roywiggins

Only if the replacement is global and not, say, only looking and inserting it into the actual (eg) Authorization header. If something is only transparently altering the Authorization header, then an agent inserting the dummy value somewhere else is totally safe.

It has a sort of metronomic quality. It never slows down or speeds up or modulates its tone. It plods forward at a relentless pace and never has a light touch with anything.

I think this is one reason why LLM text is pretty exhausting to read for long stretches.

I guess ideally you'd want to force it to pipe each step back into the agent/IDE for display (and optionally wait for approval) rather than shoving it into the context. And with some sort of heuristic to skip commands that are piped into sed and similar.

or use just use actual files instead of pipes with a permissions dialogue gate on write. That would work already in Claude I think, and with IDE integration it would show each step as it went.

Of course buffering every step of a pipeline will have its own side-effects.

Claude had some pretty good ideas about how to end the war in Iran when I asked it a while ago (The US should sue for peace, I believe was the conclusion).

I'm so tired of opening another link and finding it's just Claude blogging under a pseudonym

There is no lack of more elaborate logics:

https://en.wikipedia.org/wiki/Non-classical_logic

What if "way short of AGI" is still dangerous in ways we can't easily handle?

Pi Agent is even leaner feeling, it's worth a try.

It's cheaper to destructively scan.

Good news, airlines already barely profit:

https://www.iata.org/en/pressroom/2025-releases/2025-12-09-0...

Even more worrying if it's not deliberate and we are all just naturally writing like that now.

Sometimes that's me. I wish more submissions didn't read like AI. I worry that people can't detect tiresomely LLMy output and upvote it to the front page. I'm not mad or anything, I just find the internet less and less useful, because so much that circulates pings my personal AI detector (running in my skull). Sometimes I just close the tab in frustration but sometimes I comment, because it sucks.

(I don't actually care that much if it is LLM text, I care when something reads like it is, because it means everyone writing blogs using this style sounds like everyone else. it's alienating!)

My problem is that my personal AI detector running on wetware just goes ping and it's hard to articulate why. I agree with the person you're replying to on all points but also, when I think something is AI, it's very much just a little brain ping that I would have to laboriously try to work backwards from.

LLM text often just is missing the cadence of human-written text. It never slows down or speeds up; it keeps a relentless pace. Sure, a human can write like that, and I'm sure people do, but I find it enervating to read, whether it was a human or an AI that wrote it.

A shame that Pangram flags this as AI.

https://www.pangram.com/history/c0a9cde2-7a5c-4588-83a3-0269...

ai; dr, unfortunately

Long sequences of short sentences. Not X, but Y.

And Pangram flags it, too.

https://www.pangram.com/history/126831e1-562f-4e65-9874-5250...

Weirdly not dissimilar from MUMPS systems.

ai; dr

Very likely. I'm just saying, people in the CIA seeing where the tech might be going and hedging their bets is not that unlikely.

Common side-effect of letting Claude write your landing page.

Apparently it's mostly this:

    // ENCODE — pack(html) → one .hmml (a Uint8Array you store or send)
    //   · lifts every data: image out of the HTML into raw bytes (no base64)
    //   · gzips the HTML/CSS/JS and frames it all as one binary blob

Well, define "drone warfare"- the CIA and the Pentagon has been operating Predators and friends for a long while.

The idea is to expose it as a tool to your LLM agent so it can run calculations on its own initiative.

Your rewritten post is far easier for me to read now, fwiw.

LLM output has conditioned in me a near reflex response to just close a tab as soon as I smell LLM-authored text. Like, I'm not mad or anything, I just frequently find most default LLM-voiced text very unpleasant to read so I just don't continue reading.

"quietly" seems to be the new one recently

If something's a bad tool that misleads people into doing bad work, it would be good to know that.

> A simple distributional analysis of every rsync release with bug data. No model. No assumptions. Just placement.

If you want me to read your analysis, you are going to have to make it not read like Claude wrote it. What does "placement" even mean here?

Pangram seems to be tuned to avoid false positives most of the time.

it can operate at the level of a mere mathematics professor, who everyone knows are barely conscious, basically automatons. wake me up when it's Einstein

Pangram flags this as 100% LLM output fwiw

It's that companies like copilot/cursor are in real trouble if they are in the business of reselling expensive Anthropic tokens