I’ve been thinking about this and wonder if it makes more sense to have Claude make native mobile stuff for each platform because you’re vibe coding anyways, there’s static types and probably more training data?

Opus 4.5?

When you say the benefits don’t outweigh the costs, do you mostly mean $ to research, engineer, train, and serve the LLM, societal costs, or all of the above?

TY for the recommendation, will check it out.

I haven’t seen any indication that the SWE role will disappear. Like in my experience, even with a significant productivity boost, it still takes an expert to orchestrate things make sure the code isn’t slop, handle things the AI can’t.

I do worry about the job market for juniors, and if there will be SWE layoffs if companies think they can do more with less, whether it’s true or not, or if we will just say that we can get more productivity out of the same amount of software engineers.

Anecdotally, Claude Code has greatly accelerated my productivity. I do have to supervise/catch mistakes/offer creative direction when I know how to solve a problem in a way it didn’t think of. New Opus 4.5 is great. Idk if it’s true, but I hear that Anthropic is on a path to near term profitability.

I’m old enough to remember when the knives were out for Hegseth

Easy.

Just suffer through living in the South for an extended period of time, then you’ll have a ya’ll card

Hope is the thing with feathers

I am under the impression that GPUs like the H100 have are architected to excel at deep learning workloads at the expense of others, and aren’t particularly valuable for gaming compared to cheaper GPUs, so at least that market is probably out.

Big if true

I personally can’t wait to get my hands on those cheap surplus gpus

I also want to know the answer outside of that

Statute of limitations has expired so as good as?

Or maybe things are just impressive when you first use them and then you get used to things and see the errors more often. But I feel like Claude has noticeable significant variations in quality over time with the same types of tasks on the same alleged model version.