AI is not doing the jobs; the remaining employees are.

Interesting read. Here’s a link without the paywall: https://archive.ph/20260507223322/https://www.bloomberg.com/news/articles/2026-04-27/why-china-s-deepseek-qwen-and-moonshot-are-a-worry-for-us-ai-rivals

Edit:

Anthropic has accused DeepSeek, Moonshot and another Chinese AI lab, MiniMax, of “industrial-scale distillation attacks” — illegally extracting capabilities from its proprietary Claude model using 24,000 fraudulent accounts to gain an edge

Anthropic’s says:

Illicitly distilled models lack necessary safeguards, creating significant national security risk

So open weight models are a national security risk? Guess we’d all better pay subscriptions, let the data centers be built, and let a couple companies building proprietary models have a monopoly to protect our security, then. 🙄

Right? Qwen 3.6 27B is fantastic, for example. Data centers need to stop hoarding all the hardware so we can afford to run models like that.

Spanking the code monkey.

I think the key point is that you’re not outsourcing critical thinking to LLMs, but are instead using it as a tool to do grunt work that you could’ve done yourself, but the LLM can pump out faster. This means constantly being critical of everything it does, asking questions, asking for links to credible sources, asking it to provide info to help evaluate the pros and cons of multiple approaches, with you making the decisions and learning along the way. Overall, any work a LLM produces that will have your name on it should be work you entirely understand and agree with. For coding, I find agent markdown files to be especially helpful to make sure the LLM follows my desired practices without me constantly making it refactor.

Largely, my assumption at this point is that LLMs may not always be around, so I definitely don’t want to be left holding the bag with a bunch of slop I can’t manage on my own. I think I’ll feel better when I can run open weight models on my own hardware that are fully competitive with cloud models. With models like Qwen 3.6 27B, it seems we are getting closer to that.

deleted by creator

Are all these data centers really going to be running at full capacity with open models like Qwen 3.6 27B that have performance approaching frontier, but can run on consumer hardware? Sure, it’s slow as of now, though there are tweaks to optimize it, and how long until we see open models that run reasonably fast and give frontier models a run for their money? My company MacBook can run models like this, so will there be a point where companies stop paying hundreds per user per month for cloud AI and have devs run open models on the laptops they already have? I definitely won’t be surprised if that’s the case.

An official member of the US military industrial complex is making a phone with a proprietary OS that hoovers up your data and shoves AI slop in your face 24/7. What’s not to like?

Maybe try old.reddit.com?

Lots of fond memories watching Planet Earth in the evenings with a cup of herbal tea.

I got an espresso machine a few years ago and learned to make a proper latte with it. At this point, a $9 cup of charry sugar water made by a teenager in a fast food restaurant doesn’t really appeal to me.

Just as open weight models are getting good. Qwen 3.6 27B just dropped with claimed performance approaching Opus 4.6, but it can run on a Mac with a M-series SoC. I tested it out today on a M4 Pro with Ollama and Cline and was impressed with its reasoning, but it was slow. Going to try with llama.cpp tomorrow and mess around tweaking it for speed.

https://ai.rs/ai-developer/qwen-3-6-27b-local-coding-model

AI coding agents are useful, but it’s time for the cloud-based models to chill out so we can get cheap RAM again to run our shit locally.

If we are going to eschew open source projects from shitty tech companies, then there’s a pretty long list.

Quite an interesting development and here’s hoping this makes it to production.

Cool, let me know when the model leaks. 🥱

I know people who lied about having a degree, could do the job, and never got caught. I suppose speed running a degree from a degree mill yields a similar level of education, except with a piece of paper.

I should not have laughed so hard at that. I’m a horrible person.

I decided against Backblaze for server backups because they charge for certain API calls, and I ended up exceeding the quota when I was testing with the free tier. I was experimenting with encrypted backups and not sure how I exceeded it, but it really put me off that I could potentially have a surprise bill from experimenting without exceeding my storage quota. I went with iDrive e2 specifically because they don’t have API fees and it has worked fine the last couple years. My storage utilization has grown and I’ve been charged extra, which is expected, whereas API calls would be harder to predict depending on what I do in a given month. For self-hosting, I want easy, predictable pricing and don’t want to deal with surprise bills. It’s enough of a chore to manage cloud spend at work without it being a headache at home too.

I first heard pizzle in Kingdom Come Deliverance: “Are you pulling my pizzle?”