miatz™
Log inSit Chuun
Free concept playground · no account

The Jargon Decoder

Paste any AI text. Get plain English and the receipts.

All 90 playgrounds Join the free cohort

The Jargon Decoder

AvailableFree

Paste any AI text — a vendor pitch, a press release, a research abstract. It spots the jargon, rewrites each term in plain English, and scores how much the copy is hiding behind vocabulary.

Plain-English rewrite
Our AI agent (a model that takes actions in a loop) uses RAG (look-it-up-first) over a vector database (a store that searches by meaning), with fine-tuning (extra training on your own examples) and quantization (shrinking the model to lower precision) to cut latency (how long you wait for a reply). Each prompt (the instructions you send) fits the context window (how much text it can read at once); temperature (the randomness dial) is tuned to reduce hallucination (a confident made-up answer) during inference (running the model to get an answer), and an embedding (text turned into a list of numbers) index speeds token (a chunk of text, roughly three-quarters of a word) lookups.
35.1fog score · dense fog
13terms decoded
37words
Every buzzword, with receiptsthe fog score is the jargon, counted

This copy leaned on 13 AI terms across 37 words — a fog score of 35.1 (dense fog). Here is what each one actually means, term by term — no hand-waving:

  • agenta model that takes actions in a loop

    A system where the model plans, calls tools, reads the results, and repeats until a goal is met — not one reply.

  • raglook-it-up-first

    Retrieval-Augmented Generation: fetch relevant documents first, then answer from them instead of from memory.

  • vector databasea store that searches by meaning

    A database that indexes embeddings so you can find the nearest vectors — the retrieval half of RAG.

  • fine-tuningextra training on your own examples

    Continuing a base model's training on a task-specific dataset so it adopts your tone, format, or domain.

  • quantizationshrinking the model to lower precision

    Storing weights in fewer bits (e.g. 4-bit) so a model fits smaller hardware — a little accuracy for a lot of memory.

  • latencyhow long you wait for a reply

    The delay between sending a request and getting the response — often dominated by time-to-first-token.

  • promptthe instructions you send

    The full text handed to the model for one call — system rules, any context, and your actual question.

  • context windowhow much text it can read at once

    The maximum number of tokens a model can consider in one call — your prompt plus its reply must fit inside it.

  • temperaturethe randomness dial

    A setting from 0 upward: low makes output focused and repeatable, high makes it varied and more creative.

  • hallucinationa confident made-up answer

    When a model states something false as fact — fluent, plausible, and wrong, because it predicts words, not truth.

  • inferencerunning the model to get an answer

    Generating output from an already-trained model, as opposed to training it. Every request you send is inference.

  • embeddingtext turned into a list of numbers

    A vector that places text in a space where nearby points mean similar things — the basis of semantic search.

  • tokena chunk of text, roughly three-quarters of a word

    The unit a model reads and bills by — words are split into sub-word pieces before the model ever sees them.

The trick the industry plays is to sound expert by hiding behind vocabulary — the more a page hides, the higher its fog score climbs. The decoder does the opposite: it shows the receipts instead of asserting authority.

Was this playground useful?

The Jargon Decoder

Paste any AI text — a vendor pitch, a press release, a research abstract — and the decoder spots the jargon, rewrites each term inline in plain English, and gives you a fog score: how much the copy was hiding behind vocabulary versus saying something concrete. Every decoded term ships a one-line receipt, so instead of asserting expertise the tool shows its work — term by term. The higher the fog score, the more the writing leaned on buzzwords. Runs entirely in your browser (0 uploads, works offline).

How to use it

  1. Paste any AI-related text into the box — a competitor's press release, a vendor slide, a research abstract.
  2. Read the plain-English rewrite on the right: each known term is wrapped inline with what it actually means.
  3. Check the fog score — the share of the copy that is jargon — and the term-by-term receipts in the amber panel.
  4. Hit Load sample for a deliberately foggy example, or Clear and paste your own.

What this clears up (the fundamentals)

  • Jargon is not the same as substance — a sentence can be dense with AI terms and still say almost nothing concrete. The fog score makes that visible as a number.
  • Every term has a plain meaning — "embedding" is text turned into numbers; "RAG" is look-it-up-first; "quantization" is shrinking the model to lower precision. None of it needs to be mysterious.
  • Multi-word terms count as one idea — "context window" and "vector database" are single concepts, not four scattered words. The decoder spots them whole.
  • Whole-word matching avoids false hits — "tokens" is not flagged as "token", and "agentic" is not flagged as "agent"; only real, standalone terms are decoded.
  • Receipts beat authority — the honest move is to show what each word means, not to sound expert by piling on vocabulary.

Where it's used

A free, top-of-funnel learning widget for anyone trying to read AI marketing, papers, or docs without a glossary open in another tab. It is a miatz build-lab concept playable — a teaching widget you can play with here and then learn to build yourself in the lab, wiring a term-spotter, a glossary corpus, and a fog-score metric.

FAQ

What is the "fog score"?

It is the jargon density: the number of distinct AI terms found divided by the total word count, as a percentage. A higher score means more of the text is vocabulary rather than concrete claims — a quick read on how much a page is hiding behind buzzwords.

Why did a word I expected not get decoded?

The decoder matches whole words only, so "tokens" will not match "token" and "agentic" will not match "agent". This avoids false positives where a term appears as part of a larger, unrelated word. It also only knows its built-in glossary of core AI terms.

Does it change or rewrite my whole text?

No. It leaves your words exactly as you pasted them and only wraps each recognised term inline with its plain-English meaning, preserving your original casing. Nothing is paraphrased or removed.

Is anything uploaded?

No. The text is matched against the glossary and scored entirely in your browser — nothing is transmitted, stored, or logged. Turn off your Wi-Fi and it still works.

Limits

The glossary is a fixed set of core AI terms, so newer or niche jargon may slip through, and the fog score is a rough signal — not a verdict on whether a claim is true. It measures how much a text leans on vocabulary, not whether the underlying idea is sound. The plain-English rewrites are teaching-grade summaries, not exhaustive definitions.

Related

Part of the Demystify Playgrounds. Explore the rest from the Playgrounds home.

Bookmark this page (Ctrl+D, or ⌘D on Mac) — it works offline the next time you need it.

More playgrounds

Ninety playgrounds. Zero setup.

Every concept here is playable free, no account — and inside the program you learn to rebuild the machinery yourself.

Browse all playgrounds Join free