"Just for checking my work," he told himself at 2 AM, sipping cold coffee.

But Leo had anticipated that. Version 2.0 didn't just solve—it learned . When DeltaMath changed the variable phrasing or added a twist, the bot scraped the new patterns from public student posts and updated its solver within hours.

And Leo? He still uses the bot. But only after he's already filled two pages with his own shaky, human handwriting.

At 4:01 AM, someone asked: "How do I get my dad to stop drinking?"

It started innocently. A Python script that read the problem text from his screen, parsed the variables, and ran them through a reverse-engineered solver. Paste a DeltaMath problem, click "Fetch," and the bot would spit out the answer in under two seconds.

Leo was not a math prodigy. He was a logic prodigy. And to him, DeltaMath—the merciless, green-and-white online platform his school used—was not a test of algebra or calculus. It was a test of patience. Each problem was a tiny fortress: different numbers, same structure. Same attack pattern.