Wei Dai
@_weidai
Followers
15K
Following
5K
Media
235
Statuses
3K
Research Partner @1kxnetwork. CS PhD @UCSanDiego. Research & investing for the future of crypto. Alpha in highlights. Personal opinion, not investment advice.
Joined November 2021
1/ Why I'm excited about investing in crypto infra 🧵. Crypto and trust-minimizing technologies will transform foundational infrastructures across industries. As an ex-academic and now investor, I strive to find, fund, and partner with 0-to-1 innovators and 10x disruptors.
58
15
320
TL;DR of Matt's response on why Tempo is an L1: the desired product spec for the payment chain is easier to build out as an L1 instead of as an L2. Lesson for infra projects: your target ecosystem projects should naturally use your infra/platform if they reason from first
What apps and infra should takeaway from the rise of application-specific L1s like Hyperliquid:. Product specs of apps determine the required infra. What it mean for apps: build your own infra if none on the market works for your intended product spec. Yes, you have to have a.
17
11
126
Fragmentation of chains is great for interoperability protocols. Now that Tether, Circle, & Stripe are all building different payment L1s, it is increasingly unlikely that we will have one globally-dominant payment L1. As a result, interoperability between payment chains.
77
28
323
What happens if a major hack (like the Bybit hack) happens to an RWA asset instead of Eth? Which part of the stack should be involved in undoing the damage of the hack, if any?. I believe the truth to the question lies somewhere between both @malekanoms and @CampbellJAustin's.
LIVE NOW - Debate: Is Ethereum Ready for Real World Assets?. @CampbellJAustin (NYU) and @malekanoms (Columbia) debate a simple question: is @ethereum ready to host real-world assets? . They walk through stress tests—what if an exchange is hacked, a stablecoin breaks, or a court
2
0
8
Decentralized CR layer + centralized but trust-minimized sequencing (via TEE) is perhaps the best response from the Ethereum ecosystem to Solana's ACE.
Enshrining Rollup-Boost with censorship-resistant auctions on pod. L2s suffer from spamming by MEV bots, making them slower and expensive for users. Rollup-Boost with pod solves this problem using 200ms censorship-resistant priority auctions. Here's how it works👇 1/5
9
1
26
I call this application-controlled bridging (ACB). Canonical (token) bridges for general-purpose L2s are (1) complex, (2) high latency, and (3) permissionless. Whereas an application-controlled bridge can be designed to trust specific L2 application logic (e.g. matching and.
@gluk64 One main difference here (and this is from an app UX standpoint), is that if you have a yield-bearing vault that, even when tokenized, is not particularly liquid (e.g. due to withdrawal requirements or non-atomic deposit/withdrawal semantics), then you can't "deposit" it directly.
2
0
15
Having academic research funded by VCs is an interesting idea. Yes, it'd help with funding certain types of research, but it is not a complete solution to the funding shortfall for scientific research. As a VC, I'd fund research in either Pasteur's or Edison's quadrant, but not
what if?! college labs were VC-funded. today: labs get funding cuts, professors chase papers, students grind on la-la-land problems. imagine instead:.– professors act like executives.– top students are hired like early employees.– every breakthrough has a pre-signed path to.
13
1
29
The first step towards wide adoption of onchain privacy is for more people to understand the nuances of different levels and types of privacy.
the four levels of blockchain privacy:. 🪟 transparency.🥷 anonymity.🌫️ confidentiality.🥷 🌫️ total privacy. let's break it down ↓
9
4
38
What gold, Bitcoin, and Eth have taught us is that as an asset matures into a store-of-value asset, the demand for transactions in the asset in its native form (physical gold, and BTC/ETH on their respective L1s) actually decreases. - Physical gold transactions are infrequent.-.
27
3
44
I will take the opposite position on FOCIL: FOCIL is good for Ethereum and should not increase liability for running a block proposer. Since validators have no choice but to include transactions from FOCIL without forking the chain, Inclusion of FOCIL transactions effectively.
The case against FOCIL (which I learned about today):. ETH devs, I love you. You mean well. But when you create an EIP to solve the problem of "filtering out transactions with sanctioned addresses" and your solution is "to allow validators to impose constraints on builders by
7
1
26
The gap between tech & product traction vs. price action is the decoupling that many have painfully missed this cycle. Be patient if you are here for the tech and real value creation.
6
0
18
Great introductory/first read to understand the scaling bottleneck that Optimum is solving.
Ever wondered why your transaction feels slow?. Our DevRel @SinhaSwarnabha has the answers and an excellent breakdown of OptimumP2P's solution to the blockchain gossip bottleneck 👇.
9
6
33
Onchain privacy hasn't reached product-market fit because the product wasn't right, not because the market is not there. The product (past onchain privacy solutions) wasn't right primarily due to:. 1. Lack of zero-compromise privacy. Privacy cannot come at the cost of access to.
79
13
195
Grouping Hyperliquid trading fees together with general-purpose L1 base & priority fees is simply misleading. (Many revenue reports do this currently, like the dashboard blockworks.). The equivalent of Hyperliquid fees for general-purpose chains would be measuring chain & app.
If Hype trading fees are blockchain rev (per this chart), is BNB on top w buybacks of ~$1B BNB quarterly? What about unichain/uniswap?.Let's leave blockchain fees as what people pay for txs. July L1/L2 fees: ~$200M (based of 120 of 450+ L1/Ls). HyperEVM: ~0.15M (see @Blockworks_).
6
0
18
When there's a gold rush, you sell shovels.
8
5
55
No company would dare to start their own network competing with the internet. The only way for crypto network to replicate this dynamic is to build network effects so strong that the opportunity cost of not joining is high. No L1s today can do this. We are still early.
33
3
107
0
13
0
Yes, crypto infra has ways to go to match the performance of tradfi systems. But we have the tools, and they are improving fast: ZK, erasure-coded multicast, low-latency orderless consensus, application-specific sequencing / application-controlled execution, . We will see.
@solana Nasdaq's SIP is designed to handle 10 million operations per second, with a latency of 0.02 milliseconds (though realistically, constrained by speed of light, it still gets to 0.2 ms between NYSE or Cboe and Nasdaq), and is readily parallelisable for infinite throughput. (There.
15
14
61