"Even as we've grown, we've remained a lean organization. We do not have room for people who don't do the actual work." That is a priceless comment. It exposes Jason's huge blind spot. Worse, it is on this undefended flank that great future pain may be inflicted. I look forward to the post-learning article. The underlying premise/assumption is that a 'manager' is not only not doing 'the' work, but they aren't really…
I strongly disagree, and I think the metaphor leads one astray.
In a startup there is not a navigator and then an army of oarsmen. Instead there's a team with diverse skills attempting to understand and solve a complex problem. A manager can only direct rats to the cheese if they themself fully understand mazes, cheese and shoveling shit. But if each rat has diverse and individual ways to contribute to getting closer to the cheese, the job of a single manager providing vision and directing becomes geometrically harder as each new skill/role is added. Imagine each rat speaks one of a babel of different languages, and we start seeing how hard the job is for our imaginary leader.
It's incorrect to assume that the rats shoveling the shit don't have their own vantage on the cheese, or that the oarsmen can't see the stars and where the ship is headed on their own.
This is the difference between an industrial structure and a a research structure.
Top down command structures are extremely effective when everything is already understood. It's what you want when you're solving a known problem with known tools, like building commodity products on a production line.
But top down networks face enormous challenges when trying to achieve only partially understood goals within complex changing environments. Integrating feedback as it travels back up the tree is very difficult to do effectively. Top down command networks are at their peak efficiency when feedback doesn't need to go up very high. Otherwise, the top of the structure becomes overwhelmed.
This divide is almost entirely why large technology organizations end up innovating primary by acquisition rather than internal discovery (example: MS). They have to buy both solutions to new market opportunities as well as the people who understand and built those solutions, because their internal structure is very poor at learning and adaptation.
Collaborative problem solving is different. It relies on frequent and open ended horizontal communication. It often appears directionless. But what it facilitates is overlapping shared understandings far more detailed than could ever be centralized in one person. It allows points of command to emerge when necessary, and disappear when not. The ease with which a small team does this organic communication and discovery is one of the huge advantages of small companies. It's not one you should shed lightly.
The only way to do this in large companies is to maintain a very horizontal structure where individual business units have a great deal of autonomy in accomplishing their goals, and where employees can fluidly move between teams to spread knowledge.
There's another option for the big company: create an internal market structure. But AFAIK no one has a repeatable template for quite how to do that (and apparently Google's attempt at it has somewhat failed).
In any case, returning to flogging your metaphor:
The smart navigator will listen to the oarsman: perhaps they've noticed something I haven't? And the very smart navigator will teach all who are interested about his skills, so that the whole boat becomes more valuable than it's competitors.
And notice that no where in here have I mentioned any terminology about prestige, respect, or "actual work". This stuff is cancerous, and whenever you see it, get rid of it immediately.