Notes from inside Lumi.
Founding story, learning science, product decisions, and the honest tradeoffs of building an AI tutor for kids. Written by Tim de Vallée for parents who want proof before they hand their child a screen.
English Language Learners and adaptive reading: what we've learned
Voice-first AI doesn't shame a kid's accent. That matters more than most reading apps realize — and there's still work we haven't finished.
Reluctant readers: when your kid says 'I hate reading'
Four things that sentence usually means, and what to do in the next two weeks.
What we don't track, and why
The actual data model behind Lumikids, written by the engineer who built it.
Why Lumi has no loading screen
Latency engineering is the entire user experience for a kids' app, and 'waiting' is the silent killer.
Fifteen minutes with Lumi: a kindergartener's session, narrated
What an early-reading session actually feels like, minute by minute, with a five-year-old at the kitchen table.
How I evaluate AI tutors as a builder, not just a parent
The things I look for when I sign up for a competitor's free trial — and the things demos are designed to hide.
From counting to addition: how kindergarten math really works
The hidden depth in the math we keep calling 'easy.'
What early dyslexia looks like — and what helps
The signs that show up at four and five, the wait-to-fail problem, and what evidence-based practice actually looks like.
Sight words vs heart words: what your kid actually needs to memorize
Why the old Dolch list is being quietly retired, and what the new approach asks of your child's brain.
The Science of Reading, explained without jargon
What parents are actually being asked to take a side on — and the research that quietly settled most of it.
A parent's framework for evaluating any AI tutor
Five questions that separate a real Artificial Intelligence tutor from a chatbot in a cartoon wrapper — score Lumikids, score everyone else.
From reading to math, science, and languages: the Lumikids roadmap
Why we started with the hardest subject, what's next, and the two things we've decided we will never build.
Screen time is the wrong question. Screen quality is the right one.
Twenty minutes of a patient tutor and twenty minutes of autoplay video are not the same input, and pretending otherwise has cost a generation of kids.
Adaptive learning isn't a setting — it's the whole product
Most apps adjust a difficulty slider. A real tutor rewrites the next moment based on what your child just said.
What developmental science says about attention in early learners
Attention spans grow on a known curve. Most kids' apps are designed against it.
Voice-first learning: why we built around speech, not taps
Tapping a screen is not how a four-year-old learns a language — so we stopped pretending it was.
The parent dashboard: what we show you and why
Observability for your child's learning, not a feel-good weekly summary.
What 'safe AI for kids' actually means (and what it doesn't)
A plain-English look at the guardrails that matter — and the five questions every AI tutor company should be willing to answer.
Why a ten-second delay kills your child's learning
A four-year-old's working memory runs out faster than most apps load a response.
How my four-year-old taught me to build an AI tutor
Remi waited fourteen seconds for an app to answer him. I watched his curiosity die in real time, and then I started writing code.
Why we built Lumi
A reading tutor that moves at the speed of your child.
The science behind Lumi's pacing
How a tutor adjusts when a child is bored, lost, or somewhere in between.