Exactly how your pie gets baked

The four steps

1. Read your posts, locally

   Pasted text or your data export is opened in your browser. Pasted posts are split on lines of ---. Exports are read from the platform's own files: CSV for LinkedIn and YouTube, JSON for Facebook, Instagram and TikTok, and the tweets.js file in an X archive.

2. Measure 12 habits

   Each feed has 12 habits: some only that site has, some shared by the social feeds. Each is a simple rule: a list of phrases, a line-length check, or a count.

3. Score each habit from 0 to 100

   The measurement goes through one curve. Each habit has a target rate that earns a 90.

4. Bake the pie

   Posts that trigger no habit at all become the slice. Everything else is "% serious," split among your top five habits.

The scoring curve

Every habit uses the same formula. It rises fast at first, then levels off, so a few matching posts move your score a lot and the last few points are hard to earn.

• rate is what was measured. For most habits that's the share of your posts that match, so 3 matching posts out of 10 is a rate of 0.3.
• target is the rate that earns a 90. Each habit's target is listed below.
• Hitting twice the target earns . Nobody quite reaches 100 except through rounding.

The habits

Each habit has a pie flavor, a rule, and a target. Phrase lists match regardless of capitalization. “Only on ” flavors are unique to this feed; “Social classic” flavors appear on several social feeds.

Reading the phrase lists: (a|b) means a or b. A ? makes the character before it optional, so honou?red matches honored and honoured. ^ means "at the very start."

Baking the pie

Worked example. Say someone scores the numbers below, and of their 10 triggered no habit at all. That makes them serious. Here's how the top five split that share:

Try a sentence

Type or paste any post to see which rules it triggers. This is the same code the quiz uses, and it runs only in your browser.

What it gets wrong

• It reads words, not meaning. "I am not humbled" still counts as humbled, and sarcasm is invisible to it.
• It's English-only. Posts in other languages mostly score as chill.
• Pasted posts are missing information. They don't show photos, reposts or likes, so those habits fall back on phrases and the slider. An export is more accurate.
• Export formats change. The readers for every feed except LinkedIn follow those companies' published export formats, which change from time to time. If an export won't read, pasting always works.
• Small batches swing hard. With 3 posts, one post is a third of everything. Ten or more is steadier.
• The phrase lists are opinions. They're meant to be funny and recognizable, not fair or scientific.
• The numbers are comedy, not statistics. The targets and thresholds were tuned to make results fun.

Privacy

• Your posts never leave your browser. The site never uploads or stores them, and it uses no cookies, accounts or analytics. Close the tab and they're gone.
• Your browser does download a few things to show the page: the site's own files from Cloudflare, which hosts it; fonts from Google Fonts; and two open-source libraries from cdnjs, JSZip to open the export zip and PapaParse to read its spreadsheets. Like any website, those services see a normal page request from your IP address. They never see your posts.
• The merch preview draws every design in your browser. When you open it from your results, your pie's numbers travel in the part of the link after the # sign, which browsers never send to any server. Your posts don't go along, only the pie.
• The merch poll is the one thing that talks to our server. When you vote, it sends which company you picked, and nothing else. To stop ballot-stuffing it keeps a scrambled, salted fingerprint of your connection for one day, never your IP address itself. The quiz pages never call it.
• You can check. scoring.js contains no network calls. Open your browser's developer tools on the quiz page, watch the Network tab while you score, and nothing gets sent.