the general theory of founders, managers and systems

notes on “founder mode”

Someone asked me what I thought about Paul Graham’s post from a couple of weeks ago on “founder mode”.  The instinctive reaction is to groan and go something like “oh no, tech bros reinventing the wheel again”, but I think that would be a mistake for two reasons.  First, in my experience Graham is considerably more thoughtful and interesting than the run of the mill for Silicon Valley. 

And second, reinventing the wheel is underrated.  I happened to be working on two books at once in 2021 – one about cybernetics and one about the Brompton Bicycle company.  Although this did me quite a bit of brain damage and I don’t recommend it as a lifestyle, there was a synergy, because at the same time that I was learning about control theory from the theoretical ground up, I was talking regularly to Will Butler-Adams about the actual business of management. 

Ever since then, I’ve been fascinated by the fact that left to themselves in charge of something, intelligent people with an engineering background will always seem to independently come up with something that looks quite like Stafford Beer’s cybernetics.  This was the basis for my previous post about “intellectual carcinization”.  As well as being intrinsically interesting, it makes me always want to read this sort of thing, because I think it might be important to get a handle on which bits of management cybernetics people tend to reinvent, because that’s likely to say something about what’s important or correct about the theory.

Obviously therefore the first thing that struck me about the “founder mode” essay is that there’s a hell of a lot of cybernetics-adjacent material in there – he even talks about black boxes!  The second thing that struck me is that there is indeed quite a bit of wheel-reinvention in the pejorative sense.  I don’t agree, for example, that having “split level meetings” which cross the hierarchy is an unusual thing, or that there’s any “principle that the CEO should engage with the company only via his or her direct reports”.  When I was at Credit Bloody Suisse, they used to have global offsites with 100 key employees from different levels – I went to one when I was a junior VP because I had made a spreadsheet model of the first regulatory stress tests that got forwarded around a lot.  I don’t necessarily want to gainsay Paul Graham’s experience here, but if VCs are in the habit of imposing the kind of structures that he describes on their portfolio companies, then I think every business school prof in the world would agree with me that they are being dicks and should stop.

But dividing through by that and getting back on my hobby-horse, “founder mode” really is talking about the same things that Stafford Beer was, because its central point is exactly the one that’s the subject of “Diagnosing The System” – that there is no reason at all to believe that the best way to design a system of information flows and allocate bandwidth will produce something which looks like a normal org chart.  And even less reason to believe that the relationship between the org chart and the correct information design will be stable at different moments in time.

A lot of people who read the “founder mode” essay seemed to latch on to the idea that it’s a justification for micromanagement, based on this passage:

“Hire good people and give them room to do their jobs. Sounds great when it's described that way, doesn't it? Except in practice, judging from the report of founder after founder, what this often turns out to mean is: hire professional fakers and let them drive the company into the ground.”

I admit, this does look pretty much like an apologia for micromanagement.  But if we ignore the drive-by at “professional fakers” (which I doubt are more common in tech than other industries), then let’s compare it to this passage:

“Complex organisations … need to be aroused when a threat to their survival appears and to recognize quickly when a desired state has been reached.  Too often, however, the non-routine channels for such information in a large organisation are inadequate or non-existent and indications of danger are filtered out rather than amplified.  Management must design non-routine channels so that it may act on non-routine information in time to avert disaster or seize opportunity”.

This is from Allenna Leonard’s “Cybernetic Glossary”, at the back of Vanilla Beer’s wonderful biography of Stafford Beer.  It’s part of the discussion of what the jargon calls “algedonic signals”.  These are low-frequency, high-priority messages which bypass the usual channels to direct signals from the immediate environment to the level of organisation at which the system itself can be redesigned. (My go-to example is the red handle in a train driver’s cabin, which not only physically stops the train, but has an organisational role in beginning a process of rewriting the timetable for that day).

I think “founder mode” is partly a rediscovery of the need for this kind of thing.  “Founders” don’t really go around suddenly jumping into operational tasks at random for the joy of nitpicking; they do things when they think that they have new information that requires them to get involved. 

And, of course, in order to identify what information has the red-handle property, you have to know what the purpose of the company is – “purpose” in this sense meaning “the criterion of what makes something relevant to the company”.  That’s why Graham identifies it with foundership – in context this means someone who has the right to decide the purpose, and therefore occupies the right spot in the information processing system. 

But I think it’s a position in the system rather than a unique property of start up guys; my go-to-example here would be John Birt, who didn’t create the BBC, but who was in the right position to take two pieces of information (“this internet thing has arrived” and “the BBC can no longer count on license fee growth”) and turn them into a pretty radical set of organisational changes.

Anyway, this post is already too long.  Although Paul Graham is correct to say “There are as far as I know no books specifically about founder mode”, I’d argue that there are a few very good books on general principles of information engineering, which have founder mode as one of their special cases.  There’s a list of them in the back of my book if he doesn’t mind paying academic prices.