You suggest they lay off 90% of their employees?

No, just a hair over three quarters. Anyone can see that's perfectly reasonable.

(...he said, not realizing it was on the close order of what someone else here is advocating. Whence comes the idea that Twitter is overpopulated? Do we hear it from current or ex-Twitter-ers?)

Not going to argue with you. 4000 employees is going to lead to a lot of inertia when doing development (even if only 10% of them were devs). However, the culture and process changes you need to institute in order to make such a change work is not trivial. Probably necessary for the future growth of the company, but could quite easily end up killing it in the process.

Not asking you to argue with me. I'd just be interested to see some substantiation for the claim that Twitter is overstaffed. It seems like that claim turns up a lot, and I'm sort of surprised that no one seems to have taken the time to explain why he thinks so.

In my case, it's merely a matter of having worked in large organisations before. The communication overhead is too large. If you have 400 programmers, that's at least 40 teams. If they are all working on the same product, then you have a significant problem with communication.

I once worked on a product with 5000 programmers. Average production was 1 line of code per day per programmer. While LOC may be a poor measure of productivity, at the 250 lines per year level, you've got a serious problem no matter how you look at it.

With those numbers, we were churning 1.25 million lines of code a year, which obviously could not be done without a very large team. But the cost is crippling. Not only that, but there is just no way in hell you are going to hire 5000 good programmers. Deadwood is going to be a massive problem.

What's the solution? Simplification, prioritisation and diversification. If you have to churn a million lines of code a year to say competitive, then it may be too late. Before it gets to that point you need to simplify your code base and throw away the things you don't need. You need to think very clearly about what makes money and what doesn't and you need to have clear priorities for how to downsize your engineering work to a point where the costs don't eat the profit. Finally, once you've got a handle on that, you can build your business by diversifying into separate business areas -- there is no problem churning a million lines of code if it is not all in one product. 10 products, churning 100K LOC each will be much more manageable. But probably you will be much more profitable than that if you manage 10 products each churning 20K LOC per year (on the assumption that a simpler product will be easier to monetise).

Again, LOC is not a reasonable metric, I'm only using it to indicate approximate scope in terms that most programmers can understand.

I don't know how many programmers Twitter has. Like I said, even if it is only 10% and the rest are ops, sales, etc I think they have got problems. TBH, I have trouble imagining Twitter utilising 3000+ sales people effectively as well, but that's not my area of expertise so I will keep my mouth shut.

The claims come from current and ex employees.

Basically, in the least offense terms, doing nothing all day and got nothing to do.

Jesus Christ. Yes, perfectly reasonable.