KronisLV

20 ms·

361.
by KronisLV 6mo ago
> As of May 2026, how much money do I need to spend to buy hardware to have a local model that is 80% as good as SOTA services for assisting me in writing code? https://llm-stats.com/benchmarks/swe-bench-verified SO
362.
by KronisLV 6mo ago
Wonder if there's a way to launch the desktop Claude app like that, especially on Windows, not just the Claude Code TUI/CLI. Might not be possible and you'd just have to use --remote as a workaround.
363.
by KronisLV 6mo ago
> And it's managed. Can’t use cloud stuff on-prem and also if your clients have a server room of their own. Same for homelab. Also it’s nice not to shift the pets attitude from servers to clusters and instead treat everything as cat
364.
by KronisLV 6mo ago
I’m still waiting for the Intel Arc B770 since a 5060 Ti and 9060 XT are already overpriced and if they just committed to something for once it wouldn’t be marginally worse. Not that releasing the GPU would be something super innovative, th
365.
by KronisLV 6mo ago
> They could have shipped a good product with all those billions they spent in reinventing Clippy. I really liked Copilot - it gave you a lot of tokens across a bunch of models and their agentic features were perfectly serviceable, along
366.
by KronisLV 6mo ago
Here’s their pricing docs, they’re running a discount for now https://api-docs.deepseek.com/quick_start/pricing/ The 150M assumption of mine is for 100 USD at the regular prices (though even that needs sufficient
367.
by KronisLV 6mo ago
I'm currently paying for Anthropic's Max subscription (the 100 USD one) and I quite often hit or approach the 5 hour limits, but usually get to around 60-80% of the weekly limits before they reset (Opus 4.7 with high thinking for
368.
by KronisLV 6mo ago
> Yeah, and ultimately no body cares. More like hiding their heads in the sand in circumstances that are outside of their ability to fix. None of the tooling or practices out there push you in the direction of not being at risk, or even
369.
by KronisLV 6mo ago
> Reproducing everything Anthropic has already built with Claude Code, Cowork, and all their connectors would be nontrivial, and they're just getting started. They're one org with presumably some specific direction. As the actu
370.
by KronisLV 6mo ago
> "It's still cheaper than a human" they'll say. The question is how much friction there will be for people to switch over to Gemini, GPT or maybe even DeepSeek or Mistral or whatever. Even if price hikes are inevitab
371.
by KronisLV 6mo ago
> In successful open source projects you eventually reach a point where you start getting more PRs than what you’re capable of processing. Given what I mentioned so far, it would make sense to stop accepting imperfect PRs in order to max
372.
by KronisLV 6mo ago
> If a PR was mostly written by an LLM, why should a project maintainer spend time reviewing and discussing that PR as opposed to firing up their own LLM to solve the same problem? That's a fair thing to ask, though it seems like pe
373.
by KronisLV 6mo ago
> Anthropic apparently won't take responsibility for issues their own systems handling billing cause. I think with every org, especially the big ones, trying to dodge responsibility (setting the intent of "customer support&quot
374.
by KronisLV 6mo ago
For it's size, that's really good! Though I bet it being a dense model probably helps a lot, if it was MoE at that size, I bet the benchmark performance would go quite a bit down (which consequently would also mean that I'd a
375.
by KronisLV 6mo ago
Is it? Evolution also seems to be a result of semi-random crap over the span of millenia and nobody is critiquing it like that. Why should throwing ideas at the wall in regards to optimizing code be any different: as long as you can measure
376.
by KronisLV 6mo ago
> What do you mean, are they still making GPUs? There was recent talk of them pulling back from the consumer segment, though obviously the leaks have also predicted Battlemage not being a thing so go figure: https://youtu.be&#
377.
by KronisLV 6mo ago
> There are a multitude of reasons that a system can go down. Start doing post mortems then! At the very least, them using any off the shelf service that's shitting the bed would inform others to stay away from it - like an IAM solu
378.
by KronisLV 6mo ago
It's a bit odd, since RooCode was genuinely great (those various agent modes, vs just plan/act). At the same time, something like KiloCode (which is a further fork) seems to be doing well at the project level! Then again, I almost
379.
by KronisLV 6mo ago
> The dashboard wasn’t that complex: View information, perform some API calls. Replicate what we were doing in the DB and API manually but in a web page. > People weren’t comfortable with frontend development. Indeed, some shared that
380.
by KronisLV 6mo ago
> You make your own subscription. I'm asking because with most providers (most egregiously, with Anthropic) it doesn't work that way because the API pricing is way higher than any subscription and seemingly product/company
381.
by KronisLV 6mo ago
> On a side note, any self hosted model I can get for my PC? I have 96 GB of RAM. Try the 8 bit quantized version (UD-Q8_K_X) of Qwen 3.6 35B A3B by Unsloth: https://huggingface.co/unsloth/Qwen3.6-35B-A3B-GGUF Some
382.
by KronisLV 6mo ago
> Deepseek v4 is good enough, really really good given the price it is offered at. Do they have monthly subscriptions, or are they restricted to paying just per token? It seems to be the latter for now: https://api-docs.deepse
383.
by KronisLV 6mo ago
Decent amount of features but definitely way more expensive than what you can get on Hetzner: https://stackit.com/en/prices/cloud
384.
by KronisLV 6mo ago
Same! I wonder what other alternatives there might be for autocomplete.
385.
by KronisLV 6mo ago
> I was curious why a company would still use the VS Code + Copilot sidebar method for coding, rather than something like Claude Code. I use Claude Code, but I kept my Copilot subscription around mostly for really cheap usage of other mo
386.
by KronisLV 6mo ago
> but even the smallest startup wants to be able to avoid data loss Seems true at a glance! > and downtime. Maybe less so - I think there’s plenty out there, where they’re not chasing nines and care more about building software instea
387.
by KronisLV 6mo ago
> Eventually we have converged to Jira and instead of doing a few things really well we now do everything poorly. Is a system that does everything within its scope well not conceivable? If it is, does systems ending up like Jira come as
388.
by KronisLV 6mo ago
I don’t think Godot is any worse than other engines inherently, other than it moving forwards pretty quickly and the latest versions not being in the training data. I wanted to evaluate which engines would be the best for working with LLMs
389.
by KronisLV 6mo ago
> Mid-price enthusiast players are under the risk of irrelevance from cheaper/better competitors. The higher-end of the market-- the Steelseries, Corsair, Razer, Das Keyboards-- are being perpetually undercut by the Redragons, Akkos
390.
by KronisLV 6mo ago
Debian is good! If you want something desktop oriented and Ubuntu based without the focus on snaps, take a look at Linux Mint: https://www.linuxmint.com/ (there's Cinnamon, Xfce and MATE versions; personally I think Ci

More ›