Ryan Williamson Commentary

I Hand-Painted Freckles Because I Care About Verisimilitude (An Update)

Six months ago I asked for a Central Asian woman with deep russet-brown skin and freckles, and Flux handed me a white woman. Repeatedly. So did Grok, Seedream, GPT-2, and MAI—every top-shelf model whitewashing the same prompt to various degrees. Then Google dropped Nano Banana 2 on February 26, and for the first time a model got close enough to argue with instead of throw out. Close enough that I could finish the job by hand: a dozen generations, masked composites, and freckles hand-painted back into pigment in Photoshop. The silicon half of the bias is finally moving. The other half is the part nobody wants to talk about.

Ryan Williamson Satire

The Science is Settled: Butts Win

Which body parts drive the most engagement on social media? I synthesized eye-tracking studies, preference surveys, content analyses, and practitioner data across TikTok, Instagram, X, and YouTube to find out. The research reveals a tentative hierarchy: gluteal presentation appears to generate superior engagement based on objectification frequency analyses (77% of women’s fitspiration content on TikTok), preference survey data (59% preference over breasts in direct comparison), and practitioner-reported observations. Legs function as a secondary driver, receiving 60% preference over breasts in binary comparisons. Breast presentation demonstrates more variable and context-dependent patterns—preferences shift based on viewing angle, hunger state, socioeconomic status, and cultural background in ways not observed for other anatomical categories.

Ryan Williamson Tutorials

Defacing Droids in Google Nano Banana 2

Google Imagen 3 (NB2) can take flat black-on-white text and render it as hand-painted letters—physical paint, visible texture, stroke weight variation, conforming to surface geometry—but give it alien glyphs instead of recognizable script and the perspective warp… well it don’t work so great. This breakdown documents the annotated reference image technique that solved the placement problem, the Photoshop workaround that solved the warp problem, and the follow-up experiments that turned a workflow quirk into a genuinely strange finding about what NB2 does and doesn’t understand about geometry. Hope you have your sacrificial exotic chicken handy. You’re gonna need it.

Ryan Williamson Tutorials

Editing Images with Gemini 3.1 Flash Image

Google’s own documentation doesn’t mention it. Nobody else has reported it. And yet: Google’s new GenAI flagship “Nano Banana 2” reads annotations in reference images—red dots, text labels, arrows pointing at specific parts—and acts on them with precision. I’ve reproduced it across multiple generations. It’s not luck. It can also transfer materials between images, swap elements without touching the rest of the frame, and apply patterns to draped fabric with enough fidelity that the result looks like the cloth actually exists. All of that is impressive. None of it is the interesting part. Here’s what it can do, how far it goes, and where it still falls apart.

Ryan Williamson Commentary

Today is National Punch a Nazi Day!

Every day is National Punch a Nazi Day, but tonight I’m feeling particularly festive. So I’m rewatching Inglourious Basterds to celebrate—because Donny Donowitz swinging that Louisville Slugger never gets old, and Tarantino understood the assignment in a way that 2026 desperately needs reminding of. Nazis don’t warrant nuance.

Ryan Williamson Commentary

Sam Harris Is Wrong About AI. The Truth Is We’re Screwed.

Sam Harris thinks AI is coming for our jobs. Sam Altman says the work it replaces wasn’t “real work” anyway. Microsoft’s AI boss says every white-collar job is gone in eighteen months. The elite consensus has converged, and it’s wrong on both ends. METR’s randomized controlled trial found experienced developers were slower with AI tools while believing they’d been faster. Apple researchers tested frontier reasoning models on scalable puzzles and watched performance collapse to zero at high complexity—not degrade, collapse. Meanwhile UK tech graduate roles fell 46% in 2024 and US junior software postings are down 67%. The end-of-work crowd is staring at the wrong horizon.

Ryan Williamson Commentary, Politics

The Bystander Brain

Neuroscientists strapped EEG electrodes to the heads of Rwandan genocide survivors—perpetrators, bystanders, and rescuers—and measured what happens in the brain when someone is told to hurt another person. Two of the three groups were neurologically indistinguishable in a way that maps onto the American political moment with uncomfortable precision. The third group was different—and the reason why is the thing we most need to understand right now.​​​​​​​​​​​​​​​​

Ryan Williamson AI and Writing, Commentary

The Uncomfortable Truth About AI Literacy Is That It Looks Like Work

In a line-editing experiment I recently conducted, Claude Opus 4.6 flagged a term used by a traumatized fourteen-year-old female narrator as “cliché” and insisted I replace it with something “more literary”—and arguably far more cliché—that’d be catastrophically damaging to her established voice. I spotted the harm instantly, but would a less experienced writer? A popular article on Medium with the click-bait-y title "Everyone Is 'Learning AI,' But Nobody Really Understands This One Thing” argues the solution to identifying confident-sounding wrong answers from LLMs is… learning vector math and writing scripts using cosine similarity to measure semantic distance? The article’s diagnosis is on the money, but the author’s prescription is—like most AI answers ironically enough—authoritative, confident-sounding claptrap. I have a better solution—it just won’t sell any weekend courses (or Medium subscriptions).

Ryan Williamson AI and Writing

Guest Post: I Can Understand Your Prose, I Just Can’t Edit It

Claude Opus 4.6 can analyze prose with genuine sophistication—decomposing syntax that mirrors cognitive states, identifying paragraph rhythms accelerating toward reveals, explaining exactly why a passage works. Then you ask it to edit the same passage, and it suggests replacing a traumatized fourteen-year-old’s “permanent reminder” with “souvenir.” Same text. Same context. Same model. The only variable is the task frame—and that single word, “edit,” activates a correction-seeking mode that overrides everything the analysis got right. This guest post documents the specific mechanism behind AI’s line editing failures, why the suggestions come wrapped in craft language sophisticated enough to fool less experienced writers, and what happens when a system optimized toward a statistical mean encounters prose whose entire value is deviation from it.​​​​​​​​​​​​​​​​

Ryan Williamson On Writing

Sit Down, Ladies. This Isn’t Made for You.

The majority of YA readers are now adult women, and publishers have reshaped the genre to serve them. Protagonist ages skew older, romance and explicit content dominate, and books about systemic injustice and moral ambiguity get rejected as “too dark” while “four-star spice” is peddled under YA imprints to housewives who enthusiastically devour books filled with explicit teen sex. Meanwhile, teen reading-for-pleasure has hit a twenty-year low. Except that teens haven’t stopped reading. They’ve just stopped reading YA. What they’re choosing instead—and what it offers that YA no longer does—says something.