Reflections, 17th of May
On asymmetry, symbiosis, and the role we do not yet have a word for
There’s a crack in everything. It’s where the light gets in.
Leonard Cohen
Where the Puck Are We Going?
Wayne Gretzky’s father, Walter, coached him to move to where the puck was going to be, not where it had been. It was a line famously used by Steve Jobs when he talked about the iPhone.
In the frenzy to capitalise on a technology we do not yet really understand completely, I think we may have forgotten this lesson.
When I look around the business press, social media, and increasingly, unfortunately, here on Substack, people are obsessed with where the puck has been. I think they’re missing the point.
The critical factor in any organisation, particularly in volatile and emerging sectors, is not the tools or the technology. It’s the social friction between people. AI can speed up tasks, coordinate, and reduce infrastructure and overhead, but it cannot build trust or relationships or the deep knowledge of who does what and why.
AI does not change the nature of small groups. It changes the relationship between them and large institutions in a way the larger institution’s structure cannot match. The asymmetry has always been there, but AI is magnifying it. AI is collapsing the time cost of cognition: the research, synthesis, drafting, modelling, and experimenting. What remains as the binding constraint is the coordination cost: trust building, mutual legibility, conflict resolution and the psychological texture of who knows what about whom and which questions to ask. AI is lifting the task ceiling but leaving the social ceiling where it is. Return on group size, in terms of work produced — both quantity and quality — shifts towards smaller numbers, probably substantially smaller, for speed-critical work.
Research at MIT Sloan found that organisations may benefit from assigning entire chains of tasks to AI, even when humans could perform some steps better. That is because each handoff requires review, validation, and adjustment, and those checkpoints slow the system more than the quality differential costs. A February 2026 arXiv paper framed AI as “coordination-compressing capital,” a class of capital that does not substitute for labour, but rather reduces the friction between agents. The implication is that human-to-human handoffs have become the expensive operation. Asana’s 2025 research, based on insights from over 9,000 knowledge workers, found that 65% of workers say AI creates more coordination work between team members. That share rises to 90% among the most productive workers. Hyper-enabled individuals generate output faster than the surrounding social system can absorb it. Asana used the analogy of Formula One drivers racing on roads designed for horse-drawn carriages. The bottleneck is not production, but absorption.
Where things are emerging and fast-moving, we can see this in real life. Gamma, the AI documentation platform, was serving 50 million users in thirty days with 30 people and has now scaled past a hundred million dollars with over 70 million users. Bolt, the ride-hailing app, reached $20 million ARR in 60 days with 15 people. (We used Bolt when we were in Malta and never waited more than a minute for a cab. It was remarkable.)
This is not just a productivity story. It creates a schism. Small companies can move much faster than big companies when it comes to establishing new ideas and enterprises, but they can quickly reach a point where they cannot manage the scale they have reached with the structure they have. Big companies have benefits of scale, but have all the social friction that no amount of technical expertise will reduce. They are very good at reducing cost, so that’s what they do, whatever the PR may say.
The challenge seems to come when the big companies buy the small companies and then effectively destroy the culture that has created them. The tech sector in general and the game sector in particular seem to exemplify this. They have become leadership deserts.
The analogy that springs to mind goes back over a century, when Heinz Guderian made a nonsense of the conventional wisdom of trench warfare with Blitzkrieg tactics. By moving around and beyond obstacles without consolidating them, the Germans made startling progress and tidied up behind them as they went. It all then fell apart in the Russian campaign when they used the same tactics but didn’t have the backup to do the tidying up.
I think we face the same issue. The small, fast companies need the big companies as backup, but the cultures are so varied that they cannot work together.
The pattern is visible enough once you look for it. Bungie inside Microsoft, and Instagram inside Meta in its early years. Countless smaller cases that never made the press because the small thing was simply absorbed and forgotten. The acquisition closes; the meeting cadence imposes itself; the people who built the thing leave within eighteen months; what is left is the brand and the dashboards. The acquirer reports the synergies. The acquired returns to the Substacks and the small ventures and quietly starts again.
