> ## Documentation Index
> Fetch the complete documentation index at: https://docs.puddin.ai/llms.txt
> Use this file to discover all available pages before exploring further.
# Writing Process Evidence: What Puddin Captures and How
> Explore the authorship evidence Puddin captures — typing activity, paste events, revisions, writing speed, pauses, and session history.
Puddin builds a picture of authorship from the writing process itself, not from the finished text. Every assignment produces a structured evidence record made up of distinct signal types, each capturing a different dimension of how the document was created. Understanding what each signal means — and how to read it in context — helps you make well-grounded assessments.
This page explains each evidence type in depth, including what Puddin captures, how it is displayed, and how to interpret what you see.
## What Puddin captures
Puddin records every character entered via the keyboard — letters, punctuation, spaces, and control characters such as backspace and delete. This produces a complete, character-level record of direct typing activity across the entire writing session.
## How it is displayed
In the **Activity** panel, typing activity is visualised as a density chart over time. Taller bars indicate periods of higher keystroke frequency (faster or more sustained typing); gaps represent pauses or session breaks. The timeline view also marks typing bursts as distinct blue segments.
## How to interpret it
Direct typing activity is the baseline signal of organic writing. A document built primarily through keyboard input, with natural variation in speed and frequent small corrections, is consistent with a student composing at the keyboard.
Consider the following when reviewing keystroke patterns:
* **Consistent density with natural variation** is typical. Most writers speed up and slow down as they think, draft, and revise.
* **Very short total typing time relative to word count** may indicate that a significant portion of the document arrived via paste rather than direct entry. Cross-reference with the Paste Events panel.
* **Isolated bursts with long preceding pauses** can reflect a student who paused to think before writing a passage — or may warrant further context from the student.
Keystroke data tells you *how* the document was entered. It does not tell you what the student was thinking during pauses, whether they referred to notes, or the context of their writing process. Use it alongside other signals.
## What Puddin captures
Every time content is pasted into the writing environment — via keyboard shortcut, right-click menu, or any equivalent method — Puddin records:
* The **full content** of the pasted text
* The **position** in the document where the paste occurred (character offset)
* The **timestamp** of the paste
* The **session** in which the paste occurred
## How it is displayed
Paste events appear in the **Paste Events** panel as a numbered list, ordered chronologically. Each entry shows the pasted content in full, the approximate document location (shown as a section anchor or character offset), and the time. In Session Replay, paste events render in amber so they are visually distinct from typed text.
## How to interpret it
Pasting is a normal part of academic writing — students paste their own notes, quoted material, references, and self-written drafts. The presence of paste events is not inherently significant; the content and context are.
When reviewing paste events, consider:
* **What was pasted?** Short passages such as citations, URLs, or copied headings are routine. Long, complete paragraphs pasted at the start of a session warrant closer attention.
* **When was it pasted?** Paste events early in a session before any typing has occurred may indicate pre-written material was brought in. Pastes late in a session may reflect a student incorporating notes they had open alongside their writing.
* **How much of the document is pasted content?** The Process Summary shows total pasted characters as a proportion of the final word count. A very high proportion (for example, 80% or more) is worth reviewing.
Puddin captures pasted content but cannot determine its source. You will need to assess whether pasted material is appropriately attributed, self-sourced, or otherwise consistent with the assignment requirements.
## What Puddin captures
Puddin records all deletions and rewrites — any instance where previously typed text was removed and replaced. This includes:
* Backspace and delete key presses (character-by-character deletion)
* Selection-and-replace actions (highlighting text and typing over it)
* Cut operations (text removed via Ctrl+X or equivalent)
* Undo and redo sequences
## How it is displayed
The **Revisions** panel displays a chronological log of revision events, each showing the deleted text, the replacement text (if any), and the timestamp. In Session Replay, deletions appear as strikethrough characters before disappearing, making revision sequences easy to observe in context.
## How to interpret it
Revision activity is one of the strongest indicators of active, engaged composition. Writers who are genuinely working through ideas on the page tend to produce a rich revision record — false starts, rephrased sentences, restructured paragraphs, and refined word choices.
What to look for:
* **Frequent small revisions** (single words or short phrases changed) are characteristic of careful writing. This is a normal and positive signal.
