Work tools
Synapse
A research studio for institutions, connecting sources, shared writing, and the papers they publish.
Synapse is a research studio for institutions. I’m building it to help people collect sources, investigate a question together, write, and publish work they can trace back to its beginnings.
A claim in a paper may outlast the team that wrote it. Someone returning months later should be able to find the passage that supported it, the version of the source, and the material the authors had available. That connection runs through Synapse’s design, from reading and annotation to a shared manuscript and its release.
Organizations, teams, portfolios, and guest workspaces give the collaboration a home. Access groups decide which sources people may use. The same permissions apply before material is sent to a model.
Source import and search, shared LaTeX editing, access checks, model context, and release handling are implemented. I’m developing the wider synthesis workflow and testing the pieces together. The aim is for each published paper to leave the next investigation with something it can use and examine.
How it works / Explore the workflow
From a source to a published paper
The design follows a source through reading, shared writing, and publication. Permissions decide which material a researcher or model can use. A release connects the PDF to the snapshot and build that produced it.
Learnings
Published work leaves the circumstances in which it was made. The people, working files, and tools may all change. A future reader still needs to know which sources supported a claim and which files produced the paper they are holding.
That history has to be captured while the work happens. Synapse once built a publication from the current files rather than the snapshot named for the release. A manifest added afterward could describe the intended build, but it could not establish that the PDF came from it. The release now requires the snapshot, compiler image, successful build, and PDF to be connected before publication.
The same problem appears in smaller forms during development. Code can describe a database that no longer exists. Preparing the project’s queries against a real schema found missing tables and columns that compilation had missed. I keep coming back to the point where one system hands work to another. That is where assumptions become dependencies, and where a direct check often tells us more than another description of the design.