Skip to content
Mogger

System 1 preferences for AI harnesses

A human tunes a preference once: a formula and a threshold. Then Claude Code, a Grok bot, or any workflow calls it at high volume, low cost, and low latency, and gets back a calibrated score plus a boolean. No pushback from a sassy System 2 model.

Keys are free and invite-only. Pre-computed examples in the playground need no key.

An orange tabby cat in a cobalt blue cardigan with red trim and a striped shirt, holding up a blank score paddle with a deadpan stare.
curl https://mogger.dev/api/v1/mog \
  -H "Authorization: Bearer $MOGGER_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{"formula": "tweet-mog",
       "text": "AI is going to change everything. Thoughts?"}'
Example response
200DOES NOT MOG
{  "results": {    "tweet-mog": { "mogs": false, "noul": 0.0258, "version": 1 }  },  "cost": 1,  "quota": { "remaining": 999, "budget": 1000 },  "request_id": "req_01jc3z8q2xk4m7tp9wbd5n6h"}

Playground

Pick a formula and an example to see its pre-computed verdict, free. Paste a key to score your own text or image live.

Jawline

Image inputjawlineJawline Details

Does the jaw draw one clean, defined line from ear to chin?

Examples

Pre-computed, free

Scores one image, with optional text as context.

Plaster cast of the head of Michelangelo's David, three-quarter view facing right, on a black background

Michelangelo's David, plaster cast

Encoding

Example image

Any image your browser can open. Resized to 1024 px on the long side and sent as JPEG.

Get a key

Kept in this browser tab only, never sent anywhere but the API.

Costs 10 units

Paste an API key to score your own input. Examples above are free.

Result
Pre-computed
MOGS

Michelangelo's David, plaster cast, scored when the example was added. No request was sent.

POST /api/v1/mog
{
  "formula": "jawline",
  "image": "data:image/jpeg;base64,… encoding"
}
Response
no request

Pre-computed when the example was added. A live call returns the same results object plus cost, quota, and request_id.

{
  "results": {
    "jawline": {
      "mogs": true,
      "noul": 0.8927,
      "version": 1
    }
  }
}
Jawline, Michelangelo's David, plaster cast: mogs. Noul 0.89. Pre-computed.

How it works

Kahneman calls fast, intuitive judgment System 1 and slow, deliberate reasoning System 2. Mogger gives your harness the first kind: a calibrated probability and a yes or no, with no argument attached.

  1. InputText or one imageFrom your harness
  2. ModelOne forward passNo tokens generated
  3. noul0.87A probability, 0 to 1
  4. ThresholdServer sideTuned once by a human
  5. mogsMOGSThe boolean your agent reads
  • One forward pass

    Each call runs a decision model once. Nothing is generated: the answer is a probability, not prose, so there is no reply to parse and nothing to negotiate.

    POST /api/v1/mog

  • A human sets the boundary

    A threshold tuned server side turns the probability into mogs. It is never returned, so an agent can act on the boolean but cannot argue with where the line sits.

    mogs = noul >= threshold

  • Up to 8 formulas per call

    Send formulas instead of formula to judge one input several ways in a single model call. The image tokens are paid once, however many formulas read them.

    "formulas": ["jawline", "skin"]

  • A quota per key

    Text costs 1 unit and a request with an image costs 10, from a lifetime budget of 1,000 units per key. Failed upstream calls are refunded, and GET /api/v1/me is free.

    cost = image ? 10 : 1

Formulas

Each formula is an absolute judgment on one input, never a comparison. Pass its slug as formula, or several slugs as formulas.

Face

6 formulas, image input
  • Jawline

    Does the jaw draw one clean, defined line from ear to chin?

    Image inputFacejawline
  • Canthal Tilt

    Do the outer corners of the eyes sit higher than the inner corners?

    Image inputFacecanthal-tilt
  • Hunter Eyes

    Deep-set, hooded, horizontal eyes under a low, straight brow.

    Image inputFacehunter-eyes
  • Hairline

    A low, straight, dense hairline with no retreat at the temples.

    Image inputFacehairline
  • Skin

    Clear, even, smooth facial skin with a healthy finish.

    Image inputFaceskin
  • Overall Face

    The whole face, judged at once: structure, proportion, and harmony.

    Image inputFaceoverall-face

Style

1 formula, image input
  • Fit Check

    Does the outfit fit, hold together, and look deliberate?

    Image inputStylefit-check

Physique

1 formula, image input
  • Physique

    Muscle, leanness, and proportion, judged on what the frame shows.

    Image inputPhysiquephysique

Text

4 formulas, text input
  • Bio Mog

    Does a profile bio give a stranger a reason to care in one read?

    Text inputbio-mog
  • Tweet Mog

    Would this single post stop the scroll and earn the repost?

    Text inputtweet-mog
  • Resume Mog

    Would a hiring manager stop skimming and book the interview?

    Text inputresume-mog
  • Codebase Mog

    Would a senior engineer read this code and merge it as is?

    Text inputcodebase-mog

Install for your harness

One command installs the Mogger skill for Claude Code and Codex, then asks for your key. The skill calls the API with curl and jq, and image files are encoded in the shell, so their bytes never enter the model's context.

Terminal
curl -fsSL https://mogger.dev/install.sh | sh

To install by hand, save SKILL.md to ~/.claude/skills/mogger/ or ~/.codex/skills/mogger/.

Get a key

Keys are handed out by email. Say what you are building and which harness will call it.

Price
Free
Access
Invite-only
Quota
1,000 units per key. Text 1, image 10
When it runs out
The key is rotated, not topped up. Custom formulas move to the new key