A year ago, on a winter day like this, Arlyn Gajilan’s 11-year-old son, Tobey, came home from school in low spirits. “Everyone keeps telling me I’m smart, but I’m just so slow,” he said.
“It crushes you as a parent,” Gajilan shared. A veteran journalist, she was used to complex problems, but nothing prepared her for this. Her journey to jump-start her son’s confidence ultimately led to Tobey’s Tutor, an AI platform for kids with dyslexia, attention-deficit/hyperactivity disorder (ADHD) and other learning differences.
The origin story
That journey had its start during the pandemic, when Tobey first flagged in school as falling behind in reading. An evaluation revealed he was dyslexic — information that was “scary but helpful,” Gajilan said. She and her partner transferred Tobey to a school specializing in his learning differences.
They were confident they could support him, and at first things seemed to go well. But last February, his confidence began to crumble. Around the same time, Gajilan was working with custom versions of the artificial intelligence bot ChatGPT in her role as global editor for AI development and integration at Reuters. She built such a GPT, as the customized bots are known, to help Tobey with reading fluency, spelling and with organizing his writing.
Gajilan fed the program Tobey’s report cards, neuropsych evaluation, hobbies and information about his competitive nature, a step that ensured that his lessons were gamified. After each session, she asked the bot, “Why this choice?” or “What is he learning?”
Within weeks, Tobey was gaining confidence. The GPT had been programmed to encourage progress even when he struggled — something that tests and other traditional measures of achievement often fail to do. One day, Tobey asked: Could she build something for his friends?
“Famous last words,” Gajilan said.
From GPT experiment to business
Gajilan noticed parallels between her work helping Tobey and what she was doing for colleagues asking how AI could support them. She didn’t know how to code but had a journalist’s instinct: Ask the right questions, find reliable sources and center people’s needs.
When so-called “vibe coding” emerged — when AI models write code to build tools users dream up but couldn’t make by themselves — she drafted the architecture of the platform that became Tobey’s Tutor. During spring break in Florida, she built the first version at the kitchen table. By late June, a beta launched with eight paying subscribers, some of them strangers.
She spent the summer consulting research papers, parents, educators and the learning-differences community, iterating quickly. By September, she had rebuilt the platform into today’s Tobey’s Tutor. The platform now has about two dozen subscribers.
Not a human replacement
“Human tutoring and teaching is absolutely 100% the gold standard,” Gajilan said. “Do it, get it, benefit from it.” But with tutoring costing upwards of $150 an hour, her goal was to provide an affordable tool to bridge resource gaps.
For parents concerned about AI, Gajilan empathizes: her son, now 12, doesn’t have a phone or social media and his internet access is tightly monitored. It’s also why Tobey’s Tutor lessons last only 20 to 30 minutes.
The challenges of bootstrapping
As a one-person, self-funded operation, early responses were slow. Sometimes the AI suggested solutions requiring “10 engineers and half a million dollars,” Gajilan said. She’d remind it: “It’s just me here with my laptop. Let’s be realistic.”
Like choosing which news stories to pursue, she prioritized features based on Tobey’s experience. “I don’t want to make it sound like it was all sunshine and roses and unicorns and rainbows,” she said. There were late nights cursing at her computer and lingering doubts.
An insomniac, Gajilan was used to getting little sleep. But building the platform demanded more: every weekend and every available moment outside her full-time leadership job in a large newsroom. Her wife, Robin, kept family life running.
During one of Gajilan’s coding marathons, Tobey brought her food and designed a scavenger hunt in the house in which the hidden objects spelled “L-O-V-E”.
“This is amazing,” she recalls thinking. “Now let me go build this thing to deserve this little person.”
When the student became the teacher
Tobey wasn’t just her inspiration; he became the platform’s best critic. He flagged confusing lessons and praised helpful ones. And his learning extended beyond academics — into helping his mom cope with road rage.
“If you’re a New Yorker, you will know that when you’re on the West Side Highway and you get cut off, your first impulse is not to say, ‘Oh, please go ahead,’” Gajilan said. When she cursed another driver, from the backseat came Tobey’s guidance: “Take a deep breath. Calm down. It will get better.”
She realized he had internalized lessons from the AI tool — a “meta” moment, she said: “This whole tool you designed, helping him build those tools to then repeat back to you.”
Tips for other entrepreneurs
Gajilan isn’t necessarily leaving her newsroom role, she said — she loves journalism and Reuters.
“But I also feel like we’re building a thing that could have lasting impact for a community of kids who aren’t always the focus of technology, who aren’t the focus of entrepreneurs or focus of large tech companies,” she said. “This has the potential to level the playing field for them.”
Her advice for other entrepreneurs:
- Leverage AI to build with a one-person team: “AI is a democratizing technology where you don’t need a team of developers or software engineers to build meaningful products.”
- Build resilience — something Gajilan is both teaching Tobey and knows from covering tech start-ups as a reporter: “Entrepreneurs fail a lot. People who try to do something different will fail often.”
- Keep your mission in sight: For Gajilan, that was recognizing that many kids in NYC and across the country lack the low teacher-to-student ratio her son enjoyed. Even then, “if you’re surrounded by a resource-rich environment, but those resources are not benefiting your child, it doesn’t feel like those resources exist.”
For families: How does Tobey’s Tutor work?
- Families complete a questionnaire and can upload report cards or IEPs, the individualized plans developed for special education students by their schools and parents.
- The AI builds a personalized learning profile, drawing from more than 450 publicly available studies, academic papers, worksheets and curricula to create fully customized lessons.
- The system sets goals and delivers lessons through a chat interface with speech-to-text and text-to-speech so kids can talk naturally with the tutor as they progress.
- A newly launched homework-support feature helps turn assignments into teaching moments rather than simply providing answers; according to Gajilan, the AI’s guidance is aligned with each student’s existing lessons and goals.
