Comment by cs702 | Hacker News Reader

Comment by cs702 | Hacker News Reader

Musk is right in wanting "high velocity of features," i.e., the ability to evolve Twitter faster. How deeply he understands the technical details does not matter that much, as long as he has capable technical people working with him. It's likely that they are the ones telling Musk they need to rewrite big chunks of Twitter's infrastructure. Given that WhatsApp was able to grow to ~500 million users with only 32 employees, it's quite possible they're right.

I'm reminded of this confrontational interaction between a technical person and Steve Jobs: https://www.youtube.com/watch?v=oeqPrUmVz-o -- in hindsight, Jobs was right: "You've got to start with the customer experience and work backwards to the technology," not the other way around.

In Twitter's case, they've got to start with what's important -- being able to try out a lot of new ideas quickly and cheaply to find out which ones work. The exact details of how they accomplish that are of secondary importance.

> Elon is right in wanting "high velocity of features,"

He's proposing a ridiculous solution that will have the exact opposite effect of what he wants.

The only thing in common with the Steve Jobs clip is that both interactions are somewhat confrontational. Jobs is talking about how the user experience is important and engineering is secondary, Musk tries to solve Twitter's problems by asking the engineers to rewrite things that already work fine.

So the current stack supports a high velocity of features?

Geohot mentioned 3TB of codebase and resigned basically because he wanted to refactor shit while elon wants featurezzzz

At Twitter scale. What stack would? Every new feature needs to work for 200m people in 100s of languages, across devices and browsers. Tell me what stack can support uncautious throw anything at the wall strategy. Hell tell me one team of 20 people that can do that.

> ... that will have the exact opposite effect of what he wants.

You don't know that.

Time will tell who is right!

> Musk is right in wanting "high velocity of features

The desire for velocity isn’t the part being criticized, Musk's proposed approach to that uncontroversial goal and the evident conplete lack of understanding on which it is based is.

> How deeply he understands the technical details does not matter that much

It does, given that he is very hands on on the tech decisions and has proposed handing over the CEO job while retaining control of “Software and Services”.

I wonder: Does he (a) truly believe he knows anything about tech, or (b) desperately try to maintain what he knows is a fiction, for his brand of the celebrity wunderkind?

If I assume he's smart, I'd think it's (b). But to do (b) well, he probably would have experts tell him what to say.

I remember him talking about how Hyperloop is an air hockey table in a vacuum - I'm not a physicist but that doesn't seem to make sense. I know that what he says about software makes zero sense. Does nobody tell him that? Does he really not see it?

It's important to remember that Twitter is not, in fact, Whatsapp. A person-to-person messaging app is going to be far less complex than a global social network.

Bringing up "Whatsapp only had 32 employees" seems like a lazy point to make to handwave away any actual arguments.

That's true. But: If Twitter was built using Whatsapp's tech stack (Erlang, which they've let go in the meantime) then chances are that they would have used a lot less staff than they did. It is extremely well suited to this kind of problem.

Eh... I don't really see it. Erlang is good for the WhatsApp problem; shunting binary blobs around between lots and lots and lots of clients, and maintaining some sort of shared state of what clients exist and whether they're online or not.

This is not really a problem that Twitter has, though. Or at least not a significant part of Twitter's problems; they do have to do some web-push stuff, and sure, they could use Erlang for that, why not, but it's not a big part of their problems. And as for the rest of Twitter... well, Erlang is notably bad at handling _text_, for a start.

Reply on news.ycombinator.com