I see garbage text on the tower in the centre, indicating it's AI-generated slop imagery, and I instantly assume any code from this soi-disant IDE will be as full of mistakes as the marketing graphics.
-
@cstross Oh, Charlie. It's far far worse than you imagine. It's not an IDE; it's an entire Russian-doll stack of workflows populated by LLM coding agents.
As art, it's mindboggling. As engineering, well, it's mindboggling.
@vivtek The engineering is mind-boggling in the same way the original Tay Bridge was mind-boggling (especially to William McGonagall).
-
@vivtek The engineering is mind-boggling in the same way the original Tay Bridge was mind-boggling (especially to William McGonagall).
@cstross Exactly that.
-
@cstross Exactly that.
@cstross Maybe it would have been more apt to say that as art it's mindboggling, while as engineering it's more like mindFLAYing.
-
RE: https://furry.engineer/@talisyn/116029594929462963
I see garbage text on the tower in the centre, indicating it's AI-generated slop imagery, and I instantly assume any code from this soi-disant IDE will be as full of mistakes as the marketing graphics.
-
RE: https://furry.engineer/@talisyn/116029594929462963
I see garbage text on the tower in the centre, indicating it's AI-generated slop imagery, and I instantly assume any code from this soi-disant IDE will be as full of mistakes as the marketing graphics.
@cstross having read through this, I can see why it is ridiculously expensive for him to run. It requiring two Claude Code accounts now with a projected third is... interesting. Running this project must light money on fire.
I can't imagine what the cost to Anthropulic for this one user is.
This article is from July:
' “But a few outlying cases are very costly to support. For example, one user consumed tens of thousands in model usage on a $200 plan.” '
-
@cstross having read through this, I can see why it is ridiculously expensive for him to run. It requiring two Claude Code accounts now with a projected third is... interesting. Running this project must light money on fire.
I can't imagine what the cost to Anthropulic for this one user is.
This article is from July:
' “But a few outlying cases are very costly to support. For example, one user consumed tens of thousands in model usage on a $200 plan.” '
@cstross @GreatBigTable oh, don't worry, the crypto grifters minted a coin for him so now he has extra money to light on fire. Here's a sober review of the whole situation https://maggieappleton.com/gastown
-
RE: https://furry.engineer/@talisyn/116029594929462963
I see garbage text on the tower in the centre, indicating it's AI-generated slop imagery, and I instantly assume any code from this soi-disant IDE will be as full of mistakes as the marketing graphics.
@cstross Reading through Stevie's article, I can't tell if this is brilliant genius, performance art that went off the rails, biting parody, or incipient schizophrenia. Maybe all of the above?
-
@cstross Reading through Stevie's article, I can't tell if this is brilliant genius, performance art that went off the rails, biting parody, or incipient schizophrenia. Maybe all of the above?
-
RE: https://furry.engineer/@talisyn/116029594929462963
I see garbage text on the tower in the centre, indicating it's AI-generated slop imagery, and I instantly assume any code from this soi-disant IDE will be as full of mistakes as the marketing graphics.
Aha! Apparently, like every sideshow spiritualist, medium, and psychic has said, the presence of doubters will spoil the thing and stop it working. Only doubt-free people may cross the palm with cryptocurrency.
So: This Gas Town …
… I suspect that we know what sort of lighting it uses.
-
@anthracite @cstross But it's good to compare & contrast to @simon post today about fully automated coding at StrongDM:
https://fedi.simonwillison.net/@simon/116030108431644256
Their approach seems slightly more ... sane ... than Stevie's. And they are *making it work*.
Inefficient? Surely. Potentially error-prone? Yep. Deeply unsettling? Very much so.
But shit like this would have been considered *impossible* just six months ago. And the one of the first principles of software engineering is to get it working first, and worry about optimization later.
We dismiss these technologies at our own peril. This future is happening *now*, like it or not.
-
RE: https://furry.engineer/@talisyn/116029594929462963
I see garbage text on the tower in the centre, indicating it's AI-generated slop imagery, and I instantly assume any code from this soi-disant IDE will be as full of mistakes as the marketing graphics.
@cstross didn't even need to look at the text. Piss filter... Just has The Look