Buterin discussed the limitations of the present methods used by centralized exchanges to demonstrate trustlessness and offered solutions as he tracked the development of proof of reserves.
Due to these problems, there were debates in 2013 on how exchanges might demonstrate both the entire amount of user deposits and their ability to cover those deposits through assets. The Merkle Tree method was developed as a result and is currently in use.
However, Buterin highlighted that there are several security flaws in the Merkle Tree algorithm. He pointed out that although it is useful for proving liability, it poses a privacy concern. ZK-SNARKs, however, provide a superior approach for exchanges to demonstrate reserves and liabilities as a result of technological improvements.
In addition, Buterin pointed out that while transferring money from cold wallets to public addresses can serve as proof of assets, doing so can be difficult. The dual usage of collateral and the expense of signing off messages to demonstrate control of an address are indeed the issues.
Buterin also mentioned using Plasma and validiums to stop exchanges from misusing users’ money. But there are drawbacks to this, such as the possibility of losing money if the operator goes missing.
In his closing comment, he noted that while decentralised exchanges have benefits, the CEXes can support in account recovery if a user forgets their password. Buterin also foresaw the possibility of cryptographically restricted CEXes and partially custodial exchanges that only hold fiat.