Skip to content

A research observatory

Read the papers.
See the connections.

Bring the literature into view. Collect papers, trace claims back to source passages, and explore how ideas relate in one connected research workspace.

The scripted demo is open to everyone. No model calls.
Personal workspaces are invitation-only.

The connected viewA new perspective

A reading list, with another dimension.

Illustrative concept map · not your library

Explore the demo
Made for the work between papers
Literature reviewsResearch directionsShared reading

From a paper to a perspective

Keep your question.
Widen your view.

Research moves between reading closely and seeing the bigger picture. Your workspace should move with it.

  1. /01

    Collect a working library.

    Bring in arXiv papers by topic or ID. Organize them into collections that follow your research questions.

    A place to begin

  2. /02

    Trace evidence to its source.

    Inspect passages about findings, methods, and limitations. Keep the original words within reach of the interpretation.

    A reason to look closer

  3. /03

    Explore the relationships.

    Follow citations, shared authors, and semantic neighbors in 3D. Filter the view and focus on the papers that matter to you.

    A different perspective

  4. /04

    Ask grounded questions.

    Ask across available source passages and inspect the cited evidence. Q&A is available in enabled workspaces.

    A path back to the literature

The evidence stays in sight

A useful insight
has a paper trail.

Read the passage behind a finding. Inspect the source text and model provenance behind a suggested relationship. Decide what belongs in your argument with the evidence in front of you.

A citation is a source link.
Follow the reference back to the original work.
Similarity is a starting point.
Nearby ideas may be worth reading together.
An inference is a suggestion.
Review its evidence. Uncertain results can remain unclassified.

Analysis depends on available text, sometimes only an abstract. Source references support your reading; they do not replace critical judgment.

At the sourcearXiv · 2017

Vaswani et al.

Attention Is All You Need

From the abstract

“based solely on attention mechanisms, dispensing with recurrence and convolutions entirely.”
Read the original paper (opens in a new tab)
An attributed source example. This illustration is not an AI analysis or a record from your workspace.

For the research in progress

Make room for the next question.

I

Find your bearings in a new field.

Start with a paper you know. Explore its neighborhood, search your saved library, and build a reading path through unfamiliar work.

For the literature review

II

Build an argument you can trace.

Return to the source behind a claim. Separate a recorded citation from a similar idea or a model-suggested connection.

For the question that matters

III

Give your reading a direction.

Gather a focused collection for your next project or discussion. Keep it private, or share a reading list with a revocable link.

For the work ahead

A considered beginning

Open to exploration.
In private preview.

Explore curated paper graphs and scripted research conversations in the public demo. Personal research workspaces are currently available by invitation.

Take a look around
Can I explore without an account?

Yes. The interactive demo is public. Choose AI, climate, biology or space; explore eight real papers in each topic, inspect editorial notes and try prewritten conversations. No account or AI model calls are needed. Thematic links are editorial examples and carry no scientific scores.

Who can open a personal workspace?

Literix is an invitation-only preview. Personal workspaces are provisioned individually; creating an account does not automatically create a research workspace. Invited researchers can sign in using Open workspace. Paid upgrades are not currently available.

What does the AI evidence tell me?

Analysis uses the text available in your library, which may be an abstract rather than the full paper. Source quotations, review candidates, and inferred relationships are shown separately. Model confidence is not a calibrated probability of scientific correctness, and the system can abstain when evidence is insufficient.

How does grounded Q&A work?

In enabled workspaces, answers draw on available library passages and link back to those sources. Citations help you inspect an answer; they do not guarantee it is correct. When available, a separate faithfulness model checks whether cited passages support individual sentences; unavailable or partial checks are labelled. Q&A is not included with every preview workspace. The public demo uses prewritten responses and editorial notes, with no verification or model calls.

Follow a paper.
Find a new direction.

There is more to a reading list than a list.
See what a different perspective reveals.

Curated papers · scripted responses · no sign-in required