4 ms·

This x100000. There is no scenario - none - where thousands of engineers at Google working on search wake up in the morning and say "we sure have made it good

This x100000. There is no scenario - none - where thousands of engineers at Google working on search wake up in the morning and say "we sure have made it good enough wr2 SPAM. I think I'll have another Danish."

compiler-guy 5y ago
When the cafes were open, you can bet they said, "I'll have another Danish, and then get back to work on this problem that never seems to go away."
- techdragon 5y ago
  I sure wish I had problems that were totally unsolvable, they are so easy to measure progress on. /sarcasm I think it’s more likely that because they are just building hundreds of tiny tweak experiments and it’s someone else who desides what to build and if it even worked. Search quality is such a meta-problem that it goes beyond any real hope of simply working on it in anything beyond piecemeal trial and error fashion on their dataset.
  - siva7 5y ago
    What is a danish?
    - compiler-guy 5y ago
      A breakfast pastry something like a donut. https://en.wikipedia.org/wiki/Danish_pastry https://en.wikipedia.org/wiki/Danish_pastry
ericbarrett 5y ago
I agree with this and the grandparent comment wholeheartedly. That said, there's a kind of institutional blindness that can build up in companies—especially ones that dominate their sector. It may have roots in intransigent upper management, ossified and inflexible process, wide-scale burnout, a culture of passing the buck, or any number of other pathologies. I don't claim that Google has any of these and certainly have no insight into their search group. But I've personally been at powerful companies with best-of-the-best talent that were blind to the decay in their own living room, so I would caution against immediate dismissal of PG's take.
UncleMeat 5y ago
Especially since "made a change that improved search result relevance by X%" is an extremely compelling story for promotions. If indeed there is a launch-driven culture for promos at Google then there'd be extra incentive for new mechanisms to reduce low quality search results.

compiler-guy 5y ago

When the cafes were open, you can bet they said, "I'll have another Danish, and then get back to work on this problem that never seems to go away."

techdragon 5y ago
I sure wish I had problems that were totally unsolvable, they are so easy to measure progress on. /sarcasm I think it’s more likely that because they are just building hundreds of tiny tweak experiments and it’s someone else who desides what to build and if it even worked. Search quality is such a meta-problem that it goes beyond any real hope of simply working on it in anything beyond piecemeal trial and error fashion on their dataset.
- siva7 5y ago
  What is a danish?
  - compiler-guy 5y ago
    A breakfast pastry something like a donut. https://en.wikipedia.org/wiki/Danish_pastry https://en.wikipedia.org/wiki/Danish_pastry

ericbarrett 5y ago

I agree with this and the grandparent comment wholeheartedly. That said, there's a kind of institutional blindness that can build up in companies—especially ones that dominate their sector. It may have roots in intransigent upper management, ossified and inflexible process, wide-scale burnout, a culture of passing the buck, or any number of other pathologies. I don't claim that Google has any of these and certainly have no insight into their search group. But I've personally been at powerful companies with best-of-the-best talent that were blind to the decay in their own living room, so I would caution against immediate dismissal of PG's take.

UncleMeat 5y ago

Especially since "made a change that improved search result relevance by X%" is an extremely compelling story for promotions. If indeed there is a launch-driven culture for promos at Google then there'd be extra incentive for new mechanisms to reduce low quality search results.