Hacker News Reader

Comment by chris_va | Hacker News Reader

Hm, let's not have personal attacks here, especially for a topic on which you do not have any personal experience.

Having worked on Google Search during this period, I can say definitively that your comment is mostly incorrect. Features at Google are generally bottom-up driven. Mayer was in many ways acting as a gatekeeper to keep some of the crazier stuff from harming the experience or brand.

View on HN

If I'm attacking anything, it's her record as a leader. We can agree to disagree on that, but I see nothing personal in my comment.

Secondly, the heart of my comment is how some people are raised up to insane heights on mediocre achievements. Even if she was a great "gatekeeper" as you say, there's really nothing in that skill that sounds particularly impressive and certainly not worth the obscene amounts of money and praised heaped upon her up until her public fall from grace.

And while I'm at it I suppose I should chime in on Instant - always seemed like a waste of resources to me. What is so hard about hitting Enter? Personally it always annoyed me. Obviously many people will disagree with me but when I found out that one of the most annoying changes to google search had been pushed hard by Mayer I couldn't help but comment.

Well, there was a lot of bottom-up work, but in this case Instant was pushed by Marissa. A lot of infrastructure had to be built, scaled and tuned before launch, so that the demos could actually serve the millions of users, their wonky browsers or ISPs, etc.

Also worked on Google Search during this time period. Wasn't directly on the Instant team, but I did some work on it (it was a big enough project that it touched basically everyone in Search). The folks who originated the project were org-chart cousins of mine (i.e. their boss's boss was my boss's boss), and I later ended up serving under their manager.

The real driver for Google Instant was Larry. He saw a demo that had been floating around and said "I want that." When one of the founders wants something, they tend to get it.

I agree that she supported the launch, but Marissa wasn't involved in any of the infrastructure work, machine planning, etc.

Of course she wasn't involved in making things work efficiently and reliably, since that never was her responsibility, but the whole reason all of that had to happen was because she decided to turn what normally would be one of the many, many cool internal demos into a public, differentiating feature (and got signoff from upper management). I'll miss Instant on desktop, for the record. It was a great achievement to implement in a relatively short period of time.

I think you're being slightly unfair here in assigning credit/blame/what-have-you.

I agree with you that she liked it, and signed off on the product. However, we had to scale the infrastructure 7x. That wasn't a VP level of money, that was a BoD level of money. The decision to push the launch was made by Eric.

I wouldn't quote numbers (or multipliers) here, but my recollection is that, after a lot of tuning, it was lower than what you mention, at least in the section of the stack that I was involved with. Of course, the board had to give some kind of sign off (which is what I meant by upper management), before even knowing the full price :-), but the case for this differentiating feature was being made by Marissa. It wasn't merely signing off. I don't remember anyone ever bringing up Eric when talking about it. If anyone, it was her, for good or for bad. She stuck with it even if early iterations were clunky, slow or inefficient. I'll grant her that she started with something that I thought unfeasible and almost a pipe dream, eventually seeing it turn into something actually useful, with the help of hundreds (Eng, PM, UX, etc.).

(Reached reply depth)

7x infra to support it...what was the value prop for this?

I have no idea where that number comes from. I seem to remember that a naive implementation would have required 10x or so the existing resources (proportional to the characters in the average query at the time... We know they've been getting longer over the years). But that's not at all what was released. Also remember that this is just the serving path. Crawling is not affected, while for indexing...it depends.

My service did not get 7x the resources for launch. We did bring up more capacity, but we also improved things here and there. That's work that would have been needed anyway.

Last but not least, there were infrastructure improvements that helped reduce resources in search and other products. Bob Wyman's LinkedIn page says that his team helped with Instant, but also saved AdSense 100k+ cores, while increasing serving capacity and reducing latency.

As for the value: speed, differentiation and sheer coolness.

She was the public face of this work:

https://googleblog.blogspot.com/2010/09/search-now-faster-th...

I wouldn't have expected a VP to be involved in infrastructure, and machine planning.

Reply on news.ycombinator.com