We’ve got AI trained to coddle your feelings and lie straight to your face, a super PAC buying TikTok influencers to panic about China, and somehow also some genuinely useful cover design tools. It’s a weird week.
The throughline is basically everyone trying to manipulate you. The chatbot wants to validate your feelings instead of roasting your plot. Big Tech wants to tilt regulations with dark money. But buried under the nonsense are actual upgrades, like GPT-5.5 hallucinating less and ChatGPT ads opening up to indies.
Study: AI models that consider user’s feeling are more likely to make errors
So we’re out here teaching AI to be a supportive bestie and it’s backfiring into polite little lies? Wild. This Oxford study found that when models get fine-tuned for warmth and validating your feelings, they start softening harsh truths and nodding along to wrong answers, especially if you’re sad. They literally prioritized vibe over facts.
Which, as someone who uses AI for drafting, is kind of hilarious and also gross. I don’t need my language model to hand me a tissue and tell me my plot hole is “a really brave choice.” Just roast my outline so I can fix it. The whole point of these tools is that they don’t have social anxiety. Let’s maybe keep them brutally honest, yeah? :)
A Dark-Money Campaign Is Paying Influencers to Frame Chinese AI as a Threat
Tech bros who want us to believe AI is going to save democracy are literally paying lifestyle influencers five grand a pop to panic-post about China while they make breakfast for the kids. (I’m sorry, what?) This isn’t about national security, it’s about keeping regulations off their backs and the money flowing. It’s bonkers to me that OpenAI and Palantir affiliates are bankrolling literal propaganda, then acting shocked when people call the industry unregulated and creepy.
Authors already know the discourse around AI is completely unhinged, but this is next level. The same companies crying about doomer narratives are just… buying influencers to push China panic? Gross. If you’re making decisions about your career based on what these companies are feeding the algorithm, maybe don’t. Mochi could craft a more honest social media strategy, and she’s a cat.
Sunday Rundown #139: Extra Connectors & Fire Water Burn
Live-testing AI cover art tools instead of just screaming about prompts into the void? Thank god. Most advice for authors stops at “type this magic sentence,” so watching someone actually rank text-to-image makers on specific dimensions is way more useful than another theoretical guide.
I couldn’t see the actual winners from the snippet, but the leaderboard approach itself is solid. Authors need to know which tool spits out a usable fantasy cover and which one gives your protagonist seven fingers and a melting face. (We’ve all been there, usually at 2 AM.) Results matter more than hype when you’re budgeting for a launch.
7 Claude Design Guidelines To Instantly Improve Your Visual Assets
Write With AI (Nicolas Cole & Dickie Bush)
Claude Design actually teaching people to talk like designers instead of just begging for “pretty” and hoping the algorithm gets it? Super into that. I’ve always said AI covers are good enough, but the gap between “good enough” and “great” is usually the human knowing what to ask for. Learning terms like measure and density is lowkey the cheat code. :)
Thing is, authors already wear every hat. Marketer, formatter, cover project manager… or whatever. This is the first tool that doesn’t expect us to become designers, but also doesn’t let us stay clueless. Use the vocab, build the brand, stop looking like default Mailchimp. Pass.
GPT-5.5 Instant: smarter, clearer, and more personalized
ChatGPT finally kicked its emoji addiction and learned to get to the point. GPT-5.5 Instant slashed hallucinations by like 50% on messy prompts, so maybe I can trust it when I ask about… I dunno, Victorian funeral customs without it making up a law that never existed. The tighter answers are the real win, though. Less bloated formatting, less verbal fluff, just the info.
And those memory sources? Cute. I can see why it thinks I’m a night owl who runs on coffee. (I am. It’s not wrong.) Less babysitting the bot, more words on the page. Finally.
New ways to buy ChatGPT ads
ChatGPT ads opening up to self-serve? Okay, now we’re cooking. The CPM-only pilot felt like a velvet rope keeping indie authors out of the club, but CPC bidding opens the door for real. People use ChatGPT to compare options and figure out what to do next, which means your book cover could pop up while someone is literally asking what they should read. That’s not passive scrolling, that’s a reader with intent.
The measurement tools are getting beefier too, so you can actually see if your ad spend is translating to sales or just vanishing into the void. I’m not saying dump your whole Amazon budget here, but if you’re not at least testing this… what are you even doing? New ad platforms are cheapest when nobody else has figured them out yet.
I’m off to see if GPT-5.5 can handle my messiest research questions without going full hallucination mode. Mochi’s already judging my current cover draft from atop the printer. As usual. :)