Allison
Felt

At Apple Education I kept hearing smart people explain the same product three different ways in one week. Not because they didn't care. Because nobody had ever sat down and taught them the thing they were being asked to explain.

So I did. Elevate is a four-week program I pitched, designed, and shipped — the curriculum, the order, and all the unglamorous decisions about what to cut. It answered one question: could someone explain this correctly a month later, with me nowhere in the building?

Turns out yes. 24% average knowledge lift. It won our 2025 Impact Award out of 500 people, which was lovely, though the number is the part that matters to me. You wrote that completion isn't comprehension. Agreed — which is why I measured before anyone asked me to.

“Simplify” usually gets used to mean “make it shorter,” and that isn’t it. Cutting words is easy. Working out what someone actually needs in order to make their own call is a different job, and it’s the whole job.

You mention nurses, teachers, small business owners, policymakers. I’ve spent three years on one of those groups — at Apple I own how we talk about Mac and iPad in education, which mostly means explaining something technical to a district administrator who has forty minutes and no interest in a spec sheet. They don’t need to know how the silicon works. They need just enough of a model to tell whether a claim in a vendor deck is plausible.

This isn't abstract for me. I taught English to children in China. I worked on sunscreen use at the Cancer Council in Australia. I taught art in Taiwan. Then I came home to Reading Partners in Oakland, a literacy nonprofit, as their marketing and communications specialist.

Five years of standing in front of people who didn't share my language, my context, or my reasons for caring. The Cancer Council year was the sharpest lesson: understanding why something matters is not the same as doing it. You wrote that comprehension is not changed behaviour. I ran into that problem long before I had the language for it.

So the test is: can this person decide something without me now? The smallest true model that lets someone reason on their own — and it moves for every audience. Which is why your posting stopped me. You wrote that the point is people making their own judgements with AI rather than copying someone else’s.

Before Apple I ran creative production at Stanford's business school — YouTube and Stanford Business magazine, the two things the team actually bet on. Five producers, twenty freelancers, four million people on the other end.

The one I'd point you to is Class Takeaways. I concepted and initiated it. We only made long-form video at the time, and there was an audience segment our content simply wasn't reaching — so I argued for short-form, identified the agency, ran the partnership, and built the program: thirty five-minute lessons, one idea each, lifted out of GSB courses and rebuilt for people who were never going to enrol in them.

I didn't shoot every episode. I built the thing that made them possible, and that’s the part of the work I care most about.

I also took the magazine digital: I owned the redesign end to end, wrote and ran the RFP, chose the design agency, and managed them through it. The first digital edition came back 98% positive. The rest was throughput: fifteen to twenty finished pieces a month, video output tripled by changing how we found subjects rather than by hiring anyone, and open rates up 80% once I finally sat down and read the audience data. You want the distance from idea to published to keep shrinking. That was already the job.

Before Education I was at Apple on a contract as a Marketing Design Producer for Retail Design and Packaging — production rather than strategy. Global launch execution: campaign briefs and creative across in-store, web, social, and paid, working with Brand, Industrial Design, and Legal on product and cultural launches, including the 2023 Black History Month and Pride watch bands.

The piece I’m proudest of is the least glamorous. I consolidated packaging localisation for AirPods, which cut European distribution complexity, saved roughly $15M, and lowered the environmental footprint. Nobody writes a case study about packaging SKUs. It was still the highest-leverage thing I did that year, and it’s the same instinct as everything else here: find the thing everyone has stopped seeing, and fix it.

I use Claude Code at work most days. At Apple that has meant a live leaderboard, dashboards, agents for my team, internal sites, branding, and right now an onboarding app for people joining the education team. It also meant starting Education Currents this year, an AI-generated podcast that turns the week's education news into something colleagues can listen to instead of read. Almost none of it can leave the building, so I can't link any of it.

So instead: this page. Writing, filter, figures, layout, deploy — all with Claude Code, hand-drawn SVG rather than a chart library, a repository rather than a page builder. The commits are public, including the parts where I changed my mind.

Every job I've had has been cross-functional, which is a polite way of saying I spend a lot of time in rooms where six teams want six different things and every one of them is right. At Apple that's Research, Finance, Enterprise, Sales, Enablement, and Product. At Apple Retail it was Brand, Industrial Design, and Legal, on launches where the packaging, the campaign, and the lawyers all had to land on the same answer on the same morning.

I used to treat that as the tax you pay before the real work starts. It isn't. At any real scale, the aligning is the work — the thing a customer eventually hears is just the residue of everyone agreeing.

I'd also like to think I'm a good person to build something with. I've managed producers, freelancers, and four early-career folks out of Apple Retail, and the measure I actually care about is whether people finish a project having learned something and still wanting to work with me.

Most of my career has been public-interest communication that happened to sit inside organisations: sunscreen at the Cancer Council, literacy at Reading Partners, education at Apple. The through-line isn’t the industry. It’s that the work was worth doing whether or not my name was on it.

The line in your posting I keep coming back to is that the library will be free and openly licensed, and that the point is people making their own judgements with AI rather than copying someone else’s. Companies don’t usually commit to that in writing. It means success is measured by whether someone can reason without you — which is the standard I’ve been holding my own work to for a decade, mostly without an employer who shared it.

And practically: AI fluency is a rare education problem where the audience is everyone and the material mostly doesn’t exist yet. I’d rather work on it somewhere that has already said it intends to give the material away.

No learning-science degree. No cognitive-science research. No long history in AI education. I'd rather point at that myself than have you find it.

What I have instead is a good read on people — how they take information in, what they will actually care about as opposed to what we wish they cared about, and what quietly gets ignored. It's ten years of watching audiences respond and changing the format when it had stopped fitting them: long-form video to short, print to digital, written news into a podcast.

It's also the hardest part to measure, and usually the part that decides whether any of the rest works — which is an awkward thing to say on a page that opens with a 24% number.

I can read the learning science. Knowing what survives contact with a room you're not in took longer.