Conversation

Having gone through the process of launching centralized and decentralized products, I can confidently say that it's way harder to release the typical CEX than the typical DEX these days.

@arjunblj

-- in which sense were you comparing it to a rocket?

Quote Tweet

Arjun Balaji

@arjunblj

Oct 14, 2020

Releasing a DeFi application today is closer to launching a rocket than launching a consumer mobile app.

Replying to

and

What I meant: 1) Low/no margin for error, emphasis on FV/testing 2) Structural constraints which limit ability to patch or iterate post-launch But also a shitpost so I wouldn’t read too much into it

Replying to

hehe got it agreed on the 2nd, not really the 1st--it's pretty simple stuff all things considered, testing less necessary than for most things though obv the immutability pushes in the opposite direction. Audits are frankly overrated.

Replying to

and

For things that are temporal in nature like an orderbook, sure, audits can be overrated. For things like a multisig, where funds are deposited and just sit there for an unspecified period of time, audits are absolutely necessary.

Replying to

and

People can, and *will* hack you. The only reason they haven't, yet, is the unaudited product is not valuable enough.

Replying to

and

I think more of my point is that I think audits are not always the best way to catch bugs -- that either: 1) low-value version people play around with for a while 2) fuzzing 3) having another dev read the code 4) testnet are often better. But I'm not 100% sure in that!

5:11 PM · Oct 14, 2020Twitter Web App

Replying to

and

But this also really depends on the protocol -- some are way more complicated than others, and the more complicated and novel ones need way more testing

Replying to

and

Those are all important, especially for complex things like an operating system, but they don't come close to finding all bugs. We need more. Blockchain programs are trivial in comparison. Ideally we would have a formal proof of correctness, an audit is a (flawed) approximation.