I don't necessarily think that these companies have much room for market cap growth but it is definitely interesting that right at this moment the value of local has gone way up due to Claude Code (plus Cowork and competitors). I suspect that will change in the next several months but I know people who are actively switching from Google Docs to Office because of these tools.
"It's just too easily available to take the lazy option and just let it do the thing"
This seems to me to be one of the key problems for AI usage in general. Students have this problem where it can be incredibly helpful in actually learning but late at night with the assignment due early tomorrow the temptation is just too strong to have it do the thing.
I think that this just emphasizes how much Ive needed Jobs as a constraint. The early stuff they did together was genuinely good and changed the industry for the better. But his later work after Jobs got sick seemed significantly worse. Definitely a relationship that needed both sides.
I know this guy is doing this on actual Mac hardware but curious if there is a point of view on the best older Mac emulator out there? Ideally I'd like to run this on a current Apple Silicon Mac. It is hard to understand what is the best approach (which I realize might be because this is somewhere between legally grey and not legal). I don't want a browser based option.
Is there a writeup anywhere on what this means for effective context? I think that many of us have found that even when the context window was 100k tokens the actual usable window was smaller than that. As you got closer to 100k performance degraded substantially. I'm assuming that is still true but what does the curve look like?
I live how people used to talk about air gapping AI for safety and now we are at the point where people are connecting up their personal machines to agents talking to each other. Can this thing even be stopped now?
Oh, I love paper maps. Snowbird is trying to get rid of the paper maps and it drives me nuts. They still print a few but mostly you can't get them on the mountain. Seems like a totally misguided environmental idea.
Since I've been skiing this has been how I've experienced all the terrain. His maps just are skiing to me. But, interestingly, with the rise of smartphones/gps apps like Slopes and the late lamented Fatmap have started to move the ski world towards 3d terrain maps and away from these artistic maps.
I have a side project I've been meaning to dust off that translated GPS coordinates to locations on Niehues maps. I got it working reasonably well but the distortions were significant enough that it needs a lot of control points to do the mapping.
I've been thinking a lot about the fact that so much of our software has become engagement driven. E.g. Duolingo isn't optimized for learning a language, Facebook isn't optimized for connecting with your friends and family.
I wonder if AI coding tools might get out of this for some cases at least. Make an app that is clearly derivative but actually is optimized to do the thing it actually is supposed to do.
Harder with network effect apps but might be possible for others.
This is the thing I've found amazing about people's complaints about Apple and AI.
Historically the strength of Apple was that they didn't ship things until they actually worked. Meaning that the technology was there and ready to make an experience that was truly excellent.
People have been complaining for years that Apple isn't shipping fast enough in this area. But if anything I think that they have been shipping (or trying to ship) too fast. There are a lot of scenarios that AI is actually great at but the ones that move the needle for Apple just aren't there yet in terms of quality.
The stuff that is at a scale that it matters to them are integrations that just magically do what you want with iMessage/calendars/photos/etc. There are potentially interesting scenarios there but the fact is that any time you touch my intimate personal (and work) data and do something meaningful I want it to work pretty much all the time. And current models aren't really there yet in my view. There are lots of scenarios that do work incredibly well right now (coding most obviously). But I don't think the Apple mainline ones do yet.
There were some weird but large benefits to manufacturers for SUVs over minivans (didn't count against fuel economy standards, based on less expensive platforms). Those are mostly gone but the scale and preferences that they generated have at least partly led to the SUV takeover.
There are unquestionably some cases where Lidar adds actual data that cameras can't see and is relevant to driving accuracy. So the real question is whether there are cases where Lidar actually hurts. I think that is possible but unlikely to be the case.