I think Z.ai rushed a bit for release, for example GLM 5.2 is only available under the coding plan right now and they didn't do a big write up. Not even some charts and graphs about its performance!
This is around when people were predicting a new GLM to come out, so a couple corners clipped in order to catch the moment. I'm using it right now and it seems decent, but I haven't done heavy work with it yet. The expanded context window is great.
Starting from a literal bandwidth costs perspective definitely won't get you there. I'd start by trying to feel personally annoyed by things like that. Then maybe try to feel more annoyed, since you know it'll touch every customer forever.
In that bandwidth case I'd be annoyed by the waste which kind of pervades software already, and it'd feel great to know at least we countered it a little bit.
That's well said and where I ended up as well. This whole thing did reveal that a lot of people really never enjoyed programming at all and only saw it as some irritating necessity. I don't really like that crowds presumption that everyone else disliked it too.
For me working through the programming part is the understanding and solving. Programming languages are pretty beautiful and encourage different ways of thinking. Hopefully we can understand it and contribute.
It's a fascinating angle they've taken to give Claude your payroll. I guess we've reached this part of the AI race and they're running ahead of people realizing what it can do.
LLMs get ridiculous with elixir, especially with the repl, runtime, and ability to hot reload / directly test functions. It's really surprising to me it hasn't caught on more but I guess you have to see it to believe it.
If there was an exponential cost, I would expect to see some sort of pricing based on that. I would also expect to see it taking exponentially longer to process a prompt. I don't believe LLMs work like that. The "scary quadratic" referenced in what you linked seems to be pointing out that cache reads increase as your conversation continues?
If I'm running a database keeping track of a conversation, and each time it writes the entire history of the conversation instead of appending a message, are we calling that O(N^2) now?
My mom said, "whatever we built isn't working anymore," and I think that captures most of the sentiment. It's also funny to see the "the economy is roaring!" "incomes are up!". Great, have they increased by as much as inflation? Can I afford a home?
Work has if anything gotten worse in general. Remote's gone. Pay's less. ADHD maximum AI use required. Nobody can take a break. Pressure's on. 1.5 trillion more to the military. What are we even building? For what?
Dunno, from the WSJ scoop: "CEO Sam Altman announced the changes to staff on Tuesday, writing that the company would wind down products that use its video models. In addition to the consumer app, OpenAI is also discontinuing a version of Sora for developers and won’t support video functionality inside ChatGPT, either."
I don't get it, now I can't use Tidewave or Zed or Agent Shell in Emacs with my Claude Code plan? Why would they do this? It makes my subscription so much less worthwhile.
More than an extension, imagine us having good enough, fast enough, vision models that you never even see a real website. Maybe the whole OS in Microsoft's case if they keep putting more ads in. It will be a level of inefficiency inconceivable but really something.
I'm a big fan of it for building micro websites with LLMs, since it can keep pretty much the entire thing in context (even including the docs) it seems to perform pretty well.
I'd say you identified a difference with modern names. This split has certainly grown wider in the "code is cheap" AI era and changed meaning.
I'm firmly in the camp of actually enjoying programming. To me it was interesting to hear that some people actually don't like it all, and it's much nicer to have something "just do it".
Over my career I've leant much more heavily into programming as the art.
I wouldn't even say "how do you balance" is too much of a problem, as we all can vary between needs, you know?
I've found really hammering it with *important*, all caps, "NEVER", etc finally made it start using the tidewave MCP for elixir development well. It felt really heavy handed but it worked.
This is around when people were predicting a new GLM to come out, so a couple corners clipped in order to catch the moment. I'm using it right now and it seems decent, but I haven't done heavy work with it yet. The expanded context window is great.