dajonker

For me personally, there is no point in using more reasoning than medium or sometimes high. It just gets so much slower and expensive, meaning that i have to wait longer to evaluate the results and provide feedback or continue to the next task.

But does it generate good pelicans?

It's not OpenAI, that's what the memory industry wants you to think.

Some of these features can be turned off. Others can’t. Or if they can, it means also turning off useful long-standing features like automatic thread categorization.

This, I absolutely hate it. And like the author said, it must be intentional, so that someone at Google can show the usage numbers and get a promotion.

I recently installed an app to manually activate the fans on my MacBook Pro M1 Pro as I've never been able to trigger them over the past 4+ years. Just to check whether the fans even work (they do).

I can find it on rubygems but the link there to GitHub returns a 404.

Plug a wifi module in the M.2 slot!

that's pretty cool, thanks!

Mike Shinoda is fine with not being classified as Nu Metal https://blabbermouth.net/news/linkin-parks-mike-shinoda-says...

My father is a journalist and learned to type by himself on typewriters using his index fingers. He writes a lot and he is really fast. I don't have any issues with that.

You hit the nail on the head here. Saying and doing are two very different things. It's also especially tempting to find an excuse to use some shiny new thing that everyone is always talking about. Both for personal learning and curiosity but also for future job prospects. The reality is that it's easier to get a k8s job if you have k8s experience.

I totally agree, but that's not what happens in reality: the average devops knows k8s and will slap it onto anything they see (if only so they can put in on their resume). The average manager hears about k8s, gets convinced they need and hires beforementioned devops to build it.

Making Kubernetes good is inherently impossible, a project in putting (admittedly high quality) lipstick on a pig.

So well put, my good sir, this describes exactly my feelings with k8s. It always starts off all good with just managing a couple of containers to run your web app. Then before you know it, the devops folks have decided that they need to put a gazillion other services and an entire software-defined networking layer on top of it.

After spending a lot of time "optimizing" or "hardening" the cluster, cloud spend has doubled or tripled. Incidents have also doubled or tripled, as has downtime. Debugging effort has doubled or tripled as well.

I ended up saying goodbye to those devops folks, nuking the cluster, booted up a single VM with debian, enabled the firewall and used Kamal to deploy the app with docker. Despite having only a single VM rather than a cluster, things have never been more stable and reliable from an infrastructure point of view. Costs have plummeted as well, it's so much cheaper to run. It's also so much easier and more fun to debug.

And yes, a single VM really is fine, you can get REALLY big VMs which is fine for most business applications like we run. Most business applications only have hundreds to thousands of users. The cloud provider (Google in our case) manages hardware failures. In case we need to upgrade with downtime, we spin up a second VM next to it, provision it, and update the IP address in Cloudflare. Not even any need for a load balancer.

I don't really have the hardware to try it out, but I'm curious to see how Qwen3.5 stacks up against Gemma 4 in a comparison like this. Especially this model that was fine tuned to be good at tool calling that has more than 500k downloads as of this moment: https://huggingface.co/Jackrong/Qwen3.5-27B-Claude-4.6-Opus-...

they're almost always people who already had some pull toward software

I think this is probably true, and basically how I got into software myself.

I always dabbled in writing software and things for the web, but for some reason I never thought studying computer science would be any fun and that a career as a software developer sounded boring. But then I got an actual full time office job and oh boy, did my perspective on things change fast.

That first job did not have anything to do with writing software at all. But I saw people struggle with things that seemed to me trivial to automate, such as making annotations on paper bank statements and entering them into the system line-by-line. The bookkeeping system did support electronic bank statements, but lacked features to match certain descriptions to certain cost places. In the end it was indeed faster to go the paper route... It took me a couple of hours to write something that saved hours every week and that basically kick started my software career.

Would AI have made much of a difference here? Yes, in terms of getting to the correct solution faster, but probably not in terms of who would have done that. People would still come to the person who came up with the solution to ask for maintenance and new features.

I use voxtype on my Linux machine with parakeet. Super fast and regularly even gets the tech lingo correct. You can configure prompts and keywords to help with that as well.

Look at how many people only use their 14 inch laptop screen, it's ridiculous and terribly unergonomic.

Radeon R9700 with 32 GB VRAM is relatively affordable for the amount of RAM and with llama.cpp it runs fast enough for most things. These are workstation cards with blower fans and they are LOUD. Otherwise if you have the money to burn get a 5090 for speeeed and relatively low noise, especially if you limit power usage.

I like the UK sockets because they have a switch.

There was never a $100 billion deal. Only a letter of intent which doesn't mean anything contractually.

Wouldn't be surprised if they slowly start quantizing their models over time. Makes it easier to scale and reduce operational cost. Also makes a new release have more impact as it will be more notably "better" than what you've been using the past couple of days/weeks.

You are equally understating past performance as you are overstating current performance.

One year ago I already ran qwen2.5-coder 7B locally for pretty decent autocomplete. And I still use it today as I haven't found anything better, having tried plenty of alternatives.

Today I let LLM agents write probably 60-80% of the code, but I frequently have to steer and correct it and that final 20% still takes 80% of the time.

These LLM benchmarks are like interviews for software engineers. They get drilled on advanced algorithms for distributed computing and they ace the questions. But then it turns out that the job is to add a button the user interface and it uses new tailwind classes instead of reusing the existing ones so it is just not quite right.

Yes! This update works great. Seems to be pretty good at first glance. I'll have to setup an interesting task and see how different models approach the problem.

I use llama.vim with llama.cpp and the qwen2.5-coder 7B model. Easily fits on a 16 GB GPU and is fast even on a tiny RTX 2000 card with 70 watts of power. Quality of completions is good enough for me, if I want something more sophisticated I use something like Codex

Update: I'm experiencing issues with OpenCode and this model. I have built the latest llama.cpp and followed the Unsloth guide, but it's not usable at the moment because of:

- Tool calling doesn't work properly with OpenCode

- It repeats itself very quickly. This is addressed in the Unsloth guide and can be "fixed" by setting --dry-multiplier to 1.1 or higher

- It makes a lot of spelling errors such as replacing class/file name characters with "1". Or when I asked it to check AGENTS.md it tried to open AGANTS.md

I tried both the Q4_K_XL and Q5_K_XL quantizations and they both suffer from these issues.

I don't know whether it just doesn't work well in GGUF / llama.cpp + OpenCode but I can't get anything useful out of Devstal 2 24B running locally. Probably a skill issue on my end, but I'm not very impressed. Benchmarks are nice but they don't always translate to real life usefulness.

Yes I usually run Unsloth models, however you are linking to the big model now (355B-A32B), which I can't run on my consumer hardware.

The flash model in this thread is more than 10x smaller (30B).

Great, I've been experimenting with OpenCode and running local 30B-A3B models on llama.cpp (4 bit) on a 32 GB GPU so there's plenty of VRAM left for 128k context. So far Qwen3-coder gives the me best results. Nemotron 3 Nano is supposed to benchmark better but it doesn't really show for the kind of work I throw at it, mostly "write tests for this and that method which are not covered yet". Will give this a try once someone has quantized it in ~4 bit GGUF.

Codex is notably higher quality but also has me waiting forever. Hopefully these small models get better and better, not just at benchmarks.

Of course you can subnet ipv6, in fact I run several ipv6 subnets at home. You have to delegate a different prefix to each subnet.