I've been using a private LLM (to avoid capture of copyrighted data) to predict and estimate cancel risk of various novels to see how well they do. It turns out that everything is an outrage vector. That's... actually kinda right?

As someone who wrote about politically dangerous topics in the 2010s and nearly lost my life for it, I have mixed feelings about AI. It sucks at real writing and will never get better. And good! But it's great for "focus group" testing and spotting out-of-context risk.

Literary agents' probable response to Gemini 3 being a better predictor of literary success than their own tastes: "Yeah, but were prompt engineers before it was cool. We called it drinking."

Let me make this EXTREMELY clear: - If Clinton abused kids.. • Prison - If Bill Gates abused kids.. • Prison - If Musk abused kids.. • Prison - If Trump abused kids… • Prison THIS IS NOT A POLITICAL ISSUE. YOU’RE EITHER WITH THE KIDS OR THE PED0S.

Anyone remember that guy who lost a $500k/year career ca. 2016 because he said there were murdering pedophiles at high levels in all sectors of the capitalist establishment? I should go find him in case anyone wants to apologize.

A 10-hour read of a novel by a skilled human reader no doubt beats what AI can do, but a 45-second read by Gemini 3 outclasses the 45-second “read” that is the most anyone who isn’t royalty will ever get from a literary agent, and anyone who doesn’t see that is delusional.

In the time it takes you to read this, 37 rich white girls abroad will misread Yamato politeness and become convinced that they are Good at Japan.

I'm going to be contrarian here. Scott Adams was brilliant (as in high IQ, not as in right) and an excellent cartoonist. I loathe his post-2016 politics, but game recognize game, and he did what he did very well.

AIs are now much harder to bias against serious, good literary work than literary agents. Prejudicial prompts that worked 12 months ago now fail to make it slag a book that is actually good.

Remember when the SAT proved that all the Jews the Ivies were rejecting were smarter than most of the kids who got in? That's the existential threat AI poses to literary agents in their arrogated role of cultural curators. They are FDH.

Working with AI is weird and frustrating. It can decipher multi-language puns and find obscure Easter eggs in 250,000+ word novels. Or it might churn for 5+ minutes and produce nothing, because it thought "August 1, 2025" was in the future.

The past few decades have been an illustration of ex falso quodlibet. We refused to let Iran go socialist, as it had every right to do, in 1953. The absurdity wasn't contained; it spiraled out of control.

There's a good chance that Mia Ballard will be found innocent. Increasing evidence suggests that she was dropped for mediocre sales (the UK version had already been published) and AI was used as an excuse. The accusations haven't been proven, and plenty of TP'd authors use it.

There's a good chance Mia Ballard will be found innocent. The evidence that she used AI is dodgy at best. The UK version's sales were mediocre, and Hachette wanted out. This isn't a principled stand, but commercial optimization. All the same as ever before.

Traditional publishing: Steals the village. Also traditional publishers: Discards writers if there's even a 1% chance they used AI to replace the stolen village. I don't have much respect for "writers" who push AI slop as their own work. But a witch hunt's no good for anyone.

BRB battling archons