mobelkh

they're already knee deep with LEO/kuiper(?)

everyone is rejoicing the other part of the stack can be vibecoded.

a backend developer would tell you the backend is much more important to focus on, performance, security, data integrity etc. not losing or exposing customer data is a lot more important than what it looks like.

truth is, it's all still important, preferably no one is vibe coding any part of a serious product.

I think for me it's more the possibility of the ideas here being incorporated in other Opensource projects (vllm, llama.cpp, or in the architecture choices of the models themselves)

Deepseek manages to be ridiculously cheap and efficient with prompt caching, i spent 550M~ (1.9b input cache miss, 1.7b output) tokens and it cost me 5~ bucks, that would be ~150 bucks on Opus 5.5, if i ran that workload in parallel that would be $50 vs $1500, e.i. i wouldn't want to subagent heavily with anthropic models.

now imagine Kimi K4 or GLM 5.5 with those cost savings

comic sans is said to be pretty accessible for people with dyslexia

flattens the entire city? absolutely not, this is not a standard anyone should consider normal

cities do not get systematically flattened with bulldozers in urban warfare, you only completely wipe out cities when you don't want any the civilians to come back.

why can't you do the process when you vote?

well, it is true. there is a price ceiling for intelligence, there were news about how adoption of fable was lower than expected, only hitting 11% of enterprise spending for customers with anthropic, iirc 26% for Sol with openai customers.

in any case, I've been using open and closed models since sonnet 4, i remember when the best I could get was qwen 3 480b coder, you can definitely feel the gap closing going from that and GLM 4.5, to GLM 5.3, DeepSeek Flash V4.1, Kimi K3 etc, it's reached the point where i wish I had V4.1 at work, it's faster and bullshits me less when I use it in my personal projects. And I have unlimited access to fable 5.1

is it? it's still the same model, they can claim the quantization down to q4 still retains 98% of the performance therefore it's fine.

nothing on the fine print tells you what the weights are, you're just getting Fable 5, whatever that is

but there is a gap between benchmarks and user feel.

Opus 5 came out with better benchmark results than Fable, but it really did not feel better to use at all.

what does it do too much out if the box? afaik it only has the basic tools and two agents profiles, maybe lsp support is extra?

why can't I use the tokens i paid for anyway?

the switch, steam deck and competitors will likely look into ARM, the cheaper ones already use it, flagship and gaming phones, the latter is bigger than you think, normal mobile gaming is big in third world countries and china, and emulation continuous to get better, I can ToTK and The Witcher 3 (windows version) on my phone, and it's a few years old flagship

throw his garbage out, the time and effort taken to review that is magnitudes more than what it took to prompt it.

have him start with an overall design doc if his change is 15k, it's definitely worth a design doc.

and then have his contributions reviewed in pieces of 200-300 LoC PRs.

any other solution is trading stability and system knowledge, that's 15k LoC no one is truly familiar with, even if you do try to review it

i would argue, the NSA not being up to some bullshit is more unlikely, and thus has the burden of proof on it, to assume otherwise is to ignore history and basic incentives these orgs have.

oh yea, let me just uproot myself because immigration took too long and my company didn't want to wait.

I'll just cancel my rental agreement, move my furniture across continents, cancel my other subscriptions, oh wait that's got a few month notice period so I'll have to pay up for months I don't need anymore, let alone all the friendships and social connections I made that I never got to say a proper goodbye to.

This is extremely unappealing to skilled workers, and the loss of trust will extend past this current administration, if all it takes is one idiot in office to upend your life, you can't trust that country and its people anymore.

they want that jira <-> bitbucket integration

would be great as an opt-in though

were you running the MoE models? those perform better speed wise

> disproving this “AI isn’t making software better” claim

one company deciding a feature freeze is not disapproving anything, in fact, it means things have gotten so bad, apple who used to have higher standards for their releases, had to pull the brakes and fix their shit

can't imagine rewriting a runtime to be easier than a coding harness

I wouldn't say competition is "most of the time", any market that saturates just turns to shit, incumbents can kill off any upstart competition while squeezing customers dry, not to mention fields that should have a baseline (e.i. healthcare), because squeezing customers dry there literally kills people

the piece didn't really seem very targeted at Rust as much as it's targeted at projects claiming to be secure just because they're written in Rust

it is falling if you look elsewhere, deepseek made their 75% discount on their V4 models permanent, on one hand there's LLM improvements that make inference cheaper (e.i. MoE, hybrid attention), on the other hand we're getting more inference focused chips that break the nvidia monopoly.

i don't think a lot of people know this, but a cluster of GPUs can serve multiple clients without much of a drop in performance, e.i. worst case scenario you band together with 6-16 people to run a 2-3 H100 server to host deepseek V4 Flash or 4-6 to run Pro, and you're getting the same performance as if you ran it alone, this means a lot of companies can afford throwing 50-100k into their own LLM server cluster.

We're at a price point where if you push it further people will move, there's no real vendor lock in, your agent config, skills, MCP servers etc are all reusable with other models and harnesses, so unless you get all providers to collude on a price hike, you risk an exodus of customers