It’s very fashionable in geek circles to attack Paul Graham at the moment, particularly after his essay You Weren’t Meant to Have a Boss. I’ve been reading his essays for a few years now and I’ve wavered between agreement and an undefinable sense of unease. Now I believe I can finally pin this down.
The central point of Graham’s Boss essay seems to be that over a certain size organisations become rigidly hierarchical and once the hierarchy sets in the creativity of programmers is significantly and fatally constrained: over a certain size an organisation will be incapable of producing original software. This is a continuation of a theme through much of Graham’s writing. Startups do interesting work and software development will migrate exclusively to startups.
Well, I think that’s something I can disagree with. Of course, it’s pretty obvious that I have a vested interest. An iconoclast like Paul Graham will always get the most vicious response from those he seeks to help. Allow me to give my personal background, in the interests of disclosure.
My entire career has been in software development. In my first job out of Uni I fell into the ‘hero’ programmer archetype. In every job after that I’ve been in some form of senior position: tech lead, architect, team lead. I’ve also been an independent consultant, co-founded my own startup (we failed; be very careful about selecting your co-founders) and now I’m working for a startup. Yep, as an employee with a boss and all.
I spent sometime this morning going through all the startups listed on the Y Combinator website. For each startup I’ve tried to classify them according to the current big themes in web sites.
Advertising & Sales:
Apps on the Web:
The ‘other’ category is the interesting one: into that bucket falls a server admin dashboard, a web-based TV channel and a plugin for searching Outlook email. But, there are a lot less of those than the social networking and web-based desktop application startups.
I’m sure many, particularly the founders, will disagree with my classifications. But, these are mainly right, especially if you read ‘Social Networking’ as ‘Social Networking around Common Interest X.’ And in the end you’ll find the exact classification is not important.
All of these startups share a few things in common. They were all launched quickly and they’re all pure software development, often running on someone else’s eco-system. There is a place for development like this, but if this is the future for computer science I believe the field will be significantly poorer for it.
I am working for a company that by pretty much every definition is a startup. By the time we left stealth mode in March last year the company had been around for 13 years and had grown from a core group of computer scientists to a company of over 300, including chemists and physicists. Oh, and we invented a new type of printer. A startup like ours simply doesn’t fit into the Y Combinator model. We also don’t fit into the small company with no bosses model: building hardware takes time and a lot of people, you simply can’t avoid either.
This is my concern. Is all future computer science productisation and development really going to be latest and cool ad-funded mobile social networking site for parrots? Because, excuse me if I’m not excited by that future. I am still excited by the potential for computing and the Internet in particular, but that potential is better served by longer-term thinking and grander plans than refinements of what everyone else is doing.
This criticism may actually run deeper. A continuing trend in computing is to make programming easier for all. This has had the effect of pushing some tasks out of the realm of the programmer and back to expert users. This has been a good thing. Users have more control and programmers are free to work on interesting problems. The web has also been fantastic at improving human to human communication. Recently, the potential of the web as a mechanism for computer to computer communication is becoming more apparent. Many of the Y Combinator startups very effectively exploit this: improved experiences and convenience by combining the information on eBay with the blogosphere. It also appears that these startups are surfing a wave. The gap between technology becoming easy to use and becoming easy to program. In other words, I suspect this style of startup is not long for this industry. An historical aberration, automatic arbitrage for company startup, as it were.
This is not the end of startups of course. There will always be startups, however there will have to be some interesting, risky and difficult technology behind the curtain. Originality will once more be prized. In this world, You Weren’t Meant to Have a Boss will make a lot less sense.
Personally, I fully expect to do the startup ride at least once more. And I’m looking forward to doing that in a world that once more demands innovation rather than just another social network. After all, I really do want to add something to the world and I just don’t see that happening with late noughties startups.