Sheet 11 / 35 Human networks

Collaboration

Passport

Rock
Human networks
Depth
2 · Excavator
Time to dig
1–3 years
Capital
○ · low
Solo
✓ solo-reachable
AI
↓ under attack
Rent
✗ not for sale

Sample

Share of apps
17.3%
No-rate
17%
Median price
$15

Figures from the canivibecodeit sample. No-rate is the share of apps carrying this tag that cannot be vibe-coded — a proxy for structural strength. Only the first thirteen mechanics were measured.

Essence

The product only pays off when the whole team is on it: co-editing, presence, permissions, team workflow. What separates it from network effects is that here it is your team, there it is strangers. One of the two weakest tags in the sample.

How it is built

Figma — multiplayer as architecture, not a feature

Figma was built from day one around simultaneous editing in the browser: a CRDT-like engine, live cursors, everything on the web. Adobe could not retrofit that into desktop XD — collaboration turned out to be not a feature you add but an architecture you lay down. Once design became a team sport, Figma had no competitors by definition.

Google Docs — collaboration as a category killer

"Everyone in one document, no emailing versions around" was the only thing Docs did better than Word, and it was enough to take whole segments. The moat is reinforced organisationally: once a team has agreed that everything lives in Docs, the individual choice of editor disappears.

GitHub — workflow as an industry habit

The pull request is a team ritual GitHub invented that became a standard of the profession: training a developer includes training them on GitHub. The moat is in the muscle memory of millions and in the fact that teams describe their processes in its objects.

How it is bypassed

The single-player wedge — Obsidian, Superhuman, early Linear

A team product is weak against a tool that is better for one person: an individual chooses alone, with no committee, and brings the team afterwards. Obsidian grows against Notion on personal notes; Linear started with small teams for whom Jira was overkill. The attack from below: win one user first, add the collaboration later.

AI compresses the team

If one person with agents does the work of five, the moat of "you need the whole team" loses its meaning: presence, permissions and approvals are pointless when there is nobody to approve with. A solo founder conducting an orchestra of agents is a direct counterexample to the team premise, and a growing class of tools is designed for one person plus AI rather than for a department.

Move into somebody else's collaboration layer

Do not build your own presence — live inside an existing one: Slack and Teams became the surface through which teams interact with dozens of tools. The application hands collaboration to the platform — notifications, discussion and approvals in Slack — and competes only on its core function.

Verdict

Collaboration is a moat when it sits in the architecture and in the habits of teams. It is bypassed by a single-player wedge, by AI compressing teams, and by delegating collaboration to somebody else's platform.