The Shared Mind
On mental alignment, the weight of understanding, and why your team's shared mind matters more than your code
There is a specific, quiet moment in the life of a high-functioning team when friction disappears. You have likely felt it. A sentence is started by one person and finished by another. A critical decision is made in the negative space between messages. The architecture emerges from a collective knowing that requires little verbalization.
Then, someone leaves. Or three people join. Or the roadmap pivots.
Suddenly you’re rebuilding something you didn’t know you had built.
When Marcus Aurelius wrote that we suffer more in imagination than in reality he was speaking of anxiety, but he might as well have been describing the cognitive physics of a software team at scale.
The suffering is the coordination and not the code.
It is the twenty-three minutes required to rebuild mental context after a single Slack ping. It is the invisible web of 190 communication pathways1 that entangles when a team of eight becomes a team of twenty.
Its not a management problem but rather one of physics and no amount of SAFe can repeal it.
The Myth of the Beginner’s Mind
We romanticize the “beginner’s mind”. The idea that expertise is a trap and that knowing too much closes us off to possibility. This is beautiful in contemplative practice.
In a high-velocity environment, the opposite of beginner’s mind is shared understanding: the deep, accumulated context that allows a team to move with less explanation.
Recent research by Carraro et al., 20252 confirms what experienced leads know: teams do not just share code repositories, they share a “collective working memory.” This is a parallel internal representation of the system, the domain, and each other. It allows them to predict behavior without asking and to coordinate with less meetings.
The corollary to this is that cognition cannot be transferred. A new hire does not inherit the team’s memory they have to build it from scratch through the expensive, slow work of communication.
The Toggle Tax
We use communication tools to move faster and many make us slower.
Workers switch contexts every 11 minutes3, losing 20% of their cognitive capacity to the toggle tax alone. The platforms we built to reduce friction have become friction.
Technical debt lives in the repository. Cognitive debt lives in the mind and is the accumulated weight of context that must be reconstructed every time someone asks a “quick question.” We treat human attention as infinitely renewable. We piled on scope. We introduce tools that masquerade as help.
John Sweller’s Cognitive Load Theory names this problem. He identified three types of mental burden:
Intrinsic Load: The irreducible complexity of the problem itself. You cannot reduce this without changing the product.
Extraneous Load: The friction of the environment: bad docs, unclear ownership, meetings that should be messages.
Germane Load: The productive energy of actually learning and building. This is where value lives.
The mandate is simple in theory but oh so difficult in practice: eliminate extraneous load to make room for germane load. If a meeting can be a document, the meeting is extraneous. If a tool requires constant context-switching, the tool is a liability.
The AI Bottleneck
This conversation is no longer optional because the constraints of software development in 2025 have clearly shifted.
For the last decade, we built systems knowing that human bandwidth was the bottleneck. Now, we are deploying AI tools that generate code faster than we can specify what to build. The bottleneck has moved from development speed to specification clarity.
Specification clarity is a function of shared mental models.
You cannot clearly specify to an agent what you do not collectively understand as a team. AI does not eliminate the need for alignment but it does amplify the cost of misalignment. Every underspecified prompt with missing assumptions becomes an error generated at scale and speed and will widen the intent gap4.
The teams that thrive in this new era will be the ones with the deepest shared understanding of the problem made explicit.
Aligned Autonomy
So, how do we defend this shared mind against the entropy of growth?
We must rely on thin interfaces. The best teams do not coordinate constantly but they do coordinate intentionally. They create “aligned autonomy” which means freedom within a framework. This requires a North Star clear enough to guide decisions and guardrails sturdy enough to prevent drift.
But mostly, it requires trust. Trust that mistakes are data. Trust that you do not need to be in the room for the room to function. Marcus had a meditation fitting of this moment when he wrote…
Many lumps of incense on the same altar. One crumbles now, one later, but it makes no difference. — IV. 15
He was writing about mortality but read it as a statement on the lifecycle of a team.
We are all burning toward the same deadline, in the same shared space of attention. The incense that falls sooner isn’t better it is just sooner. The only thing that persists is the architecture we build in each other’s minds.
Because we know that shared understanding is the product. Everything else is just an artifact.
See Brooks Law
Carraro, Michela, et al. “Unlocking Team Performance: How Shared Mental Models Drive Proactive Problem-Solving.” Human Relations, vol. 78, no. 4, 2025, pp. 407–37, https://doi.org/10.1177/00187267241247962.
Mark, G., Gonzalez, V. M., & Harris, J. (2005). “No Task Left Behind? Examining the Nature of Fragmented Work.” Proceedings of the CHI 2005 Conference on Human Factors in Computing Systems.
See my September writing here