Synthesis Tutor calls itself, right there in the page title, “the world’s first superhuman math tutor.” That’s not marketing shorthand somebody exaggerated in an ad — it’s the actual tagline on their website today.
It’s a big claim. So when a claim is that big, the useful question isn’t “does that sound impressive?” It’s “what’s the evidence, exactly, and does it hold up?”
Synthesis actually makes this easy. They wrote a blog post called “Does the Synthesis Tutor get results?” and it names their source directly: a DARPA-funded program called the Digital Tutor. Follow that thread, and here’s what you find.
What the DARPA study actually was
The Digital Tutor was real, and by the numbers reported, it was genuinely impressive — for what it was designed to do. DARPA wanted to see how close a piece of software could get new recruits to the expertise of a seasoned professional, fast. So they built a tutoring system and tested it on U.S. Navy sailors training to become Information Systems Technicians — the people who keep a ship’s IT systems running.
Over 16 weeks, those sailors went through the Digital Tutor program. At the end, they were tested against two other groups: Fleet technicians with an average of ten years on the job, and sailors trained the traditional classroom way. The Digital Tutor group outperformed both, by a wide margin, on troubleshooting real IT systems.
That’s a strong result. It’s also, if you look at the actual DTIC records, from assessments run in 2010 — sixteen years ago now — on adult Navy technicians learning enterprise IT systems. Not elementary students. Not math. Not children at all.
Where the gap is
None of this means Synthesis Tutor is bad, or that the company is lying. They’re transparent about their source — they link straight to the DARPA documentation and a public summary of it. That’s more citation than most ed-tech marketing bothers with.
But there’s a real distance between “a 2010 program that helped Navy sailors master IT troubleshooting faster than classroom training” and “the world’s first superhuman math tutor” for a 6-year-old learning to subtract. The first is a specific, well-documented result in a narrow, adult, technical-skills domain. The second is a sweeping claim about a completely different subject, age group, and learning context — resting on that same study as its evidence.
That gap is the actual lesson here, and it’s a useful one to practice noticing — for you and for your kids.
A framework for checking claims like this
This is exactly the kind of moment where the third piece of metacognition — Evaluate — earns its keep. Monitor asks “do I understand this?” Regulate asks “what should I do differently?” Evaluate asks the question most of us skip: “was that reasoning actually sound?”
Applied to a marketing claim, Evaluate looks like three quick questions:
- What’s the actual claim? (“Superhuman” — a specific comparative claim, not just enthusiasm.)
- What’s the cited evidence? (A study — good, that’s better than nothing.)
- Does the evidence match the population and the product? (Adult Navy IT trainees, 2010, technical troubleshooting — vs. a math app for a 7-year-old, in 2026.)
That third question is where most impressive-sounding claims quietly fall apart. It’s not about catching companies in lies — it’s about training the habit of checking whether the receipt matches the bill. That’s a skill worth modeling out loud with your kids the next time an app, a toy, or a tutor promises something “revolutionary.”
The takeaway
Citing a source is good. Citing the right source is what actually matters. Before any tool — ours included — earns a claim like “this works,” it should be able to show you evidence that was actually about the kids, the subject, and the context you’re deciding for. If it can’t, that’s worth knowing before you buy in.
Auxesis doesn’t sell superhuman anything. We teach the slower thing that actually works: struggle, checked understanding, and frameworks kids can use for the rest of their lives — including this one.
Leave a Reply