The social layer
#8am-ai#people-and-team#deep-dive#collaboration#teams
David Olsson
Smaller than the technical threads, sharper for it. The people-and-team thread is about the layer where every individual AI gain compounds or evaporates: other humans.
The recurring finding is uncomfortable. The model was never the hard part. The team was.
the arc
2024: collaboration as opportunity. The framing is optimistic — diverse skill sets, future collaborations, the value of mixing people who know different things. AI is a reason to work together more.
2025, first half: partnerships and load. Running several projects and partnerships at once. Funding arrives, and with it the question of how to balance commitments without dropping the people involved.
2025, second half: onboarding and context-sharing. The concrete pain surfaces. Bringing a new team member up to speed, and the difficulty of sharing context across different AI tools. Everyone's accumulated their own corpus in their own tool. Getting a team to share that context is hard.
Late 2025: the systemic mirror. Ying's survey result lands cold. Canada ranks 44th of 47 countries in AI readiness, attributed to cultural and institutional factors, not capability. The team problem scales up to a national one.
2026: teams got harder. Jason's blunt question — is working in teams harder now than before the shift? He struggles to find a logical division of labor when one person plus agents can do what a team used to. The unit of work changed faster than the unit of organization.
the tension
The corpus keeps returning to a contradiction it never resolves. AI makes the individual more capable and the team more confusing.
When a single person with a swarm can do a team's worth of output, the old logic of dividing work breaks. Who does what, when the "what" is increasingly "direct agents"? The collaboration that 2024 celebrated becomes, by 2026, a coordination problem nobody has a clean answer for.
This is the social face of atomic-to-architecture. If everyone is pushed up into directing systems, teams stop being assemblies of task-doers and have to become something else. The corpus is honest that it doesn't yet know what.
Context-sharing is the team bottleneck. The tooling for "my context" raced ahead of the tooling for "our context." Readiness is cultural, not technical — what holds groups back is institutions, incentives, and willingness to change. And the division of labor is unsolved. When one person can do a team's work, "what is everyone else for" is a real, open question.
The corpus's own pattern — a shared, structured substrate everyone reads and writes — is the obvious candidate for "our context." Whether it replaces the lost logic of divided labor or just documents its absence is the next thing to find out.