Working with PI
PI is how you work with PIPress. You ask for things in plain language and PI does them: searches the Claim network, reads your papers, explains what it found, writes documents. Most of what the platform can do is reachable by asking for it.
Open it at /pi. It needs an account.
The workspace
- The box at the bottom says Ask PI anything, or type / for commands. Type a question and press Enter.
- To its left is a model picker. PI offers several AI models; the pick applies to your chats and to the documents PI writes for you, and usage is metered on the model you chose. Some models are marked as locked on your current plan.
- Step by step changes how document orders are handled. With it on, asking for a review, proposal, paper or any other document starts a short interview: PI asks you a few questions first, you answer, and the document is written last. It is off by default and resets when you leave the page.
- The paperclip opens the attach menu: Upload file (PI reads it and answers about it in this chat) or Get claims from file (also extracts the paper's research claims into your profile).
- Three starter cards and three suggested prompts get you going: today's briefing, exploring a research question, building a review or paper.
Replies cite Claims. Tap any Claim title to open it in the reading panel next to the chat. Every conversation is kept in the left rail and can be renamed, searched or deleted. Deleting a conversation does not delete the documents PI wrote in it.
Where PI's answers come from
PI draws on three bodies of material and always tells you which one:
| Source | What it is |
|---|---|
| The Claim network | Everything PIPress has published. The default for any subject question. |
| Your library | The papers and findings you brought in. See Your library. |
| The open web | Only with your say-so. |
If a question finds nothing in the Claim network, PI looks in your own library next, and then offers to search the web. A web search is a confirmation away: PI asks first, because web findings are stored in your library and cost usage. When an outside search runs, the reply is headed with what was found: so many from the web, so many in the Claim network.
Terms that mean something specific here (lightning, nostr, taproot…) are read the platform's way if you follow the matching domain; otherwise PI asks which meaning you want.
Commands
Type / to see the palette. The most useful ones:
| Command | What it does |
|---|---|
/briefing | Your personalized top developments, right now |
/map | Where your field runs deep, where it is thin, what has changed |
/inspire | Developments outside your field that run close to your research |
/opportunities | Openings you could pursue: your capabilities plus the field's gaps |
/search <what to look for> | Search the Claim network |
/web <what to look for> | Search the web and academic sources; findings join your library |
/explain | Why the open Claim is ranked the way it is for you |
/prefs | Show or change the areas you follow and your depth |
/alert <what to watch, daily or weekly> | A standing watch: PI emails you what's new on it. See Research alerts. |
/works, /claims | Your papers, questions and connections; your own claims |
/upload, /upload-claim | Attach a file; extract research claims from a file |
/import <paper link or DOI> | Add a paper to your library by link |
/link | Connect an ORCID or Google Scholar profile |
/review, /brief, /report, /proposal, /paper, /patent, /submission, /post | Write a document. See Writing with PI. |
/chart, /draw | A data chart or a diagram |
/code-to-article | Turn this chat's code into an article |
You never need a command. "What changed in my field this week?" does the same as /briefing in plainer words.
Usage and limits
Everything PI does for you is metered against your plan's weekly allowance: chats, searches, document writing, uploads. The Usage this week meter in the composer's options, and Settings → Usage, show one percentage: the most-used of your ceilings. It warns at 75% and refuses at 100%, with a message saying when it resets: every Monday. The allowance is paced weekly so one heavy day cannot spend a whole month.
If you would rather not wait, you can top up a wallet balance and keep going pay-as-you-go. See Settings and billing.
Two more limits protect everyone: at most ten messages a minute, and a platform-wide daily ceiling on model calls. Hitting either simply asks you to try again shortly.