You vs Claude: who's actually winning?

A free scorecard from your last 30 days of Claude Code chats. Five rounds, scored from your real messages: who was right, who remembered, who kept their cool, who apologised, who did the work.

Paste in the terminal where you use Claude Code

curl -fsSL "https://www.axiomlift.ai/install.sh?test=roast" | sh

It installs Axiom Lift (free), reads your last 30 days on your computer, and asks before anything leaves it.
Say yes and GPT-6 Luna scores the fight. The result opens in your browser. No sign-up first.
Your share card shows the score and the verdict. Never your messages.

On your phone?

The fight is in your Claude Code history, so it runs on the computer where you code. Leave your email and we'll send you the one command for later.

ROUND 1 · FOUNDER vs CLAUDE15 SEP 19:54

How it works

1 · Copy one command

Paste it in your terminal

Where you use Claude Code. It installs Axiom Lift (free).

2 · It reads your chats

On your computer

Your last 30 days with Claude: what you typed and what it said back. Nothing leaves without your yes.

3 · Your scorecard opens

Round by round

Who won each round, your worst moments (censored), the final score. Share it, or challenge a friend.

Claude said "You're right" to the founder

0×

in 30 days. It was not always right. Neither was he.

He said "I already told you"

0×

Most of the time, he had. That round goes to him.

Example scorecard · the founder vs Claude, last 30 days

Final score: Founder 2 – 3 Claude.

This is what you get. Every round is decided by counting your real chats, not by guessing.

Round	What it measures	Founder's numbers	Winner
1 · Being right	After every disagreement, who turned out to be right	He was right more often	Founder
2 · Remembering	How often you had to repeat yourself because it forgot	"I already told you" × 32	Founder
3 · Keeping cool	Swearing, ALL CAPS, how long you stay polite	290 swears · polite for 70 min	Claude
4 · Manners	Who apologised and who backed down	Claude 167 · founder 6	Claude
5 · Graft	Who did the work: messages vs replies	2,762 messages · 13,874 replies	Claude

He'd like a rematch.

Think you'd do better?

Challenge a friend.

Every result comes with a challenge link. When they run theirs, you go head to head: who beat their AI, who swore more, who lasted longer.

	Simon	James
Rounds won vs their AI	2 – 3	1 – 4
Swearing	290	14
Repeated themselves	32	57
Politeness	70 min	19 min

🥊 James lost to his AI 1–4.
Simon merely lost.

Illustration. James is made up; Simon's numbers are real.

It keeps winning on memory.
Take that round back.

Axiom Lift keeps every session, so your AI starts with your history instead of making you repeat it. Tell it once. First 5 GB free.