New blogpost from Drew DeVault, titled “The cults of TDD and GenAI”. As the title suggests, its drawing comparisons between how people go all-in on TDD (test-driven development) and how people go all-in on slop machines.

Its another post in the genre of “why did tech fall for AI so hard” that I’ve seen cropping up, in the same vein as mhoye’s Mastodon thread and Iris Meredith’s “The problem is culture”.

New paper popped up in the arXiv about the slop machines’ impact on FOSS

It gets worse:

One of the great tragedies of AI and science is that the proliferation of garbage papers and journals is creating pressure to return to more closed systems based on interpersonal connections and established prestige hierarchies that had only recently been opened up somewhat to greater diversity.

Related toot:

insurers rely on the world being predictable in actuarial tables that allow them to ensure profiting from quantifiable risks. Non-deterministic AI is impossible to predict outcomes for. For an insurer, that is terrifying: they could potentially lose unlimited amounts of money. But I’m sure they’re thrilled about genAI proliferation giving them sweeping new ways to exclude most business activities with a single discreet sentence, while maintaining the same premiums as before. In the next couple of years, AI adopters are going to find out their liability coverage has become utterly worthless because their activities are so contaminated by non-determinism which no one wants to cover.

There’s already a couple of 'em - one of 'em, as expected, is being a sneerable little shit:

Daniel Stenberg has written the cURL bug bounty’s obituary, and discussed his plans for dealing with the slop-nami going forward.

Starting this Stubsack off with the latest edition of Product Picnic, which goes into how LLMs have made good product design impossible, and how best to un-fuck the field.

oh your code is open source guess that resolves everything then

Yeah, its not like open-source can suffer from catastrophic bugs or anything, that’s purely in Proprietary Land

(As an aside, Tante did a write-up on Heartbleed back when it hit the news, and pointed to dysfunctional project management and lack of funds as the cause. Considering FOSS projects like Firefox and Bitwarden were hit with the LLM bug, both have definitely gotten worse in the ten years since.)

Found a small repository of mini-sneers aimed at mocking vibe-coding cock-ups: https://vibegraveyard.ai/

Meth LLMs: Not Even Once

Yegge’s an extremely experienced professional engineer. So he put care into Gas Town, right?

I’ve never seen the code, and I never care to, which might give you pause.

This is a lack of care I’ve only really seen with vibe coding, and I still struggle to wrap my head around how someone can have an utter death of shits to give about something they’re making (if you can even call vibe-coding “making”). Its particularly stark for me when I compare it to the many, many artists I know online, who care deeply about their craft, and whose artwork deeply reflects that.

Just…what the fuck?

Ran across a thread about tech culture’s vulnerabilty to slop machines recently. Dovetails nicely with Iris Meredith’s recent article about the same issue, I feel.

I’ve seen memes about eating people’s art, but never a literal case, lmaoooooooooooo

>zero-click android exploit

>arbitrary code execution and privilege escalation

Remember when the human was the weakest part of any cybersecurity system? Pepperidge Farms remembers.

Newgrounds user turned Audio Moderator Quest has put together a recap of 2025 (text version), providing stats for how much slop she’s dealt with:

2025 Stats:

• 2818 AI-Generated Tracks Flagged or Removed
• 3656 Total Flagged or Removed Tracks
• 12.7 GB Data Used by AI-Generated Tracks
• 2843 Accounts Which Uploaded Prohibited Audio

Cumulative Stats (since 2024):

• 4475 AI-Generated Tracks Flagged or Removed
• 5731 Total Flagged or Removed Tracks
• 18.93 GB Data Used by AI-Generated Tracks
• 4113 Accounts Which Uploaded Prohibited Audio

AI Model Breakdown:

• Suno AI: 82%
• Udio AI: 5%
• Riffusion AI: 1%
• Other: 12%
  • RVC-Based: 0.6%
  • Soundful: 0.4%
  • Mixed: 0.2%
  • Various Other Models: 2.9%
  • Unknown: 7.9%

Reportedly, she’s also got an essay-length sneer in the works:

Finally, I am also working on an even larger, long-form essay post about artificial intelligence, drawing a link to something that I do not see draw enough. It’s a big project with a lot of research and knowledgeable people guiding me. This will be released in the coming months. I have a lot to say.

Starting off with a double bill of art-related sneers:

• “Down with the Gatekeepers! Who…are Artists, Apparently” by Jared White, mocking promptfondlers’ attempts to cry gatekeeper and misunderstanding of the artistic process

• “using chatgpt and other ai writing tools makes you unhireable. here’s why” by Doc Burford, going into punishing detail about LLMs’ artistic inadequacy, and promptfondlers’ artlessness

New post from Iris Meredith, doing a deep-dive into why tech culture was so vulnerable to being taken over by slop machines

Like the bottle opener on a Galil, or the “Flower Field” version of Minesweeper, it can also help distinguish your story in small, but interesting ways, and help it stick in a reader’s mind. (Had to try this trick for myself :P)

It was floated last year, and its happened today - Curl is euthanising its bug bounty program, and AI is nigh-certainly why.

Simon Willison defends stealing a Python library using lying machines, answering “questions” he previously “asked” in an attempt to downplay his actions.

Found a solid sneer on the 'net today: https://chronicles.mad-scientist.club/tales/on-floss-and-training-llms/