* **Large deletions followed by large pastes** may indicate a section was removed and replaced wholesale with pre-written material.
* **Very few or no revisions** in a long document can be a point of interest — very few writers produce final-quality prose in a single unrevised pass.
* **Undo/redo sequences** are visible in the revision log and can sometimes reveal experimentation with alternative phrasings.
## What Puddin captures
Puddin calculates writing speed continuously throughout each session, expressed as words per minute (WPM) over time. This is derived from the keystroke record and represents the pace of direct typed input only — pauses and paste events are excluded from the WPM calculation but are visible alongside it.
## How it is displayed
Writing speed is shown as a line graph in the **Activity** panel, with time on the horizontal axis and WPM on the vertical. The graph includes an average WPM line for reference. You can hover over any point on the graph to see the exact speed and the corresponding passage being written at that moment.
## How to interpret it
Writing speed varies naturally between writers and within a single session. There is no universal "suspicious" speed threshold — context matters.
Useful reference points:
* **Average adult typing speed** is roughly 40–60 WPM. Students writing academically, pausing to think, tend to write slower — often 10–30 WPM when measured as effective composition speed.
* **Sudden speed spikes** (e.g. jumping from 15 WPM to 90 WPM for a sustained passage) may indicate a shift from composed writing to fast transcription of memorised or pre-prepared text.
* **Consistently high speed throughout** is unusual for sustained academic writing and may merit review alongside paste events and revision data.
* **Slow passages** often correspond to complex argumentation, unfamiliar vocabulary, or careful sentence construction — all consistent with genuine engagement.
## What Puddin captures
Puddin distinguishes between two types of inactivity:
* **Pauses** — gaps in keystroke activity within a session, from 2 minutes up to the session timeout threshold. These are captured with their duration and the position in the document at which writing stopped.
* **Session breaks** — gaps where the student closed the writing environment entirely and returned later to continue. Each break is recorded with its start and end time and the total duration away.
## How it is displayed
Pauses appear as grey markers on the timeline and in the Activity panel. Session breaks appear as distinct separators between session blocks in the Sessions panel and in the timeline. Hovering over either shows the duration and the document state at the start of the break.
## How to interpret it
Pauses are a natural feature of thoughtful writing. Extended pauses often precede complex passages — a student thinking through an argument before committing words to the page is a positive indicator of genuine engagement.
What to consider:
* **Short pauses (2–10 minutes)** throughout a session are entirely normal and usually indicate thinking, reading back, or referencing notes.
* **Long pauses (30 minutes or more) mid-session** within a single session may indicate the student stepped away. This is often benign but is worth noting if the writing before and after the pause differs markedly in style or quality.
* **Multiple short sessions spread across several days** is consistent with students working iteratively on a piece over time — a hallmark of considered academic work.
* **A single session with no pauses** for a lengthy document can be unusual, depending on the student and the task.
## What Puddin captures
For every assignment, Puddin records:
* The **start and end time** of each writing session
* The **word count at the end of each session** (and at regular intervals within sessions)
* The **sequence of events** (typing, pastes, deletions) that drove word count growth
* The **document state at any point in time**, reconstructable from the event log
## How it is displayed
The **Sessions** panel shows each writing session as a row with its start time, end time, duration, and the word count delta (how many words were added or removed during that session). The **Timeline** tab shows the full event record on a scrollable chronological view, and the Process Summary includes a word count growth chart showing how the document developed across the full writing period.
## How to interpret it
The session timeline tells the story of when and how much work was done over the course of the assignment. Document development patterns are a meaningful dimension of the authorship record.
Typical patterns to recognise:
* **Gradual growth across multiple sessions** — consistent with a student drafting, pausing, and returning over days or weeks.
* **Sudden large jumps in word count** at the start of a session — particularly if accompanied by high paste volume — may indicate pre-written material was brought into the document.
* **Document development close to the due date** is common and not inherently suspicious; many students write under time pressure. Consider whether the writing process evidence is consistent with the timeframe.
* **Very few sessions for a long document** — for example, a 3,000-word essay completed in a single 45-minute session with minimal revision — is a pattern worth examining in light of the full evidence record.