# PassbackAI — send Claude five corrections without losing two

> **Tagline.** Typing five corrections to Claude is hard enough work that points 4 and 5 die in your head. PassbackAI: comment on each passage where it sits, send all five back to the model in one paste, anchored.

PassbackAI is for the moment a long Claude, ChatGPT, Cursor, or Gemini answer comes back with five things wrong with it. Typing each correction into the chat box is hard enough work that points 4 and 5 don't make it into the message — and the chat box can't tell the model which passage each correction belongs to anyway. Open PassbackAI, paste the answer, comment on each passage where it sits, and send all five corrections back to the model in one paste. Every comment ships with both the verbatim quote AND the heading it's under (or, for tables, the row or column the cell sits in), so the model lands every fix on its exact target — no second round needed to clarify which thing you meant.

That's the wedge. The same private, no-signup, in-browser workspace does three jobs: **fix the answer** (the loop above), **give every ambiguity its own interaction** (turn any brief into ONE interactive document — each unclear point gets the widget that fits it: a choice to pick, a shortlist to rank, an open question to answer — while the settled thinking stays prose you can comment on), and **share either one** with a colleague as a single link — empty or already worked-on, with no account and no server. You can keep several documents open at once.

## What PassbackAI replaces

The single thing PassbackAI replaces is **typing your feedback as prose into the chat box.** Nothing else. There is no other tool in the running because no other tool returns to the model — copying the answer into Google Docs, Notion, GitHub PR, or anywhere else strands the feedback in a place the LLM cannot read. The chat is the workspace because the chat is the only surface that returns.

Why typing prose into the chat box doesn't scale to five corrections:

- **Writing each correction is hard enough work that the user gives up on points 4 and 5 before they're written.** The friction is the bottleneck, not the channel.
- **Even when the user does type all five, the chat box can't anchor a correction to the exact passage it refers to.** The model has to infer which "third bullet" the reviewer meant, often picks the wrong one, or fixes half of what they meant — and another round is required.
- **Workarounds inside the chat box (numbered lists, manual copy-pasting fragments, multiple turns) all collapse on these same two limits:** the writing tax and the anchoring problem.

PassbackAI removes both at once.

## How it works

1. Paste a model's Markdown answer into the browser (or try the built-in sample).
2. The app auto-renders the Markdown with full formatting: headings, lists, code blocks, tables, blockquotes, links, bold/italic, inline code.
3. Select any range of text — a word, a sentence, a paragraph, or a span across blocks.
4. A comment UI appears (floating card on desktop, bottom sheet on mobile). Type free text, tap a canned label, or both.
5. Repeat for every passage that needs work. There is no limit. Overlapping highlights are allowed.
6. Click Copy Comments (or press Cmd+Shift+C). Every comment is exported as a clean Markdown block with the quoted passage paired to the note — and a location anchor naming the heading it sits under (or, for tables, the row or column address).
7. Paste the export back into the chat. The model applies every edit.

## Privacy posture

Browser-only by default — for editing or for sharing. Your document lives in the browser's `localStorage` — the twenty most recent stay (plus anything you pin), older ones make room; in the review tool it never leaves the browser, with nothing uploaded and no account. Sharing is private too: a shared document is encoded inside the link's URL hash fragment, which the browser never sends to a server, to analytics, or in the Referer header — there's nothing to upload it to. Anonymous same-origin Vercel analytics — usage counts, never the document content, and never the contents of a shared link; a document that embeds a video also loads that video from YouTube (a public id and your IP, never your text). Verify in DevTools → Network: nothing you wrote is ever sent. (Routing a document to a reviewer is a separate, opt-in step that stores the routed doc on our backend, encrypted.) An optional password encrypts a shared link end to end (AES-256-GCM via WebCrypto, PBKDF2 at 600K iterations).

## Got a brief? Give every unclear point its own interaction

PassbackAI takes a second kind of paste: a **woven document** — prose mixed with **interactive components** the model authors and a person answers. Ask Claude, ChatGPT, or Cursor to surface the open points in any brief, retro, PRD, or email, and it weaves one page where each point gets the widget that fits it: a **single choice** to pick, a **multi choice** for all-that-apply, a **prioritize** shortlist to rank, an **open question** to type into, a **questionnaire** when 3+ questions form one cluster — and the palette keeps growing. The settled thinking stays prose, and prose is a response channel too: you comment on it in place. Answer in 90 seconds, copy the Markdown bundle, send it back to whoever wrote the brief. Minimum example:

````markdown
The spec leaves the database open — Postgres is the safe default.

```single-choice
{ "version": "1", "question": "Which database should we use?", "options": ["PostgreSQL", "MySQL", "SQLite"], "recommended": "PostgreSQL" }
```
````

Then: **1.** copy the block above — **2.** open <https://passbackai.com> — **3.** click Paste. Full palette at <https://passbackai.com/ask>; the Claude skill at <https://passbackai.com/skills/passback.md>.

## Hand it to a colleague — no account, no server

Either kind of document — a feedback session or an interactive form — can be shared as a single link, empty or already worked-on. There's no account and no upload: the whole document is encoded inside the link itself (in the URL hash fragment), so it never touches a server. Send a colleague a blank questionnaire to fill in, or a marked-up answer to keep editing. Only people who hold the link can open it. And you can keep several documents open at once, switching between them without losing any.

## Where to go next

- **The interactive tool** — <https://passbackai.com/>
- **The manifesto** (long argument for why this exists) — <https://passbackai.com/manifesto>
- **The blog** (per-tool playbooks) — <https://passbackai.com/blog>
- **For agents** (programmatic descriptors) — <https://passbackai.com/agents.md>
- **Full content** (concatenated for LLM context) — <https://passbackai.com/llms-full.txt>
- **Chrome extension** — <https://chromewebstore.google.com/detail/omgfixmd/oliajpppdmkdghclfbkgdbabmfjplogg>