graphitiTurning something visual into math, or math into something visual, is not a new impulse — it's most of the history of graphing. What's changed over time is which direction the conversion runs, and how much of the work a machine can do without a person filling in the formula first. Here's a general, careful look at that lineage, without pretending to know more precise dates or names than we actually do.
Before computers
Plotting by hand, point by point
Long before any of this was automated, the basic technique was manual: take a table of values, plot each pair as a point on graph paper, and connect them to see the shape a relationship traced out. It's a slow, foundational habit — and it only runs in one direction. You need the equation, or at least the data, before you get the picture.
Mid-to-late 20th century
The graphing calculator puts the plot on a screen
Handheld graphing calculators brought that same by-hand plotting process onto a small screen, letting you type in a formula and watch the device draw the curve instantly instead of plotting it point by point yourself. The first graphing calculators are generally placed in the mid-1980s, with wider adoption in classrooms following over the next several years. Worth noting: this is still equation-to-picture, the same direction as graph paper — just automated. The calculator has no idea what shape you meant to draw; you have to already know the formula and type it in.
Late 20th century onward
Curve-fitting software flips the direction
As computers became common in labs and offices, software for curve fitting — spreadsheet trendlines, and more specialized tools built for scientists and engineers — made the reverse trip possible on a wide scale: start with scattered data points, and let the software find the equation that best explains them. This is the direction Graphiti works in, and it's covered in more detail in how curve fitting actually works.
Late 20th–early 21st century
Symbolic regression removes the last assumption
Ordinary curve-fitting software still needed a person to pick the shape of the equation in advance. Researchers working on symbolic regression — including foundational work on evolving equations with genetic-programming-style search in the early 1990s, and later systems built in the 2000s that could examine raw experimental data and reconstruct known physical relationships without being told the formula ahead of time — pushed the automation one level deeper: let the software search for the shape of the formula too, not just its numbers. That's the technique covered in symbolic regression, explained for actual humans, and it's the core of what Graphiti runs on your sketch.
2010s
Photo-to-math apps solve an adjacent, different problem
Smartphone apps that let you point a camera at a printed or handwritten math problem and get a step-by-step solution — Photomath is the best-known example, launching in the mid-2010s — became genuinely popular in this decade. It's worth being precise about what these apps actually do: they read symbols someone already wrote (an equation, an arithmetic problem) using text and symbol recognition, then solve or simplify what's already there. They're not looking at a freeform drawn shape and asking "what formula does this resemble" — the equation is already given to them in written form. That's a related but distinct problem from the one curve fitting and symbolic regression solve.
Where a hand-drawn shape (not a written equation) fits into this
Line all of these up and there's a gap that's easy to miss: graphing calculators go from formula to picture. Curve-fitting software goes from data to formula. Photo-to-math apps go from an already-written equation to a solved answer. None of them start from a shape with no formula and no equation attached — a heart doodled in a notebook margin, a spiral drawn absentmindedly on a call — and ask "what's the equation for this."
That's the specific spot Graphiti sits in: a hand-drawn shape as the input, with symbolic regression doing the work of figuring out both what family of curve it resembles and what its numbers are, with no formula supplied by the person drawing it. It's less a brand-new idea than the next reasonable step after the ones above — pointing the same curve-fitting and symbolic regression techniques researchers have used on scientific data at something anyone can do with a finger on a phone screen.
Is Graphiti the same thing as a photo-math app?
No. Photo-to-math apps read an equation or problem you (or a textbook) already wrote out, then solve it. Graphiti does the opposite: you draw a shape with no equation attached, and it finds one that fits.
Where to go from here
For the two techniques mentioned above in detail, read how curve fitting actually works and symbolic regression, explained for actual humans. To see what the shape-to-equation trip actually produces, see what is a mathgram, or jump straight to how it works and the FAQ.