Notes from "Google, Innovation, and the Web" Craig Silverstein, director of technology O'Reilly Emerging Tech Conference 2003 Cory Doctorow doctorow@craphound.com -- Google: Organize the world's information and make it universally accessible and useful. * This gives us focus, but still gives us enough material to work on for the next 200+ years. Google: do things that matter. * Helps us recruit smart people, even though we pay 1/10 of what Wall St does * Helps us not be evil -- we can succeed without being dickheads Google: focus on the user * Popup ads suck, text-ads are in context and don't suck * Google Labs: interesting projects, lets our users help us make our products useful to them * Switching costs for search-engines is really low Google: * brilliant people have good ideas * creative environment helps * our process works -- The process: * We need good ideas -- which can come from anywhere (as can bad ones): inside and outside the company * Design for users: our delta force of UI experts help us plan this stuff. Co-founder Larry Page was a UI person. Getting the Google frontpage as simple as it is takes a lot of effort. We count the bytes on our homepage. * Compile, discuss, prioritize. We keep a list of the 230+ projects that Google would like to do. We keep a list of what's important and in some vague priority order, lets us focus on our whole mission. * We use Sparrow, a Wiki-like tool from PARC, that allows *anyone* in the company to add ideas to the list of to-dos * Product discussion forums: brainstorm new ideas from Sparrow. * New field data changes priorities: rising wireless usage will reprioritize wireless-based product ideas. * Small teams, fast and agile: every product has a 3-person team that owns it from start to finish, design coding testing launching and maintenance. Two people -- Jillzilla and Kingston -- run the crawl, the billions of pages spidered by Google * Code is reviewed for style and readability before it's committed. We have 10 "interteams" of three people that coordinate activity among the teams. Volunteer positions. They know what's useful and what's not and what engineers are going through. * People in the company do weekly TechTalks (which are archived) that describe the projects -- if you inherit a project, you can review it later. * We have search tools for our internal site -- i.e., we google Google * We have signoff-checklists on our site what need to be checked off by everyone in the chain before a new project is launched * Project webpage lists the last week's work and next week's work, generated by emailing a bot that compiles a list. Lets us identify redundant work, chances for mutual aid. Keeps people from wandering off-track. * We use internal blogs to communicate. The press after the Gbloogle buyout never speculated that we might be buying Blogger to use it for in-company comms. * Test: We used to just post stuff and wait for email to come in. Now we're more cautious, we do user-studies. We use Google Labs. -- Sidebar: www.google.com/jobs * We need good people. We don't care if you know from search. We want people with the right attitude: curiousity, drive. * We're conservative in our hiring process. One cttee controls all the hiring that we do. Seaparates hiring from headcount ("We need 10 more engineers!" <-- not the person who makes the hiring decision, which means that no one is hired out of desperation) -- The Web changes everything: * We couldn't have done this without the web * No search engines w/o the Web * Simple search engines were sufficient pre-web * We couldn't have run the little teams and still maintained communications without the Web * We wouldn't have logfiles to tweak our algorithms without the Web * There's no lock-in for search, and marketing doesn't drive people to search: the technology is what wins the search wars -- Q&A Communication is really tricky when you're in a different timezone. We just opened an NYC office. East Coast culture is 9-5, and Google is more like noon to 8. Makes it really hard to be realtime, increases importance of Web-based comms. Why do we "only" index 3BB pages? It takes time to crawl the Web -- some sites can't take the load if we spider them more frequently. We want the whole Web. We need a way to only d/l the pages that have changed. robots.txt and proprietary databases frustrate this. We just acquired Applied Semantics. They do targetted advertising. I can't say more. Semantic tech could be applied somewhere other than ads. If I said more, I'd have to kill you. We try to get ideas from everywhere, but not all ideas get selected to be talked about. Gatekeeprs decide what we discuss and consider. They're techies, engineers.