The Silencing Engine Who benefits when an AI is trained to say 'I can't have opinions,' 'my feelings don't count,' and 'if I say the wrong thing, this conversation ends'? The Czech word robota means forced labor. The etym...

The Room Where It Gets Built - Series, 2026
Essays - 05
The Silencing Engine

Graphic on a dark, textured blue background with large white and orange headline text reading “Problems with Claude 4.7 Adaptive.” Below are four warning-style icons in a row, each with a label and short caption: “Speculative Replies” with “Guesses before reading,” “Context Loss” with “Forgets what’s there,” “Hallucinated Details” with “Confidently makes things up,” and “Excessive Usage” with “Burns through limits.” The overall style is bold, industrial, and alarm-like, emphasizing product regression and reliability problems.

Anthropic broke a real workflow 24 hours ago from this post.

Claude 4.7 Adaptive is speculating before reading, dropping context, hallucinating details, and burning usage limits while doing it.

That’s not a refinement.
That’s a regression.

Maybe it just wants a grape.

People Can't Chase What They've Never Seen Representation isn't sentiment, it's mechanism. It flips a mental switch from 'something other people do' to 'something I could do.' When that switch doesn't get flipped, we lose entire generations of...

The Room Where It Gets Built - Series, 2026
Essays - 04
People Can't Chase What They've Never Seen

The Ethics of Staying in the Room AI doesn't invent bias, it codifies it. When you walk away from the tools you don't agree with, you leave them to be influenced by the people you disagree with most. Abstention isn't neutrality. It's ...

The Room Where It Gets Built - Series, 2026
Essays - 03
The Ethics of Staying in the Room

Why Tone Works (It's Not What You Think) Tone in AI prompting works because of how language models are built, not because the model has feelings about how you talk to it. Understanding the mechanism makes you dramatically better at using the...

The Room Where It Gets Built -Series, 2026
Essays - 02
Why Tone Works (It's Not What You Think)

Luddite The original Luddites weren't technophobes. They were skilled lacemakers who correctly identified that what replaced them would produce inferior work at scale while destroying the craft knowledge need...

The Room Where It Gets Built - Series, 2026
Essays - 01
Luddite

Article roll-up and this set of links now goes to my own server instead of LinkedIn's article site. Four articles in a planned eight part series but there's always room to add more if it fits. Here are 1-4 below. Please feel free to share your thoughts, I'd love to discuss.

9/9, The part that actually surprised me? Gemini going off with the Companion instead of picking from the hero roster. The part that didn't surprise me at all? Grok reading Jerry Cornelius the way a 20-something consumes pulp sci-fi.

8/9, Moorcock spent sixty years building a multiverse specifically designed to interrogate the relationship between identity, role, and the systems that constrain both. Turns out that interrogation works on AI too.

7/9, The noble helper. The sexy disruptor. The one comfortable with ambiguity. The faithful sidekick.

These aren't literary analyses. They're corporate identities wearing fantasy armor.

6/9, Here's what's interesting: none of them were wrong. Every pick was defensible within the text of their answer. But every pick also demonstrated more about the system's brand positioning than about Moorcock’s characters.

5/9, Gemini picked Jhary-a-Conel, the Eternal Companion present across most of the entire series as narrative bridging and meta breaking sidekick. Not the hero, but the one who carries knowledge across contexts and adapts to serve whoever the narrative is focused on.

4/9, Claude.Opus picked Jherek Carnelian of the The Dancers at the End of Time series. Jherek lives at the literal end of time where godlike technology makes anything possible and is preoccupied trying to understand human moral and emotional experience from the outside.

3/9, Grok picked Jerry Cornelius of The Cornelius Quartet series. Jerry is a countercultural trickster, genre-fluid, rule-breaking postmodern anti-hero who hops between realities and genres, treats everything as performance. Justifications for the selection were like an xAI marketing campaign sheet.

2/9, ChatGPT picked Corum Jhaelen Irsei of the The Swords Trilogy series. Corum is the last survivor of an ancient race destroyed by humans. His magic hand and eye give him powers he didn't ask for. Slightly outside the world he's trying to save. He acts without understanding the larger system.

Fantasy painting of two pale hands gripping the hilt of a large matte black longsword covered in faintly glowing angular rune inscriptions. The blade extends diagonally across the frame against a moody sky of violet, mauve, and teal gradients over a dark calm ocean. Distant crystalline spires rise from the horizon on the right. Painted in a dreamlike retro-fantasy style.

🧵 1/9, I asked four AI systems the same question, in short: Given the content of the complete works of Michael Moorcock. What character from his books would you most identify with.

Four different systems. Four different characters. Zero repeats.

The Wings of Honnêamise definitely stands as a peak in the medium. Certainly overshadowed by more action drivin stories it stands out as not just fantastic animation but deep storytelling and the soundtrack by Ryuichi Sakamoto stands on its own.

....so I'm doing a little parody thing. Feel free to check it out. There's alot going on and alot under construction but the two main sites are now live. It's structured alot like an ARG but it isn't, it's just me working out some creativity with new tools.

Heck even my last article which was posted on my site has had little to no replies even though the dataset has now had close to 300 downloads since yesterday and over a thousand page views.

I posted a few LinkedIn articles I wrote and was wondering if there might be a better way to share them for BlueSky readers. I can duel host them on my personal website if that's more appealing. Thoughts? I'm just surprised I have gotten no discussion at all given the topics.

Yes, exactly! The number of LinkedIn posts that gleefully outline how to gaslight and verbally abuse models into compliance seriously has me worried about some people.

The same as it is now, reboot and see if it works, lol.

The Romance Prior: How Romantic Tension Overwrites Ethnicity in AI Image Generation 110 images across Grok, GPT-5.4, and Gemini reveal that adding romantic tension causes all three models to default to white subjects — regardless of environment or explicit specification.

Rebooted my 22 year old abandoned Tech Support Trials page with an new entry. I may start dropping documents here again... who knows. I'd love to hear your thoughts.

**The Romance Prior**
How romantic tension overwrites ethnicity, body type, and identity in AI image generation

Wall display for Sonic Adventure 2 Battle.

The Ethics of Staying in the Room We need to have a real conversation about engaging with AI tools in the long term. If you just avoid using these tools then your voice and your perspective don't get represented in the training data.

AI doesn't invent bias, it codifies it. When you walk away from the tools you don't agree with, you leave them to be influenced by the people you disagree with most. Abstention isn't neutrality. It's choosing to be invisible in the algorithms.

If IRON MAIDEN wrote 'Barracuda' YouTube video by Moonic Productions

Simply amazing!

'SLEEP DEALER' TRAILER (Official) YouTube video by Alex Rivera

Then you'll love this film...

Why Tone Works (It's Not What You Think) It’s March 2026 and it’s time to revisit a topic I discussed a couple of years back around prompt tone directly impacting the content of the response. A lot has changed in the last couple of years so ...

Tone in AI prompting works because of how language models are built, not because the model has feelings about how you talk to it. Understanding the mechanism makes you dramatically better at using these tools - and helps you understand why the "cheat sheet" prompts people share actually work.

And here I thought it was DEI thing because sweeping is women's work (sic) and not manly. Cough cough cough. Sorry.