Thursday, June 28, 2007

PowerSet: Let's Re-Invent Everything

I drove up to San Francisco for the inaugural PowerLabs open house. Nearly four hours later when it was all over, I gave three PowerSet people who had missed their train a ride down the Peninsula. "What's great about this company is the business people," said one, a PhD. "They really seem to know what they're doing."

Yeah. There are a lot of smart tech guys who've been in too many dumb companies. This certainly isn't one of those.

Cover Slide

The big knocks on PowerSet -- 'PowerHype,' etc. -- can be broken out into three categories:

1) Can they build it?
2) Will anyone care / do they have a business?
3) What will Google do in response?

I've not been worried about 1) for quite a while. While there are many real technical challenges to bringing up a web-scale search service, there are many real technical guys at PowerSet who've done it before and can do it again. Since modern search is far more heterogeneous than most people realize -- a spaghetti ball of models, features, and evidence driving every instance of ranking and retrieval -- PowerSet doesn't have to climb all the way to the top of some brand-new semantic mountain to launch *something* decent. What they're trying to do is technically hard and thus quite risky, but they've certainly got the right team to try.

Scott presents demo search results

To my surprise, they made a very impressive stab at persuading me about 2) 'Will anyone care' tonight. Though they compared it to World of Warcraft several times, the PowerLabs concept bears the most resemblance to an open source project in general, and the Slashdot community as a particular instantiation. Submissions, votes, feedback, and all the other communal goodies of the open source way, wrapped in a slick interface and presented to prosumers who will give feedback and product advice, tell their friends, blog incessantly, and create a large nucleus of interest and buzz about their new service when it emerges. It's smart. It's also, take 2, very risky. It can fail in a lot of peculiar ways, and it's a brand-new innovation to the company launch problem. We'll see if they celebrate or curse this decision years from now.

I am compelled to add that it's not just the concept that's impressive about the business side of this company. Their PR has been impeccable since launch, their fund-raising was also eyebrow-raising, and this whole evening was handled very well. Steve Newcomb, COO and founder, was calm and impressive in presenting the vision, answering nearly every question, and conveying an absolute sense that this company stands for something, and is striving to do something great. I dare say it's almost impossible to fake that sort of sincerity; I admire them greatly for making this huge bet, and am intrigued to watch whether they can pull it off.

Steve discussing a query

3) remains to be determined. WWGD is a (the?) key Valley question, ignored by any technology company at their peril. The big G is not infallible, but no one has even stressed them yet. Microsoft managed at least two really outstanding strategic pivots (Office in response to Borland, and Explorer in response to Netscape) and web search is Google's game. Tonight, Steve and I had a spirited exchange about whether Google can emulate the results of natural language using statistical methods. While I concede his point that there are edge and corner cases that can't be done by any other method, I maintain that enough core cases can be derived from query behavior analysis to really limit PowerSet's perceived advantage to end-users, and thus their motivation to switch. Steve's ultimate response was, "in any case, we'll have moved search technology forward"; a good answer for the world, regardless of what that might mean for PowerSet's option-holders.

I got laugh line of the night, so I've gotta blog that. Steve revealed some numbers about their indexing time per sentence; he played the crowd well, with repeated, "I don't know if I should tell you... Barney [CEO Barney Pell] might kill me..." before the big reveal. When a minute or two later he said again, "Barney might kill me," I called out, "Who killed Steve?" in best PowerSet query style. Laughs all around.

Summary: This is still incredibly early. The core of their technology just came out of a LAB, for God's sake; they won't have a real search engine for half a year at least, if not longer; and any speculation about whether they'll be better than some multi-billion dollar competitor is kind of like wondering whether Greg Oden will break any of Wilt Chamberlain's records.

They've put together an impressive bet; they're making all the right noises; they could become General Magic, or they could become great. In any case, it'll be fun to watch.

Startup life.

Update: Dan Farber at CNET and Kevin Burton also have commentary.

No comments: