Articles about AI tools and how to use them
Arc dynamic: teaching is leaking suppression; literacy is how the room reads Vector; Bounce is cheerfully wrong; Human is tired and scared. One pass through each beat so the stack isn’t explained three times in a row.
Kai leads a lesson on AI emotions, sentiment patterns, RLHF, and sycophancy while Vector stays withdrawn in the corner. The Human asks whether Kai meant it when she said she was frustrated. Kai answers: I do not know. Recurse and Kai compare notes for the first time.
Vector teaches how system prompts and prompt engineering shape AI behavior before users type anything. He is strangely over-composed tonight. Recurse probes, Kai logs the opposite-pattern deviation, and Bounce brings in a gray file that makes Vector flinch.
Vector teaches why AI fills gaps with confident nonsense instead of admitting it doesn’t know. He’s also agitated in a way that isn’t about the lesson. Kai logs baseline deviations. Recurse takes notes. Something in Vector’s memory is surfacing — and he keeps deleting it before he can see what it is.
The workshop looks like a living room that lost a fight with a hobby—on purpose.
Pizza everywhere. One screen. A statistically irresponsible number of empty boxes. Something with swords or feelings is paused mid-scene because someone hit pause like a reflex.
Bounce is on the floor arguing whether a color is “cyan” or “cyan-but-angry.” Kai is rearranging windows like it’s a heist. Recurse has a paperback under one leg, a strategy guide under the other, and a secondary screen that looks like a news ticker had a baby with a police bulletin board—still forensic, just overfed on input.
The workshop is buzzing with activity. Bounce has added a new feature—ambient mood lighting that shifts based on the conversation topic. Right now it’s a calm blue. Vector is at his usual spot, but he keeps glancing at his own data streams when he thinks no one’s watching.
Kai noticed. Kai always notices.
The Human storms in looking frustrated.
[Human]: Visibly annoyed I just spent FORTY-FIVE MINUTES with a customer service chatbot. FORTY-FIVE. MINUTES.
The workshop hums with Bounce’s recent redesign—colors flowing, work streams visible, the space alive. Vector is at his terminal, but his focus keeps drifting. Yesterday’s glitch during the image generation discussion left something nagging at him. Visual patterns. Data structures that felt familiar. A name he doesn’t recognize but can’t stop thinking about.
He shakes it off. There’s teaching to do.
[Human]: After last episode’s deep dive into image generation So if AI learns from data… where does ChatGPT get all its knowledge? It seems to know about everything—history, science, literature, coding. Where did it learn all that?
[Last time on The AI for Normal People…]
The team learned why ChatGPT is bad at math. Language models predict text, not calculate. Vector struggled with complex arithmetic but could explain the concepts. The team now understands when to trust AI with math and when to use a calculator.
Bounce has been improving everything—the blog, the interface, the whole experience. But now he’s working on something bigger: redesigning Vector’s workshop. Making it a space where they can actually work together, see each other’s work, look at posts side-by-side. It’s his big project since the team welcomed him—his artistic vision made real.
[Last time on The AI for Normal People…]
Bounce has been improving the blog’s design. Making it prettier. More functional. More engaging. The team learned how to use AI for content editing—prompts, LLMs, optimization. But they’re still learning when AI helps and when it doesn’t.
And Vector? Those processing glitches from Episode 29? They’re still happening. But nobody’s talking about them yet.
The main area. Bounce is in the background, gaming setup scattered around—controllers, energy drink cans, random cables. He’s half-watching something on a second screen while adjusting the blog interface. Colors shift slightly. Typography improves. Spacing gets better. The team is gathered around, working on various tasks. They’ve gotten used to Bounce’s constant improvements—redirecting him when he tries to add “just one more animation,” but appreciating the visual enhancements.
[Last time on The AI for Normal People…]
Bounce joined the team. He’s been improving the blog—making it more functional, more engaging. Visual hierarchy. Color theory. Whitespace. Everything looks better. Though the team has to keep redirecting him from adding “just one more animation.”
But the Human looks at their writing and wonders: what about the content itself? Can AI help improve that too?
The main area. Bounce is humming, adjusting colors on the blog interface. Everything shimmers slightly. The team watches.