Skip Navigation

InitialsDiceBearhttps://github.com/dicebear/dicebearhttps://creativecommons.org/publicdomain/zero/1.0/„Initials” (https://github.com/dicebear/dicebear) by „DiceBear”, licensed under „CC0 1.0” (https://creativecommons.org/publicdomain/zero/1.0/)P

percent

@ percent @infosec.pub

Posts
2
Comments
451
Joined
1 yr. ago

I've never really followed X (nor Twitter), Bluesky, Instagram, TikTok, etc. so I basically live under a rock. Sometimes I ask dumb questions to try to understand people a little better. Apologies if my questions inadvertently offend anyone. I mean no harm.

  • Claude-powered AI coding agent deletes entire company database in 9 seconds — backups zapped, after Cursor tool powered by Anthropic's Claude goes rogue

    Jump

  • Seems like they were operating with a pile of bad practices, then threw AI into the mix.

    Neural networks are approximation algorithms. There's a reason LLMs are generally more productive with statically typed languages, TDD, etc. They need those feedback loops and guard rails, or they'll just carry on as if assuming they never make mistakes (which tends to have a compounding effect).

    If you want to use AI safely, you should be more defensive about it. It will fuck up; plan accordingly.

  • Claude-powered AI coding agent deletes entire company database in 9 seconds — backups zapped, after Cursor tool powered by Anthropic's Claude goes rogue

    Jump

  • These protocols predate LLMs

  • Why are people speculating that the most recent Trump assassination attempt was fake?

    Jump

  • There will always be conspiracy theories for things like this. It's probably just the usual noise

  • jigglyfluff

    Jump

  • It's chest hair

  • DeepSeek ditches Nvidia for Huawei chips in V4 launch

    Jump

  • Gemma4:26b is also worth trying. I find it runs much faster on my hardware.

    Edit: Qwen3.6:35B might be the sweet spot. It's bigger than the 27B, but actually more lightweight when running. TIL the 27B is not a MoE model; it's a dense model. The 35B is a MoE model with only 3B active params.

    So far, I think Qwen3.6:35B might be giving me better results than Gemma4:26B. It's a bit slower than Gemma4:26B, but definitely faster than Qwen3.6:27B.

  • DeepSeek ditches Nvidia for Huawei chips in V4 launch

    Jump

  • +1, exactly the same experience. Except Gemma4:26B really sucks with OpenCode. Works great with Pi though

  • The Pulse: ‘Tokenmaxxing’ as a weird new trend

    Jump

  • Do you not use it enough because yet get bad results? I discovered that, no matter how smart the LLM might be, its first attempt is never its best work. Tell it to review its work (or its plan, if using planning mode). If it makes any changes, tell it to review its work again. Repeat until there are no more changes.

    (You don't actually have to do this repetition manually; just tell the AI to do it in a loop. I recommend making it into a SKILL.md so you don't have to explain the loop every time.)

    With these loops, I get better results AND burn lots of tokens. (Yes, it feels strange that excessive token consumption is actually considered a good thing)

  • Go on...

    Jump

  • The one that goes "siiimply haaaving a wonderful Christmas time"

  • Microsoft drops Copilot branding in Notepad for Windows 11 for everyone, but it’s really just a rename

    Jump

  • That's Notepad in the screenshot? I haven't used Windows in years, but I remember Notepad being the one that didn't do rich text. Did they just fold Wordpad into Notepad and add Copilot?

  • Do I belong in tech anymore?

    Jump

  • Before AI, I was spending a lot of time fixing slop left behind by people who didn't seem to care about quality.

    Now, I (am forced to) use AI to clean up slop from the same people, but they're also armed with AI (though, to be fair, I think AI adoption actually had a net positive effect on their code quality).

    So yeah I agree, it's definitely magnified.

    Honestly though, slop cleanup doesn't affect my mood as much anymore, now that I don't have to do it manually. I had AI agents identify ~15 bugs yesterday. Then, I had agents handle the entire flow for each one in parallel: Create a ticket in the issue tracker, fix the bug (TDD approach), open a PR, monitor CI while it runs, wait for review (and address change requests/feedback), merge, then close the ticket. It was a lot more in a single day than I used to do manually.

  • Wonder why?

    Jump

  • 🤔 I wonder what they did to the people who offered the bribes

  • Software names

    Jump

  • Wow, I haven't heard of BitchX in many years! Decades, I think. Thanks for bringing up some old memories 🙂

  • Do you wash your hands after going pee?

    Jump

  • It is cleaner before reaching the urethra, but not sterile.

  • What's your backup plan if AI takes your job?

    Jump

  • Flint's water is okay now

  • Conforms to Marxist theories regarding the proletariat.

    Jump

  • Maybe it's both. I would guess that men's feet would get tipped more than no feet, but less than women's feet.

  • School of hard knocks

    Jump

  • When I had to hire people, I was much more interested in seeing a portfolio than a degree.

    It depends on what the job is though. I definitely want my doctor to have a degree

  • Free resources

    Jump

  • Not in a troy ounce though

  • Deleted

    Permanently Deleted

    Jump

  • Why?

  • Artemis vs. SpaceX

    Jump

  • SpaceX is by far the cheapest way to launch a satellite. As a taxpayer, that seems like a good choice, IMO.