How many startup engineers does it take to change a lightbulb?

2 - a technical co-founder and a business co-founder. Since they don’t have enough money for proper lightbulb they use an old socket and lightbulb they found in the garage and since they don’t fit the technical co-founder builds a makeshift adapter from duct tape and aluminium foil. This actually works as long as you don’t tilt it too much. During seed negotiations the co-founders break up and the technical co-founder leaves.

Although they now have money for proper sockets and lightbulbs the new R&D team is unsure why the adapter was put in place so they decide to leave it and refactor it to be easier to work with. The new adapter fails Q/A due to flickering at certain angles and the Q/A team leader insists on a better implementation or he can’t sign off on it. Reluctantly the new VP R&D agrees and they spend the rest of quarter adding code to deal with various edge cases. The company has now grown to 25 people and build takes 1 minute. Immediately after launch it turns out customers don’t like the product since it ships with black light bulbs - it seems nobody wrote anything about color in the specs. The company immediately hires a VP Product to help get in line with market demands and a 5 person marketing team to convince clients black light is EXACTLY what they want. The new CMO hired 2 more people to do SEO. CEO hires VP HR, office manager, IT guy, 2 dev team leads and a Chief architect. Build time increases to 10 minutes.

Meanwhile, VP Product discovers that there are actually 2 competing standards for light bulbs: 220v and 110v. He pressures R&D to support both and starts working on UniversalLightBulbAdapter v2 which will allow connecting various light bulbs to various sockets regardless of voltage, physical connection or color. Unfortunately, due to pressing deadlines only 1 implementation (BlackLightBulb220v) is actually completed and it is hardcoded in various places in the code. Build time is now 30 minutes and a CI team is formed to deal with the busy CI cluster.

At this point, the company actually launches an experimental WhiteLightBulb implementation but it is rejected by the company’s 2 enterprise clients, since marketing whitepapers produced by the company’s marketing team makes it clear that BlackLight (TM) is far superior. The chief architect resigns in disgust and is replaced by 3 new chief architects (they all threatened to resign unless promoted). This results in everything having at least 2 incompatible implementations held together by proxy classes, adapters, ESB and ETL. Build time is now 2 hours and so complicated that only the CI team can build the product. VP R&D actually learns of this, but is too busy with the company’s new products: LightBulb management studio, LightBulb insights and LightBulb enterprise security (helping clients manage, monitor and secure their lightbulb adapters) - which are all programmed in Java SWING. To speed up development the company hires 3 new team leaders, 4 project managers and 2 full time SCRUM masters. Build time increases to 6 hours.

In the meantime, the newly formed Professional Services and Training division is booming - they have just launched LightBulb Academy giving CLBE & CLBA (certified light bulb engineer/architect) degrees; This is met with enthusiasm by enterprises' LightBulb integration departments which were formed to deal with the growing complexity and management of LightBulb adapters. 10 new solution architects join the company and help clients make sense of the growing portfolio - the company has acquired 5 smaller companies offering companion products to lightbulb adapters but they are hard to integrate with the core platform and sometimes incompatible. The new LightBulb utilities division vows to solve this ASAP.

The growing integration pains prompt management to bring a high profile consultant - which after careful analysis of the situation suggests the problem is that the company does not control light bulb and socket manufacturing and must dance to whims of other manufacturers. The CEO approves a 3 year plan to rewrite the core product from scratch so it will be “seamlessly compatible” with UberLightBulbs whose manufacturer has just been acquired by the company. In addition, they strike a deal with a Chinese factory to produce custom sockets but these are not actually used due to previous deals with other socket manufacturers. The 10 person legal team missed this completely as they were too busy with ongoing lawsuits - 3 cases suing opensource projects offering lightweight HTML5 based management interface for lightbulb adapters (allegedly they reverse engineered the management protocol) and 2 patent cases involving a 10 year old patent on “light emitting device”.

While preparing for the upcoming IPO, the company mistakenly join the OpenDaylight project - apparently the community manager did not understand this is an SDN project (as he is non-technical and did not confer with engineers) and not related to lightbulbs; however, to avoid a PR disaster just before IPO the company publishes a press release stating “our recent investment in the OpenDaylight project shows our commitment to the community and SDN market penetration” - they are starting a SDN innovation center headed by a professor but staffed with disgruntled engineers whose managers seized the opportunity to get rid of by means of a transfer. The professor resigns in anger and is replaced by a former Qualcomm manager.


In his “LightBulb world 2016” keynote speach, CEO Barry Elisson responds to the new software-defined lightbulbs trend by calling it “a hipster toy incapable of delivering the reliability, features, security and support modern enterprises require”. A crowd of 1500 lightbulb integration engineers cheers frantically in response. Stock prices are at an all time record.