Harvester is just Kubevirt with some UI atop it, the same as Redhat Virt. Works fine if you’re hosting datacenters or whatever, haven’t seen it be suitable in smaller manufacturing environment
Sysadmins needs logs that tell them what action they can do fix it. Developers need logs that tell them what a system is doing.
Generally a sysadmin needs to know "is there an action I can do to correct this" where as a dev has the power to alter source code and thus needs to know what and where the system is doing things.
I mean that is also the job of existing call handlers.
"We are experiencing an greater than usual call volume, please wait while an agent becomes available" only to be randomly disconnected has been a thing for most of my life.
Everyone seems to be hyping open claw at the moment soon its just going to be LLMs talking to LLMs.... I wonder if they will develop a short hand and start talking in wingdings.
Qwen3-Coder-30B-A3B-Instruct is good I think for in line IDE integration or operating on small functions or library code but I dont think you will get too far with one shot feature implementation that people are currently doing with Claude or whatever.
I could be doing something wrong, but I have not had any success with one shot feature implementations for any of the current models. There are always weird quirks, undesired behaviors, bad practices, or just egregiously broken implementations. A week or so ago, I had instructed Claude to do something at compile-time and it instead burned a phenomenal amount of tokens before yeeting the most absurd, and convoluted, runtime implementation—- that didn’t even work. At work I use it (or Codex) for specific tasks, delegating specific steps of the feature implementation.
The more I use the cloud based frontier models, the more virtue I find in using local, open source/weights, models because they tend to create much simpler code. They require more direct interaction from me, but the end result tends to be less buggy, easier to refactor/clean up, and more precisely what I wanted. I am personally excited to try this new model out here shortly on my 5090. If read the article correctly, it sounds like even the quantized versions have a “million”[1] token context window.
And to note, I’m sure I could use the same interaction loop for Claude or GPT, but the local models are free (minus the power) to run.
[1] I’m a dubious it won’t shite itself at even 50% of that. But even 250k would be amazing for a local model when I “only” have 32GB of VRAM.
I have been adding a one shot feature to a codebase with ChatGPT 5.3 Codex in Cursor and it worked out of the box but then I realised everything it had done was super weird and it didn't work under a load of edge cases. I've tried being super clear about how to fix it but the model is lost. This was not a complex feature at all so hopefully I'm employed for a few more years yet.
Presumably not, but the better approach is anyways to first plan using a powerful/expensive model like Opus, then you can use something less capable and cheaper for the coding part. This would be true even if you just want to use Anthropic models, but makes even more sense if you want to use something cheaper like Qwen3.5 or Kimi K2.5 for the coding part.
It takes your query, computes the complexity of the request, and tries to route it to the appropriate model. There is a /manual command i think, to pick the right model.
They mask the 429s well in Gemini-Cli - if an endpoint is rate limited, they try another, or route to another model, etc to keep service availability up.
Your experience on the 429s is consistent with mine - the 429s is the first thing they need to fix. Fix that and they have a solid model at a good price point.
I use my own coding agent (https://github.com/hsaliak/std_slop) and not being able to bring my (now cancelled) AI account with Google to it is a bummer.
I'd still use it with the Code Assist Standard license if the google cloud API subscription allows for it but I have no clarification.
> It takes your query, computes the complexity of the request, and tries to route it to the appropriate model. There is a /manual command i think, to pick the right model.
That is what is should do, but there is no > 2.5 model shown in /model and it always picks a 2.5 model. Ive enabled preview models in the google cloud project as well.
If I pass the 3 model in start param it shows 3 in the lower right corner but it is still using 2.5.
I know google has issues dealing with paying customers but the current state is a shit show. If you go to the gemini-cli repo its a deluge of issues and ai slop. It seems there is a cadre of people jumping to be the first person to pump an issue into claude and get some sort of PR clout.
It might be good but it needs more time to cook, or they need to take a step back and evaluate what they should consider a paid product.
Plus if you look at the commit cadence there is a lot of commits like 5-10 minutes a part in places that add new functionality (which I realize doesn't mean they were "written" in that time)
I find people do argue a lot about "if it is reviewed it is the same" which might be easy when you start but I think the allure of just glancing going "it makes sense" and hammering on is super high and hard to resist.
We are still early into the use of these tools so perhaps best practices will need to be adjusted with these tools in mind. At the moment it seems to be a bit of a crap shoot to me.
As bad as cloudflare is there is a reason people use it.
If you try and run a site that has content that LLMs want or expensive calls that require a lot of compute and can exhaust resources if they are over used the attack is relentless. It can be a full time job trying to stop people who are dedicated to scrapping the shit out of your site.
Even CF doesnt even really stop it any more. The agent run browsers seem to bypass it with relative ease.
One of the things that a lot of LLM scrapers are fetching are git repositories. They could just use git clone to fetch everything at once. But instead, they fetch them commit by commit. That's about as static as you can get, and it is absolutely NOT a non-issue.
No... Basically all git servers have to generate the file contents, diffs etc. on-demand because they don't store static pages for every single possible combination of view parameters. Git repositories also typically don't store full copies of all versions of a file that have ever existed either; they're incremental. You could pre-render everything statically, but that could take up gigabytes or more for any repo of non-trivial size.
I did not imply that it does, I meant to have a budget allocated for 'unauthenticated deep history queries', when it's over it's over and you only handle dynamic fetching for authorized users until cooldown.
Is it pretty? No, but it also is a pretty niche thing overall (git repo storage).
Granted, but there are open source alternatives that don’t have the same obsession with meaningless digital signatures. Turnstile is just a terrible product.
They are not super helpful fixing it either.
reply