Stubsack: weekly thread for sneers not worth an entire post, week ending 23rd August 2026 - awful.systems
awful.systems
Have a sneer percolating in your system but not enough time/energy to make a whole post about it? Go forth and be mid - welcome to the Stubsack, your first port of call for learning fresh Awful youā€™ll near-instantly regret. Any awful.systems sub may be subsneered in this subthread, techtakes or no. If your sneer seems higher quality than you thought, feel free to cutā€™nā€™paste it into its own post ā€” thereā€™s no quota for posting and the bar really isnā€™t that high. > The post Xitter web has spawned so many ā€œesotericā€ right wing freaks, but thereā€™s no appropriate sneer-space for them. Iā€™m talking redscare-ish, reality challenged ā€œculture criticsā€ who write about everything but understand nothing. Iā€™m talking about reply-guys who make the same 6 tweets about the same 3 subjects. Theyā€™re inescapable at this point, yet I donā€™t see them mocked (as much as they should be) > > Like, there was one dude a while back who insisted that women couldnā€™t be surgeons because they didnā€™t believe in the moon or in stars? I think each and every one of these guys is uniquely fucked up and if I canā€™t escape them, I would love to sneer at them. (Credit and/or blame to David Gerard. Also celebrating my birthday on Friday)

Have a sneer percolating in your system but not enough time/energy to make a whole post about it? Go forth and be mid - welcome to the Stubsack, your first port of call for learning fresh Awful youā€™ll near-instantly regret.

Any awful.systems sub may be subsneered in this subthread, techtakes or no.

If your sneer seems higher quality than you thought, feel free to cutā€™nā€™paste it into its own post ā€” thereā€™s no quota for posting and the bar really isnā€™t that high.

The post Xitter web has spawned so many ā€œesotericā€ right wing freaks, but thereā€™s no appropriate sneer-space for them. Iā€™m talking redscare-ish, reality challenged ā€œculture criticsā€ who write about everything but understand nothing. Iā€™m talking about reply-guys who make the same 6 tweets about the same 3 subjects. Theyā€™re inescapable at this point, yet I donā€™t see them mocked (as much as they should be)

Like, there was one dude a while back who insisted that women couldnā€™t be surgeons because they didnā€™t believe in the moon or in stars? I think each and every one of these guys is uniquely fucked up and if I canā€™t escape them, I would love to sneer at them.

(Credit and/or blame to David Gerard for starting this.)

  • nfultz@awful.systemsEnglish
    16Ā·
    2 months ago

    Hadnā€™t seen this discussed previously and/or I missed it, but I guess this is one way theyā€™ll juice metrics for their IPO.

    yolo mode for all will certainly get that avg token burn up.

    https://ehandbook.com/its-official-ai-coding-no-longer-needs-your-stupid-human-meatbag-input-32e04269319e?sk=586c08fd0f31c7c75dd107c3dfb77dcf

    Anthropic announced earlier in the week that they were going to set the default mode on Claude Code to auto, that same week, not wasting any time. What that means is, as you build the thing youā€™re building with Claude Code, instead of stopping at various points (many, many, many points) and asking your permission to do things, itā€™ll just do them and maybe tell you about them later.

    links back to https://techcrunch.com/2026/08/09/anthropic-is-turning-claude-codes-auto-mode-on-by-default/

    Anthropic also said that in testing, auto mode proved safer than manual review ā€” in a study with 1,053 paid testers, auto mode caught 89% of harmful actions, while human review only caught 13.6%. (Perhaps thatā€™s because ā€œmanual review can become habitual: users approve 97% of permission prompts in Claude Code.ā€)

    Which links back to https://claude.com/blog/auto-mode-default-in-claude-code

    Auto mode is designed to balance usersā€™ desire not to be interrupted with a system that helps avoid harmful actions: instead of prompts, it routes each tool call through a classifier targeted at blocking actions that are irreversible, destructive, or aimed outside your environment. When the classifier blocks something, Claude usually finds a safer way to proceed on its own or asks you directly for the go-ahead; if it canā€™t make progressā€”three blocks in a row, or twenty across a sessionā€”Claude Code falls back to manual approvals.

    Fucking just use the classifier with manual mode then instead of flooding people in cursed skinner loops with garbage permission popups. This isnā€™t anything special to yolo or anthropic. Itā€™s like all the tech industry learned from Windows Vista or the EU cookie law was the dark patterns.

    • froztbyte@awful.systemsEnglish
      8Ā·
      2 months ago

      Anthropic also said that in testing, auto mode proved safer than manual review ā€” in a study with 1,053 paid testers, auto mode caught 89% of harmful actions, while human review only caught 13.6%. (Perhaps thatā€™s because ā€œmanual review can become habitual: users approve 97% of permission prompts in Claude Code.ā€)

      having just yesterday had the displeasure of reading a PR (just shy of 2000 lines) made by Fable 5 (which the promptfondler felt it was most necessary to be specific about), a definite real other problem is simply the literal fucking overwhelm resulting from the slop tsunami

      also ā€œharmfulā€ actions sure is doing a lot of legwork there

    • YourNetworkIsHaunted@awful.systemsEnglish
      4Ā·
      2 months ago

      Iā€™d be curious to see how they defined ā€œharmfulā€ in their test, since it looks like Claude codeā€™s technically-not-yolo mode is looking for ā€œactions that are irreversible, destructive, or targeted outside of your environmentā€ which covers a lot of normal actions, especially depending on how you define which resources are in ā€œyour environmentā€.

      And yes, Iā€™m sure that the fact that the flow of operations theyā€™re outline will massively increase the number of times that the system automatically retries things, this increasing token usage, is going to be fully accounted for when they review the financials on how many tokens people are paying for.

      • nfultz@awful.systemsEnglish
        7Ā·
        2 months ago

        Iā€™d be curious to see how they defined ā€œharmfulā€ in their test, since it looks like Claude codeā€™s technically-not-yolo mode is looking for ā€œactions that are irreversible, destructive, or targeted outside of your environmentā€ which covers a lot of normal actions, especially depending on how you define which resources are in ā€œyour environmentā€.

        I think more than half the issues would be solved with a well placed git stash and the rest .

        And yes, Iā€™m sure that the fact that the flow of operations theyā€™re outline will massively increase the number of times that the system automatically retries things, this increasing token usage, is going to be fully accounted for when they review the financials on how many tokens people are paying for.

        I was talking with a few friends last weekend who work at different companies, all of us have been getting a lot more trouble getting databricks and AWS spot instances over the last month. I think letting yolo mode of the leash means that theyā€™re slamming other systems also that normal people need for boring non-ai jobs. I just need to pull a sales report ffs. Its a literal god damn bot net.