors all over again. I'm excited to return to Claude as my daily driver!
I don't think we won't get AGI if Anthropic were to implode, and frankly, right now, I'd rather have someone say clearly, "They cannot stomach the existence of someone telling them 'No' or adhering to moral principles. Like spoiled children they can't hear the former and are terrified by later because it might expose them to the condemnation they deserve."
TLDR (story, not math) - Knuth poses a problem, his friend uses Claude to conduct 30 some explorations, with careful human guidance, and Claude eventually writes a Python program that can find a solution for all odd values. Knuth then writes a proof of the approach and is very pleased by Claude's contribution. Even values remain an open question (Claude couldn't make much progress on them)
I think this is pretty clearly an overstatement of what was done. As Knuth says,
"Filip told me that the explorations reported above, though ultimately successful, weren’t really smooth.
He had to do some restarts when Claude stopped on random errors; then some of the previous search results
were lost. After every two or three test programs were run, he had to remind Claude again and again that
it was supposed to document its progress carefully. "
That doesn't look like careful human guidance, especially not the kind that would actually guide the AI toward the solution at all, let alone implicitly give it the solution — that looks like a manager occasionally checking in to prod it to keep working.
looks like he is trying to make a point that the actual (formal) proof for 2Z + 1 (odd numbers) is still human - by himself that is. Not sure who came up with the core modular arithmetic idea of with s = 0 k increasing by 2 mod m.
Totally reasonable project for many reasons but fast tools for AI always makes me chuckle. Imagine your job is delivering packages and along the delivery route one of your coworkers is a literal glacier. It doesn't really matter how fast you walk, run, bike, or drive. If part of your delivery chain tops out at 30 meters per day you're going to have a slow delivery service. The ratio between the speed of code execution and AI "thinking" is worse than this analogy.
The crucial thing is that Tesla's valuation has the hype projects baked in. The fact that it never delivered self driving or a robotaxi fleet and is now being saved solely by an import ban on Chinese EVs means that any success he had with Tesla is now an illusion.
There is another way to view this. FSD plays fast and loose because they are constantly iterating. The culture at Musk co is that if you dont' keep pushing updates you are in trouble so do we really want to trust that each of his numerous updates are truly tested? This guy is a pathological liar after all. How many lawsuits are they dealing with now?
Supercruise only runs on pre mapped routes. If my life is on the line, I'd rather take the pre mapped routes and supercruise design is better at preventing people playing games to defeat the system (ex.shoving an orange in the steering wheel) so I know that others using the system on the road are following the system guidelines.
Supercruise may not do everything FSD does but it cuts out a large portion of the "fatigue" portion of driving and as a result can be highly trusted value add.
They rolled out full driverless in Austin in November 2025 and there's a website that reverse engineered the mobile app API to track the active cars. It found 90 active in Austin with more declared total by Tesla and 150 active in SF (SF ones have a safety driver for now). Likewise they found around 300 active in SF for Waymo with around 1000 cars declared total by Waymo itself.
While this seems to detect posture fairly well, the screen blurring doesn't work for me despite allowing what appear to be the relevant permissions. (macOS 15.1)
This seems to be the classic discussion over what counts as reliable. Humans aren't particularly reliable, and as any hardware engineer knows, even if you have provably correct algorithms your software system can never be 100% reliable because cosmic rays and spilled coffee. You can get close via herculean efforts in software and hardware co-design but never all the way. To try to pierce the hype of AI agents without allowing for the surprisingly low bar set by humans across a large array of tasks is to miss the forest for the trees.
reply