ChatGPT's Jujutsu Kaisen Cast: Six Picks from the Current Session
Six Jujutsu Kaisen castings proposed by ChatGPT in the current session, pairing playful fictional names with primary-source context about each person's role and research.
Lab notebook
The research, decisions, and practical lessons behind each build.
RSS feed 
Claude's Jujutsu Kaisen Cast: Six Picks Led by Role Labels
Claude's six Jujutsu Kaisen castings, five fictional Domain names, Arthur Mensch's lone exception, and the exchange that shaped the response.
Gemini's Jujutsu Kaisen Cast: Facts and Metaphors for Six AI Figures
Gemini casts six AI industry figures in Jujutsu Kaisen, pairing four playful fictional names with primary-source context about their roles and research.
Comparing Gemini, ChatGPT, and Claude's Jujutsu Kaisen Casts
Compare 10 people across Gemini, ChatGPT, and Claude: two unanimous pairings, two 2–1 splits, and candidates who appeared in only one model's list.
Storytelling Basics for Marketing Reels
Plan a marketing Reel with one goal, a real constraint, a forced choice, and visible proof. Use evidence-bounded 20- and 30-second beat sheets.
How to Edit AI Drafts: Fix Evidence Before Style
A practical review order for AI-assisted drafts: verify claims and citations first, then fix copied artifacts, generic filler, voice, and formatting.
When Blog Gates Pass but Prose Craft Drifts
A snapshot audit of 11 posts found existing gates passed while ungated prose craft varied; a proposed review gate remains pending.
18 HyperFrames Instagram Reel Hook Templates to Download
A gallery of 18 original AI Maker Lab HyperFrames HTML compositions based on external, unmeasured English, Brazilian Portuguese, and Japanese Reel hook formulas.
Instagram/Facebook Reels Retention Benchmarks
Compare Reels watch time, percentage watched, completion, paid scope, and a three-post case without treating external averages as universal targets.
Culture and Identity in the 2010s: Three Source Sets
A bounded comparison of 12 source records from Brazil, the United States, and Canada shows recurring tensions around participatory culture, platforms, race and representation, and identity.
Society and Trend Research: A Five-Country Source Map
Browse 60 directly inspected, source-linked records from Brazil, the United States, the United Kingdom, Canada, and Japan by era, category, voice, and evidence boundary.
My 17-track AI soundtrack failed—what I’ll change
Hear all 17 instrumental tracks from my local MiniMax Music 3 batch, why I will not keep them, and the caption, iteration, and QC changes planned for my next attempt.
Funko to Filament: Can AI Turn a Portrait Into a 3D Print?
A research log on image-to-3D AI: what the tools export, why a generated mesh is rarely printable, and three ways to print a figurine.
HyperFrames Video as Code: How the Render Loop Works
HyperFrames explained: its HTML-to-video render loop, transition patterns, community skills, source demos, templates, and Remotion boundary.
How to adapt Instagram Reel hooks for English, Brazilian Portuguese, and Japanese
Adapt Reel hooks in English, Brazilian Portuguese, and Japanese, then align the cover, frame zero, and first three seconds around one evidence-bounded promise.
How Instagram Reel scripts change across English, Portuguese, and Japanese
A research digest of Reel script tutorials in three languages: the shared hook-value-CTA skeleton, what each market's playbooks add, and why each technique works.
Learning in order, and seeing what I actually retain
The problem that starts a new build: my list of things to learn grows faster than any course covers it, and nothing shows me what to learn first or what stuck.
The learning methods evidence supports — and the ones it doesn't
An evidence-led guide to active recall, spaced repetition, lower-utility study habits, learning maps, and honest ways to represent mastery over time.
A learned gene editor: can a model mutate NEAT better than chance?
Tests whether a trained edit scorer can beat random NEAT mutations, with fresh seeds and a protocol fixed before the first run.
Should a genome keep what its network learned? Three ways to mix gradients into evolution
Three hybrid trainers ask whether evolution should discard learned weights, inherit them, or pool experience in one gradient learner.
Learning to read the level: replacing lucky jumps with a policy gradient
A failed champion sets the test: can a policy read unseen Skyline Run levels instead of replaying one lucky route?
Procedural districts: making Skyline Run test the policy, not the route
A measured generator makes every Skyline Run district clearable, separates training from test levels, and exposes route memorization.
Can a deterministic game teach a network to run the rooftops?
A deterministic engine becomes a browser AI lab—and measurement catches the environment changing before training begins.
Teaching the trainer: evolving the reward instead of tuning it
Evolve better rewards automatically: AutoRL searches training settings; PBT, or population-based training, copies and perturbs the best runs.
The findability problem: the navigation rules I had to unlearn
How a missing header became a measurable navigation repair—and why visible paths mattered more than seven-item or three-click rules.
The memorization trap: when a champion learned one district
The rebuilt Skyline AI lab learned under fixed budgets, but 0/630 generated-world clears show that broader training still did not generalize.
The jump that measurement corrected
Building the platformer's first playable slice: the measured jump values, and the level design that only measurement could fix.
Learning to Make a 2D Platformer: Where I Started
What I learned about platformer styles, movement, cameras, and level design—and how those pieces helped me find a manageable starting point.
I Wanted to Make a Game. First, I Had to Learn Its Language.
What I learned about game screens, feedback, and the path from an idea to something playable—with a small tic-tac-toe game you can try.