>>>>> The example we like to play with internally is Halloween. It would be interesting if you were to see who are all the people posting Halloween costumes, and what’s the most popular costume. That would be an interesting use case if we had really great neural networks and machine learning. But right now we’re mostly using the textual content of people’s captions and their accounts.
Machine learning's most brilliant application to date
Articles like this make me think of a long lost period where everyone would listen with baited breath about the technical wisdom of Digg. Digg had a compelling product, and succeeded in spite of profound technical issues and ignorance, but invariably they became "leaders" of technology because of the position of the product (the Digg engineering team had front page after front page on here). Instagram -- and they are hardly alone -- falls in the same camp as Digg to me: This article details the most banal achievements, and would never, in a million years, be on HN if it weren't from a lauded product.
But here it is, in a long line of "Instagram did pretty basic things" articles.
As an aside, I've always carried a chip since Instagram did as so many Valley companies do and waved their hands claiming things weren't possible on Android. Since then I discovered Gallus, which is a vastly superior product than Instagram's offering, and makes none of the same excuses. A single guy did what Instagram wasted so many words claiming impossible.
I think part of the value is that it is Instagram that's doing it. It's reassuring to know that some method is actually being used at scale by a service with millions of users, rather than having to rely upon the experience of someone who merely knows the theoretical side of things.
Instagram's database sharding post was reasonably detailed and good at explaining things. Yes, the same info is probably on stackoverflow, but I'll waste a lot of time there trying to understand untested opinions.
I remember when Instagram had to shard their database to keep up with demand, and the article was so popular among developers. "Check it out, they split up the databases! Whoa!" I'm not trying to dismiss their work, but it was a relatively ordinary engineering feat.
We (at IG) aren't claiming to be doing revolutionary things on infrastructure--but one thing I found super valuable when scaling Instagram in the early days was having access to stories from other companies on how they've scaled. That's the spirit in which I encourage our engineers to blog about our DB scaling, our search infra, etc--I think the more open we are (as a company, but more broadly as an industry) about technical approaches + solutions, the better off we'll be.
This is exactly right and further I think comments like above scare a lot of people out of posting information that may be useful to others. I'm quite surprised by the sourness. Thank you for sharing!
Thanks for sharing your stories and lessons learnt. Also appreciated is your explaining in the context of the larger Fb sys infra and the services instagram engg. deliberately decide to rely on.
That is the right attitude and sorry, don't mean to dismiss that either. I was just kinda piggy-backing the OP's story with a small tid-bit that I found funny. It's more about poking fun at the "wow!" reactions than it is directly at you.
Machine learning's most brilliant application to date