Hacker Newsnew | past | comments | ask | show | jobs | submit | doctoboggan's commentslogin

What agentic harness do you use deepseek with?


AI generated header image and a heavy scent of LLM prose, but this guy still complains about the "insane climate costs" of google's 4GB on device LLM?

I used this for a bit a few years ago but eventually needed something that was hard or impossible in k3sup and just went to using the k3s tools directly. My deployment script actually got simpler after removing k3sup.

Also, fun fact, k3sup is pronounced "ketchup" according to the README[0]

[0]: https://github.com/alexellis/k3sup/blob/master/README.md


I was reading the description trying to figure out what it actually does. I built remote k3s deployment over ssh into a product I worked on, and there really was very little to it. Shell into the machine, run the installer, set the config - and that last part is going to be unique to your situation anyway. It makes perfect sense that your setup got simpler after removing this.

The pronunciation ketchup is somewhat unfortunate as a popular backup operator for k8s, k8up, also claims this.

Yes, field trips were always my favorite part of school. The "How its Made" show scratches a similar itch.

I've noodled with the idea of starting a "fieldtrips for grownups" group but I feel like a wastewater treatment plant is more likely to open their doors for a group of third graders than a group of thirty somethings.


That’s probably true but I wouldn’t count it out. I think you’re more likely to get answers like “we do 10:00 on tuesdays” (timed for schools) than “no”

That is a great idea. Too often, we just grow up getting used to the idea that things are just made somewhere in a "black box" and never take the time to investigate. We could probably get into more places than we realize.

The Oakland wastewater treatment plant offers tours open to the general public: https://www.ebmud.com/wastewater/collection-treatment/wastew...

Wouldn't be surprised if others do as well, or would be willing to if you asked the right person nicely.


> ... more likely to open their doors for a group of ...

Probably true. Usually. Just keep your eyes on local news, and wait 'till the Sewer Dept. is facing budget cuts, or needs a rate increase to pay for long-delayed repairs, or is trying to get a millage passed.


Using Claude code with Opus 4.7 and xhigh effort for a few hours will definitely cost hundreds of usd.

I am not sure if you would call claude code "an auto loop", but you don't need to be running something crazy like gas town to spend a lot of tokens with Claude.


Yes the blender API feels like it sits on top of the GUI rather than the GUI on top of the API. When you are writing scripts in the blender api you basically mechanically describe the steps you would take in the UI. It can be a little fragile at times.

I've used Claude to write some blender scripts and it's an excellent use case. I look forward to even better claude/blender interaction based on this annonuncement.


I've also used genAI to write script. It works splendid up to a point, then there is absolutely no way to move the needle further. And it's not even close to renders I would ever publish.

That being said, it's about the same for the code it produces for non purely creative things, but for artistic work, I doubt an LLM in between gives any gain. After all, we do have an interface. A human interface.


Yeah I am using blender to generate models for 3d printing - no rendering and Claude doesn't have to do anything artistic for my use case.

Its an LLM comment, don't search too deeply for logical consistency

Is it honestly better than Opus 4.6 or just benchmaxxed? Have you done any coding with an agent harness using it?

If its coding abilities are better than Claude Code with Opus 4.6 then I will definitely be switching to this model.


Apparently glm5.1 and qwen coder latest is as good as opus 4.6 on benchmarks. So I tried both seriously for a week (glm Pro using CC) and qwen using qwen companion. Thought I could save $80 a month. Unfortunately after 2 days I had switched back to Max. The speed (slower on both although qwen is much faster) and errors (stupid layout mistakes, inserting 2 footers then refusing to remove one, not seeing obvious problems in screenshots & major f-ups of functionality), not being able to view URLs properly, etc. I'll give deepseek a go but I suspect it will be similar. The model is only half the story. Also been testing gpt5.4 with codex and it is very almost as good as CC... better on long running tasks running in background. Not keen on ChatGPT codex 'personality' so will stick to CC for the most part.


Their Chinese announcement says that, based on internal employee testing, it is not as good as Opus 4.6 Thinking, but is slightly better than Opus 4.6 without Thinking enabled.


I appreciate this, makes me trust it more than benchmarks.


In case people wonder where the announcement is (you can easily translate it via browser if you don't read Chinese): https://mp.weixin.qq.com/s/8bxXqS2R8Fx5-1TLDBiEDg

It's still a "preview" version atm.


That's super interesting, isn't Deepseek in China banned from using Anthropic models? Yet here they're comparing it in terms of internal employee testing.


They use VPN to access. Even Google Deepmind uses Anthropic. There was a fight within Google as to why only DeepMind is allowed to Claude while rest of the Google can't.


> That's super interesting, isn't Deepseek in China banned from using Anthropic models? Yet here they're comparing it in terms of internal employee testing.

I don't see why Deepseek would care to respect Anthropic's ToS, even if just to pretend. It's not like Anthropic could file and win a lawsuit in China, nor would the US likely ban Deepseek. And even if the US gov would've considered it, Anthropic is on their shitlist.


Who uses Opus without thinking though...?


I know the rumors were swirling for the past few months, but just 4 more months of Cook seems like pretty short notice, no?


Consider that is mostly public headway. Behind the scenes the handover, mentorship, alignment I am sure was already happening for a while. E.g. you probably don't want the incoming CEO to have to immediately clean house or people might end up doubting their decisions, getting anxious or similar. The previous CEO can start retiring, moving people around to clear out possibly problematic leaders, break up internal "gangs" and ways of work - people will be more willing to accept their decision as they've been at the head for a while and have the trust. The new CEO comes in, group dynamics and rules are still fresh and building up between everyone, they don't have a black mark for firing anyone - to me it just feels like it would be a healthier and more mature transition.

To support this I was thinking about (and obviously Googling these names because I definitely don't know them by heart, only that they recently left) the change of CFO Luca Maestri to Kevan Parekh, John Giannandrea being removed, Alan Dye leaving and being replaced with Steve Lemay.

So I take those 4 months more as like an FYI to the public than anything else. Though I am definitely not someone that knows corporate politics all that well (or at all), just mostly thinking out loud in response to your comment.


Yup. The rumor mill was talking about a CEO change for a while, and around the time you saw the rumors building you saw the departures you mentioned. Ternus was being mentioned as the likely successor at least back to November last year. So internally the shifts have already been happening for some time, only observable on the outside via the high profile departures.


In 2021 the average time from announcement to new CEO for S&P 500 companies was 3.5 months, so this seems reasonably normal.

[0] https://www.spencerstuart.com/research-and-insight/2021-ceo-...


Given how quickly Cook had to step in for jobs, first in the interim role, four months seems like plenty of time (particularly given he's still executive chairman).


4 months in his _current_ role, but he’s not going anywhere—he’s remaining on as Chairman, which is still very much involved day-to-day.


Not really. Anyone and everyone is replaceable. Even Steve Jobs.


The title is missing the fact that this was solved by an LLM (gpt5.4)

See here for more comments: https://www.reddit.com/r/math/comments/1smehbo/stunning_ai_b...


Guidelines | FAQ | Lists | API | Security | Legal | Apply to YC | Contact

Search: