In case you are in the camp of “Andrew Tridgell is vibefucking rsync” please read this.l and adjust your priors. medium.com/@tridge60/rs...

Trump et al want to be authoritarian dictators, but are not good at it and will probably fail. This is good news when it eventually happens, not good news in the meanwhile. Things could have been better. www.youtube.com/watch?v=CV4u...

🌳 pipdeptree 3.0 is out. New: render a dependency tree straight from PyPI (needs index extra) or a PEP 751 lockfile without installing anything, render it in a Jupyter notebook, and see extras by default. Try it 👉 pipdeptree.readthedocs.io/en/latest/tu...

this is the kind of merit that pete hegseth has brought to the department of defense

I support this group. A lot. Everyone can attend this virtual meeting and get fired up and have a call to action.

Hamburger helper only works if the hamburger is ready to accept help.

Linter for looking evidence of too many ignore/skips #python pypi.org/project/dont...

IDEA: LLM client, except it is an email client and you email your queries to the bot and get back email. I figure it could be an alternative to interactive chat for local models, which are so sloooooooooooooow.

I don't mind that #agy has limits, but it is dog shit for doing work within token constraints. Got no control over how many agents it want to run, no control over its plans, you can't even see that it is waiting on approval, can't see how much is left-it reports 100% on all models-nope 4 hour wait!

I like how this McSweeney's headline puts it www.mcsweeneys.net/articles/cal...

Well, I'm at Backup Mastodon again.

Why is #agy so bad? They changed the switches from gemini cli, it can't detect terminal size and appears to squish it all into 80 columns, and the docs are loath to tell you that you start the app with `agy` #antigravity And it is google so there is no way to tell them.

An LLM told an appropriate joke. Damn. "❯ Do it. <insert emperor Palpatine meme, except w/o the implication of urging you to do scifi war crimes> ● Executing Order 66... I mean, cloning all repos: "

There are fuck-ups, and then there's this krebsonsecurity.com/2026/05/cisa...

A republic, if you can an afford it. * which is now more accessible than ever thanks to monthly payments with Klarna!

IDEA: ClaudeProject. It is a tool for CIOs to give Claude some direction on what the project is and the enterprise goals then Claude tells human programmers to go write the code. ClaudeProject spawns subagents by posting job listing and hiring developers, testers who then create the application.

The bag of coffee I buy every week -- whose price fluctuated between $7.99 and $9.99 for about a decade, but never went higher -- was $15.99 when I went to the store today.

So I've been experimenting how lazy I can get with Claude Code and friends. Pretty lazy. At the limit, I'm a Product Owner, Manual Tester. I'm going to try this with languages I don't know and see what happens.

v2 of best effort sandbox for python Use cases - LLM (defense against LLM mistakes) - unit tests (defense against self) - plugins (defense of mostly trusted code) Not a use case - defense against malicious user who will take the time to craft a workaround hermetic.readthedocs.io/en/latest/

Tech giant accused of ‘cynical attempt to bust the union’ in decision to shutter location in Towson, Maryland

ChatGPT: "Let me tell you about my favorite animals"

Back of the envelope calculation says the pay back period on a RTX4090/5090 vs running Qwen3.6 Plus is 24 years. From the sound of it, if I can think up enough work for Qwen, it would be a $4/m to $20/month monthly API spend.

Can't edit. Why did they have to copy twitter's worst feature.

Has #ClaudeCode been remove from the $20 plan? claude.com/pricing

If you use an LLM to "rewrite" your text, you risk losing your voice, and it sounds like you had no input in it because stylistically it is all LLM. Better way: let the bot tell you what is wrong and you fix it. You think because you are lazy and don't give a shit about your voice, it's justified

Well here I am at backup mastodon. Regular mastodon is under a DDOS attack status.mastodon.social

New workflow: - use up all of my claude tokens on the weekend. - spend the weekdays catching up on what just happened I think the "generate code all day every day" is false productivity. It's a literal "Sorcerer's Apprentice" problem, teaching the brooms to bring you infinite buckets of water.

My Claude code now resets on Thursday and this has really rearranged all my coding plans for the week. Now I have to prompt extra Wednesday and Thursday? I bet this was happening on all accounts and causing surges on Saturday when it used to reset.

We found out several things during the wait. 1) target uses AI for theft prevention which often falsely targets people. 2) they purposly don't notify you if you don't scan all items in a self checkout lane. 3) if u hit the ok button on the self checkout and you miss an item they say thats stealing

Where the grass is [redacted] and the girls are [redacted].

Claude $20 plan life - enable spending money at end of a session so I'm not stuck in the middle of a session - disable it when there is session limit to spare - write code after 3pm EST - evenly space work across the week - interruptible work on Friday/Saturday - Opus work starting 5PM EST Saturday

I'm on the Claude $20 plan. For a while I kept hitting limits. Then changed my behavior and this week I'm having a hard time hitting my limit. - rotate thru my other plans - convos are 1-2 asks long - opus for hard stuff and finish conversation/implementation using sonnet

Eliza Code: Doesn't fix bugs. Makes you feel better about your self as you make bugs.

Why Trump talks that way. Good short video if your a language geek www.youtube.com/watch?v=-jmq...

Tell claude to talk like a caveman to cut down on thinking and tokens and token usage github.com/JuliusBrusse...

It's alive! Like-claude code but semi-agentic and no tools, except some file writing and no file editing. tuochat - agentic coding for cowards. pypi.org/project/tuoc...

Opus- hard stuff Sonnet- difficult stuff Copilot/Codex- work horse because I have so much usage quota and it is almost as good as Sonnet. Gemini- Look for bugs, do the type annotations, do the lint, write basic unit tests Kilo/Opencode's free models- not using them much...

#msteams shows your status as the white circle of death while connecting to the network. The white circle of death is the "this person has been fired" status symbol. Grey (x) means offline/don't know.

Well shit. Next war my neighborhood is going to get blown up, with all the data centers and all.

I don't know if agentic coding is going to fly in a corporate environment. I need to write in my AGENTS . md file `if you prefix any more variables with _ as a signal of "private", I will knife you` and not sure you can do that at the office.

"toilet go for use!" "that's one small shit for a man..."

Google's Gemini and Codex will do what Microsoft couldn't make me do for decades. Learn Powershell. Those bots seem to have been forced to use powershell. They keep trying to use bash syntax tho.

The anti-LLM sentiment on Mastodon has gone from tepid to incendiary over the last year or two. I'm not really sold on the incendiary arguments because it just so emotionally over the top.

Most of my side projects were either python libraries or CLI tools. I'm having the LLM write tkinter GUIs for many of them because it help with the discoverability. You might still ultimately use it as a python library or part of a bash script, but the GUI shows you how it works.

I think this fixes it.

Fix your PATH with pathkeeper - it got messed up somehow, this will try to repopulate it - backup and restore so you can keep it fixed Vibe coded, but subjected to a fairly strict build script and I did some user testing. pypi.org/project/path...

Well, I guess I'm on backup mastodon for the day. Mastodon Social is barely standing. Probably was out drinking Ouzo all night. status.mastodon.social

Who'll to do it? I'll `make claude` ```Makefile claude: claude --allow-dangerously-skip-permissions ```

Let's Pedestrize the Beltway! Positioned to experience great benefits from the creation of a pedestrianized zone; the region would see an increase in safety, economic impact, and livability, and a decrease in pollutants, injuries, and stress. More information: jeramyolmack.com/pedestrianiz...