N11 THE REALITY LAYER
From Papers to Projects: The Missing Collaboration Layer
Knowledge becomes work only when people, permissions, files, decisions and incentives share a surface.
IN THIS NOTE · MAY 2025
Publishing systems are optimized to reveal completed knowledge. Collaboration systems must coordinate unfinished work, changing assumptions, confidential material and contributors who arrive at different moments.
A paper is an endpoint
The scientific paper compresses a messy process into a stable artifact. That is useful for communication but incomplete for coordination. A team working toward the paper needs open questions, task ownership, protocols, files, discussions, decisions and evidence about why the plan changed.
When those objects live across inboxes, drives and private chats, the project loses memory every time a participant leaves.
Projects need durable context
A collaboration layer should preserve the relationship between a claim and the work that supports it. Agents can help summarize progress, surface contradictions and propose scoped tasks, but their actions need attribution and human review.
The workspace must also distinguish public discourse from protected execution. Institutional teams will not put core IP into a surface whose boundaries are ambiguous.
Graduation is the product
The strongest collaboration systems help ideas graduate. A post becomes a project. A project earns contributors. A scoped question becomes funded work. Results become a decision about the next experiment, protection or launch.
The meaningful metric is not how much discussion occurs. It is how reliably worthwhile ideas move into accountable execution.
The project is the unit of scientific memory
Papers are durable outputs, but they are poor containers for work in progress. A project needs a living map of the question, evidence, assumptions, contributors, protocols, decisions, rights and unresolved risks. Without it, every new collaborator reconstructs context from meetings and folders. Important negative evidence disappears because it never became a publication, while decisions survive only as conclusions detached from the reasons that produced them.
A collaboration layer should therefore organize around versioned project objects, not a feed of documents. A claim can point to sources and analyses. A protocol can point to the decision it is meant to support. A result can update a claim without erasing its prior state. This creates continuity across humans and agents and makes handoffs less dependent on the availability of one person who remembers where the real story lives.
Collaboration quality is measurable
The number of comments or connected researchers is a weak proxy for progress. Better measures examine whether collaboration changes the work: time to resolve a blocking question, percentage of decisions with traceable evidence, reuse of protocols, successful handoffs, contradictory findings surfaced before commitment and external reviewers able to reproduce the reasoning. These metrics value coordination without confusing activity with scientific output.
The platform should also make non-participation legible. A requested review that never arrives, a collaborator waiting for data rights or a vendor blocked on an undefined endpoint can silently hold the critical path. Workflows need owners and service expectations even in open communities. The objective is not to industrialize every conversation. It is to protect scientific attention by ensuring that the important coordination work produces an artifact and a next state.
- Connect every discussion to a project, decision or evidence object.
- Preserve context as contributors and agents change.
- Measure graduation into funded, completed and reported work.
I would reconsider if fragmented communication tools proved sufficient for complex, long-lived scientific collaboration without material context loss.
Primary and institutional sources used as the grounding layer. Interpretation and synthesis are Luca's.
01