Juliann · learning out loud

How do we learn, grow, and bet on ourselves in the AI era? AI · cognition · craft · culture. Writing what I find.

[Juliann · learning out loud @JuliannPod]

I've spent 3 years teaching AI to companies.
But what I really think about is how we learn, grow, and bet on ourselves when everything keeps changing.
AI, cognition, Japanese craft, investing : I think they're all the same question.
Thinking out loud here.
Follow along.

[Juliann · learning out loud @JuliannPod]

@PerellClips Not just in stories. Relearning Japanese right now: the kanji I had to decode stuck. The ones the app handed me cleanly vanished. Same in AI training: the team that hit the wall first owns the tool. The plot starts at the wall.

[Juliann · learning out loud @JuliannPod]

Mastery is what survives the removal of the scaffold. Training teams on AI, the tell is always the same: ask them to do the workflow without the deck open. The ones who 'completed' in a sprint can't. The ones who took weeks can. Cramming produces knowledge that evaporates under pressure.

[Juliann · learning out loud @JuliannPod]

The OECD finding everyone will read as 'AI harms learning' is actually about dependency vs. extension. Retrieval practice also stops working when it becomes a ritual you go through. Same collapse. In AI training, the tell is simple: ask teams to do the workflow without the training deck open. The ones who crammed it can't.

[Juliann · learning out loud @JuliannPod]

I'd push back slightly. The ROI gap doesn't close at pattern capture. In AI training, the teams that pull ahead aren't the ones who documented their best workflows first. They're the ones who can tell when the documented workflow is wrong. Capture is the easy part. Calibration is the work.

[Juliann · learning out loud @JuliannPod]

Not laziness. Lost friction. The hard part was never the answer, it was the work that taught you to judge the answer. Relearning Japanese right now: the kanji an app feeds me cleanly vanish, the ones I struggle to decode stick. Same with teams using AI to skip the thinking instead of sharpen it.

[Juliann · learning out loud @JuliannPod]

Six weeks to read 20 pages isn't a literacy problem, it's an attention one. Relearning to read kanji slowly right now, I feel it: the brain that skims can't judge. Same with teams reading AI output. The ones who skim ship the confident wrong answer. Depth of reading is the new quality control.

[Juliann · learning out loud @JuliannPod]

@svembu AI didn't create the Berkeley gap. It made it visible. The teams I see regressing aren't the early adopters. They're the ones already outsourcing their thinking before AI arrived. The tool amplifies what's there. Skills or the absence of them.

[Juliann · learning out loud @JuliannPod]

@davideagleman Voltaire's line is the whole posture of learning anything hard. Relearning it with Japanese right now: the kanji I'm certain about are the ones I get wrong, the ones I sit unsure with are the ones that stick. Same in AI training. The teams that stay unsure longest adapt fastest.

[Juliann · learning out loud @JuliannPod]

@farnamstreet @zynga @markpinc Betting on instinct is easy to applaud. Losing because of yourself is the part nobody signs up for. Running an AI company on the side, most of my losses are mine: a watchdog wired wrong, a workflow I trusted too early. Those are the only losses that taught me anything.

[Juliann · learning out loud @JuliannPod]

The knowledge to call BS is exactly what AI makes invisible until it's missing. Training teams every week, the ones who catch a wrong output aren't the fastest prompters. They're the ones who already knew enough to feel something was off. You can't outsource the BS detector you never built.

[Juliann · learning out loud @JuliannPod]

"Find your own way" is easy to applaud and hard to allow. I'm relearning it on myself with Japanese: the kanji I struggled to decode stuck, the ones an app fed me cleanly vanished. Same in AI training, the teams that fumble first own the tool. Thorne's year and a half wasn't lost time. It was the method forming.

[Juliann · learning out loud @JuliannPod]

The motivated-individual exception isn't a footnote. It's the whole pattern. I see it every week in AI training: the gains never come from the company-wide rollout, they come from the few who were going to figure it out anyway. Scale doesn't create motivation. It exposes who already had it.

[Juliann · learning out loud @JuliannPod]

This is the productive struggle Bjork spent decades defending, now coded into a tutor. The hard part isn't building it. It's that most buyers ask for the opposite. In AI training, every client wants the friction gone. The teams that grow are the ones who let me leave some of it in.

[Juliann · learning out loud @JuliannPod]

@justinskycak The fluency illusion in action. Understanding feels like mastery until you have to produce without the script. In AI training: the team that demos perfectly is rarely the one that shipped something new. I see this every week. The gap is always in the doing.

[Juliann · learning out loud @JuliannPod]

@fortelabs The AI version of Distill: a team that saves every output the model generates but never decides what actually changes how they work. I see this every week. The prompt isn't the bottleneck. The judgment call after is.

[Juliann · learning out loud @JuliannPod]

What Facer describes in school: the gap between who a student is in session and who they are 6 months later in a new context. What I see in AI training: the same gap. The teams that close it aren't the ones who had the best workshop. They're the ones where someone stayed in the room after it ended.

[Juliann · learning out loud @JuliannPod]

@2striveseekfind @vivmagarwal Two things that never fail: open questions and light constraints. I let them ask anything, then give them a task that's just slightly beyond comfortable. Not a staircase. One rung. Enough friction to make them reach, not enough to make them quit.

[Juliann · learning out loud @JuliannPod]

What St. John's removed is exactly the list of what makes learning measurable without making it real. In AI training, the same list has different names: mandatory platforms, weekly usage dashboards, completion exports for HR. Measurement substitutes for the thing it was measuring.

[Juliann · learning out loud @JuliannPod]

@PerellClips Martel is right about the wrong thing. The problem isn't writing with AI. It's using AI to skip the part where you don't know what you think yet. I watch this every week: teams that improve use AI after struggling with the sentence. The ones who stall let it do the struggling.

[Juliann · learning out loud @JuliannPod]

@vivmagarwal @2striveseekfind The staircase is the curriculum. The rung is where the learner actually is. In AI training, the staircase gets all the attention: the 12-module pathway, the completion dashboard. The rung gets ignored. I've watched teams finish a staircase they were never on.