Death to MRGN. Long live marginfi.
The final chapter is finally here.
This is for users who have followed marginfi since its very first chapter. If that’s not you, you can skip this article.
I previously cofounded marginfi, and I’m currently the founder of Project 0 – the first trustless DeFi-native prime broker. Project 0 is powering marginfi development now, and marginfi is one of the venues in Project 0’s cross-margining engine.
This article slightly pulls back the curtain on some of the product related hurdles MRGN faced over the last year. I feel it’s important to share this with users because a) they’ve asked for it, b) it was counterproductive and often impossible for me to walk through ops/product/structuring/etc related problems publicly before any meaningful step forward was live, and c) there’s a ton of confused people and misinformation out there, so this can at minimum act as a reference for those people.
A year ago MRGN was in a hole. The team was understaffed, suffering major reputational damage, had severe burn, had no meaningful revenue, and the team’s data / risk / monitoring / oracle systems were broken. My previous cofounder had just left in the middle of a public meltdown, insulting hundreds of thousands of users, causing panic leading to withdrawals, and gut punching the employees who had put in tireless hours to make MRGN the fastest growing DeFi product of 2023. Worst of all, users were upset with MRGN's execution. Things needed to change.
The MRGN team faced a plethora of operational, legal, and product challenges from that point. These obstacles are important, and explain a lot of the timing with Project 0, but aren’t the focus of this article. I feel it's important to focus on the product and give users more details on the questions they’ve been asking over the last year. This article will be my attempt to do that as briefly as possible, without any drama.
Looking back at MRGN, the highest impact fires were around data pipelines, platform monitoring, risk systems including oracles and liquidators, and the very small team. To quantify this a bit, MRGN’s broken data pipelines were burning $300K+/mo. The team had broken monitoring and redundancy on its internal liquidators, and oracle price data wasn't landing, causing users to have funds temporarily stuck on marginfi. Perhaps most preventable: MRGN was a team of 6 people building systems to permissionlessly handle hundreds of millions of dollars in an on-chain, adversarial environment. Hiring was desperately needed.
Fixing these problems with the scale of liquidity and users MRGN had, from the ground up, would not be easy or fast. However, it was do or die work.
At Project 0, we got to work. In order of priorities, over the last year we've cut infrastructure costs (oracles and data) from over $300K/mo to about $2K/mo. For those wondering, yes continuing at $300K/mo would've bankrupted MRGN in a very short amount of time.
We fixed MRGN’s internal liquidators. A year ago, MRGN had piss-poor visibility, monitoring systems, and redundancy on its liquidators. The team was extremely reliant on its incredible community of profit-motivated external liquidators. On a network like Solana, which has been going through upgrades at an extremely fast pace, this is a dangerous position to be in. I won't detail the sleep I lost over this, but solvency is everything – and MRGN’s position on protecting solvency with its own internal liquidators was weak.
Fast forward to now. Project 0 is about to release v2 of its own internal liquidators, we're about to innovate on how liquidations happen on marginfi & Project 0 (and lending primitives as a whole), and we're introducing new novel approaches to liquidations – custom to each asset – that will unlock further capital efficiency, improve risk, and better serve the user experience. Project 0’s data pipelines around these systems are stronger than ever and improving more every day.
Another blazing fire at MRGN was oracles. Anyone who actively uses marginfi knows this. This problem was full-stack – from the Solana protocol itself, to our transaction landing systems, to our monitoring and data pipelines. As of today, we’ve rolled out our answer to this full stack problem on P0. Our oracles now use multiple differentiated feeds, smart price weighting across those feeds, have close to 100% uptime, and are cost-efficient (from $20K/mo to essentially $0). Pricing systems are everything – they're a first line of defence for our risk systems. Now they're redundant, performant, and dynamic – and we're not even close to done.
Then, there was the small team. MRGN was understaffed, spread thin, and didn't have dedicated attention to critical product areas needed to grow. Making this harder was the PR damage MRGN was suffering. I had to go out and convince talented people on our mission at P0, on our new product, and on what we saw differently. Despite the noise, we did just that. Over the last year we’ve hired 11 world class people. Project 0 now has a dedicated, differentiated focus on risk, applications, program, backend, and ops. These teams have rebuilt every part of our business on a scalable foundation. We needed to spend time going backwards in order to go forwards 10 times faster. We fixed our team, and we're finally in a position to accelerate.
I want to emphasize how delicate this work is at scale. These moves severely hindered MRGN’s velocity over the last year as we deployed a new team and resources at these problems. This isn't sexy work, there aren't any impactful community updates we can make about it, but it's laid a foundation that will enable us to move at a compounding speed moving forward. I'm extremely excited now that we're turning the corner. The product we're releasing on the back of this infrastructure is incredible.
Looking back at MRGN points, they were introduced on top of brittle data pipelines, when Solana was dead, when points systems weren't a thing, when there were no expectations around what points would become, and with no concrete gameplan on what to do with them. The MRGN team thought quantifying user activity was important, and in the interest of transparency, decided to make that public. They started with a very primitive form of benchmarking user activity, but that drew a ton of attention and blew up before the team could appropriately react.
To be clear, a part of the reason MRGN points began was certainly to understand user behavior heading into a token event. These things aren't directly related, but points were supposed to give the team a solid foundation on identifying users who were lending capital and obtaining credit. There's a plethora of things not included in the MRGN points calculation that the Project 0 team finds valuable, such as people who mint and use LST, or people who tried beta products. The MRGN team assumed they could quickly pivot to building data pipelines that captured other strategic user activity they cared about, but – as I explained in Introducing Project 0 – the old systems weren't in a place where points could update quickly.
Putting aside all operational and legal hurdles we faced, even if we wanted to finish points and launch the token, we didn't have the bandwidth or resources we needed. There were pressing product fires that needed to be dealt with and a team to build.
So, over the last year, we've built a new product (Project 0) and we've built a new team. This is unsexy, heads down work, but it's enabled us to get to this point. Front-running this difficult work with updates about the token or product, given we knew it wouldn't be ready for many months, was counterproductive. Crypto users are tired of being told "coming soon". They want products that matter yesterday – and to their credit, we and the industry at large need to adapt.
To marginfi users everywhere: try Product 0, it’s made for you. You can also keep using marginfi – the P0 team will be refining and improving the marginfi program and dApp for all users.
And to the few who know the depths of what we overcame and were supportive throughout it all, my unending gratitude goes out to you.
Welcome to Project 0.