Shimmer for researchers

A research partner that knows your world, stays current for you, and never invents a citation.

Most AI research tools do one thing and forget you the moment you close the tab. They search a corpus, write a report, and own none of your context: not your inbox, not your collaborators, not the papers you already trust, not what you are actually working on this month.

Shimmer is the other shape. It is a persistent personal agent that thinks with your knowledge base and the people who matter, comes to you with what changed, and grounds every answer in a source you can open. For a researcher, that combination is the whole point.

Why this fits research work

Three things matter most to researchers using AI, and they rarely come together:

• Grounding you can trust. The single biggest reason researchers distrust AI for literature work is invented citations: real authors, real journals, fake paper titles. Shimmer answers from your own knowledge base and connected sources, with the source attached. If it does not have a source, it does not assert the claim.
• Your private context. Your academic and personal email, your calendar, your collaborators and labs, the projects you are running. Shimmer reasons over your world, not a generic corpus.
• Proactivity. A research brief that comes to you, rather than a chat window you have to remember to open.

What a researcher gets today

These are wired now.

• Two inboxes, one agent. Connect a university Google Workspace account (for example a .edu address) and a personal Gmail at the same time, read-only. Briefs and answers draw on both. Outlook works too.
• A research-flavored brief. Shimmer delivers proactive briefs on a schedule. One brief kind is built for research: it surfaces questions worth your attention, grounded in your knowledge base, calendar, and inbox.
• Hand off a research question. Ask your agent to dig into something and it investigates, then reports back. Pick the Research & analyze focus during onboarding to turn this on.
• Literature search that cites real papers. Ask "what's the state of X" or "what's known about Y" and the agent searches the academic literature (via OpenAlex) and answers with real papers: title, authors, year, venue, DOI, citation count. It cites sources you can open; it does not invent citations.
• Bring your existing library. Connect Zotero (paste a read-only API key) to pull your library directly, or paste/upload a BibTeX or RIS export from Mendeley, EndNote, or Papers. Each paper enters your knowledge base as a source. Imports are unconfirmed: review them under Memory → Imported papers and keep or discard each.
• Pull your own publication record. Connect ORCID (paste your ORCID iD) to import your public works, so your research profile starts from the canonical source. Reviewed under Memory → Imported papers like any other import.
• Link your code. Connect GitHub (paste a read-only access token) and pick the specific repositories that matter; each becomes a project your agent can connect to your papers and collaborators.
• Add a paper by link. Paste a DOI, arXiv link, or journal URL and the paper resolves with real metadata (via OpenAlex) into your knowledge base.
• Track a topic on a schedule. Subscribe to a research topic (daily, weekly, or monthly) and your agent checks "what's new" on that rhythm, writing a dated research note each time.
• Find where to submit. Ask your agent where to submit work on a topic and it returns real venues ranked by how much related work they publish (grounded via OpenAlex). Submission deadlines are not included yet.
• A network of people and organizations. Track collaborators, advisors, labs, and institutions as a typed graph, with warmth so a relationship going quiet gets flagged. Useful for advancing a research profile, not just managing a calendar.
• A knowledge base built for research. Under the hood the KB has typed nodes for research notes, claims (each with the evidence it rests on and the run that produced it), and projects. Answers and briefs cite why they believe something.
• Capture by voice. Send a voice note any time. It is transcribed and understood.
• A researcher soul for your agent. The Biologist soul ("trusts observation over theory, tells you what she sees, not what you want to hear") fits research work, or write your own.

A day with it

• Morning: a brief lands with what moved overnight across your two inboxes, the meetings you have, and a research question or two worth your time.
• You hand off "what is the current state of X, and where are the open gaps?" and get back a grounded answer with sources you can open.
• A collaborator you have not spoken to in two months gets flagged before the relationship goes cold.
• As you talk to your agent about a project or a person, it proposes adding them to your knowledge base. You confirm. Nothing is written as fact without you.

Not yet, on the roadmap

What Shimmer does not do for researchers today, so you know before you plan around it. All of these are scoped and sequenced behind the current work:

• More scholarly sources beyond OpenAlex (arXiv, PubMed, Semantic Scholar) wired in directly to the agent.
• Drag in a PDF and have it extracted (paste-a-link already works; the PDF drag-and-drop path is coming with chat file upload).
• Google Scholar profile + citation tracking (no official API, so best-effort).
• View figures and images in chat.
• Submission deadlines alongside the venue suggestions (which already work).
• Draft and publish to a blog or Substack.

How it compares

Tools like Undermind, Elicit, ResearchRabbit, and Semantic Scholar Feeds each do a slice well: deep literature reports, citation-graph exploration, or new-paper recommendations. What none of them do is combine source-grounded literature intelligence with your private knowledge base and context and proactive briefs, in one agent that is provably yours. That intersection is where Shimmer is aimed.

Next: plans & models →