We have a word for one form of this relationship — acquisition — and the word does the thinking for us. Once we say “acquired,” everyone knows what happens next. What we do not have is a word for the other relationships that would be possible if we were prepared to design them. Partnership tends to drift into dependence or competition. Outsourcing converts the small group into a vendor and strips out the agency that made them worth working with. The relationship the moment calls for has no name yet, and the institutions cannot manage what they cannot name.
Symbiosis is the term biologists use for organisms of different kinds living together in ways that benefit both. The conditions for symbiosis are specific. Each organism must do something the other cannot. Neither must metabolise the other. The interface must remain stable for long enough that both organisms become adapted to the relationship rather than merely tolerant of it. Coral and zooxanthellae. Lichen and fungus. Mycorrhizae and roots. The relationships are not equal, not symmetrical, not negotiated; they are evolved, and they take time.
What the asymmetry between small AI-enabled groups and large institutions calls for is symbiotic, not transactional. The small group needs what the large institution genuinely has: capital that can absorb risk, infrastructure that no five people can build, institutional memory that compounds over decades, and relationships with regulators and customers that no startup can manufacture. The large institution needs what the small group genuinely has: speed, judgment, mētis, the willingness to commit before the picture is complete.
Neither can produce the other’s contribution. Neither survives, in the long run, without something like the other’s contribution.
The historical precedents for this are older than the modern firm. The Jesuit colleges sat inside the Church but operated with their own rule, their own selection, their own pedagogy. The City of London livery companies regulated their crafts without being absorbed by the Crown that chartered them, and the BBC, in its best period, commissioned independent producers who retained their cultural distinctness and were the better for it. The university research group, properly governed, has the protection of the institution without its tempo — though this form, too, is being eroded by the financialisation of education. The form is not new; what is new is the absence of any contemporary corporate vocabulary for it.
The reason large institutions struggle with this is not lack of intelligence. It is impatience. Genuine symbiosis takes years to establish, when the cycles of corporate leadership are measured in quarters. The acquisition closes in months and the integration completes in months after that, by which point the small thing is no longer small and no longer what was bought. What is required is the willingness to leave something alone in a structured way; to give it air, money, time, and protection from the host’s metabolism, and to do this for long enough that what emerges is something neither party could have produced alone. Few institutions are designed to permit this. Fewer still have leaders who would be allowed to attempt it.
Which raises the question of the leader. The symbiosis does not maintain itself. It requires someone who can hold the interface between the two forms — someone with the patience of the institution and the velocity of the small group, who can speak both languages without being captured by either, who is willing to operate in a register that does not pay them in the institution’s coin. They are not a manager. They are not a sponsor. They are not a coach. They are something the modern organisational chart cannot name.
Every functioning symbiosis I can think of in the historical record has had this figure. The rector who protected the Jesuit college from the bishop. The master of the livery company who held the standard against the impatience of the market. The commissioning editor who fought the management for the budget and the time the work needed. They are not always visible. Their work is mostly invisible by design. What they do, in essence, is tend the conditions in which other people can do their best work without being interfered with.
I find myself reaching for the alchemical word, which is “tender”. The tender of an athanor — the slow-burning furnace in which transformation occurs — does not perform the transformation. They keep the fire steady, watch the temperature, and protect the vessel from interruption. They know when to intervene and when to leave alone.
Their craft is not in the work itself but in the conditions for the work. Without them, the fire goes out or runs too hot, and what was meant to become something extraordinary becomes something ordinary instead.
This is, I suspect, the role that the AI-era institution will most desperately need and find hardest to recognise. It does not produce visible output and does not appear on the dashboard. It cannot be measured against quarterly targets. It is the human function that holds the interface between the small fast thing and the large slow thing, that protects the agency of the operators from the metabolism of the institution, that maintains the relationship long enough for both parties to become genuinely adapted to it.
The Elvis (mētis, craft, intuition etc) I wondered about last month has left a great many buildings. He has, in many cases, found his way to small groups working with AI in conditions where his genius can still be exercised.
The question I am now sitting with is whether the buildings he left can learn to host his return without consuming him again. That depends, I think, on whether they can develop the role I have just described — and on whether anyone in those buildings would wish to inhabit it.
I do not know the answer, but I notice that the question is not, in the end, about technology. It is about the kind of leader the moment is asking for. Whether such a person can still be found in the buildings is an open question. Where they will be cultivated, if the buildings cannot produce them, is a question for another day.