Senior FullStack Engineer | Blockchain | AI Prev @0xPolygon, @0xsequence (Acquired by Polygon Labs), @biconomy

Bucharest, Romania
Pinned Tweet
I built a DeFi hedge fund in 3 minutes. Lend on @aave → borrow → deposit into @Morpho vaults Route yield from a Yarn vault into $XAUT Tokenized Gold + $POL native staking Route yield from Euler USDT into $MSFT $NVDA and $SpaceX Configure how much goes into each market. Hit publish. Others can copy it. You earn performance fees. They earn based on your knowledge, you manage the strategy. This is what composable finance looks like.
47
32
376
47,853
The whole point of having tokenized RWA's on-chain is to make them composable, things that would never be possible in Trad-Fi, are now possible in DeFi.
Took a while for DeFi to get here, but very cool to see the Magnificent 7 stocks supported as collateral on @aave. More coming soon.
1
87
I curated an automated DeFi strategy that lends, borrows, reinvests, and buys tokenized stocks, in about 3 minutes. Here’s the strategy 👇 USDC → Aave → borrow → reinvest → Lend USDC on @aave → Borrow 70% against it → Put 50% of the borrowed USDC into the @Morpho Clearstar OpenEden USDC Vault → Use the other 50% to buy $NVDA An automated strategy generating ~3.4% average yield, after accounting for the Aave borrowing cost. The Morpho allocation helps offset the borrowing cost, while the $NVDA allocation gives you exposure to a tokenized stock alongside the yield strategy. And this is where it gets interesting. You can publish the strategy for others to use. If someone deposits into your strategy, you can take a 1% performance fee from the profits. So your strategy can potentially generate 3.4% yield + 1% performance fee. You don't give up custody. Users grant the strategy a session key with strict permissions. The key can only perform the actions defined by the strategy. It cannot send assets wherever it wants, withdrawals can only go back to the user's own EOA or smart account. So even if the KMS automation key is compromised, the attacker can't simply drain the user's funds. Strategies are versioned, if a curator changes the allocation: → Users get notified → They can accept the new version → Or keep using the previous version → Or exit The curator can still publish and use the new strategy without being blocked by existing depositors, it will use version 2. Funding the strategy is super easy, you can use any token from any chain through @TrailsHQ, or fund it with fiat. And behind the scenes, the backend finds the best swap routes using @EnsoBuild. Oh and there's no gas, operations are sponsored.
10
204
Vaults can be composable but that doesn't make them safe. Kompose V1 is built on vaults. Here's one of them, an ERC-4626 Kompose vault that spreads your funds across the Magnificent Seven stocks on Robinhood Chain. It buys, holds and rebalances the stocks through @Uniswap, and you get back share tokens you can trade or plug into other vaults. So the composability works. The weak spot is pricing. This vault trades stocks on the open market and values them at the live pool price, and a pool price can be manipulated. On other chains I priced through @Ondo or @chainlink, but Ondo wasn't on Robinhood Chain, so this one runs on Uniswap directly as an experiment. Chainlink is also not fully reliable for tokenized RWA prices yet since they are pretty new. app.kompose.xyz/strategy/cmt…
1
4
180
Another use case for Jev in DeFi, this time on @solana. Vault risk classification for @kamino vaults. I pulled the signals that matter for Kamino vault risk (reserve allocations, liquidation LTV, utilization, idle cash, curator, APY sanity, liquidity, concentration and more) across 30 vaults, then asked Jev to score each one LOW / MEDIUM / HIGH / EXTREME. 30 judgments, ~700ms each, all back in 4s running in parallel. ~65% median confidence, $0.007 in costs. Result mix: 15 MEDIUM - 14 HIGH - 1 EXTREME - 0 LOW. Jev also ranked curators and vaults by trust. @SteakhouseFi looks safest among curators, Elemental riskiest. SOL Balanced (Gauntlet) is the most trusted vault. Claude, Codex and other models can do this as well but it takes much more time and money. The goal is to get to a median confidence of 80-90% and see how results change.
3
7
224
gabidev retweeted
only 10% of tokenized rwas are active in DeFi which means 90% of them are onchain but not really doing anything tokenization was always just step one, composability is the actual endgame
17
4
72
6,111
~5,000 years of iteration on financial infrastructure for Banks while DeFi is 9 years old.
Can someone explain why are so many crypto protocols so easy to hack? I never heard of a bank getting hacked
2
160
In a world with complex permissioned execution via session key policies, don't choose vaults.
The more one builds with vaults, the more one questions “why vaults?” Vaults will always exist to increase composability, reduce transaction costs, and generally increase security But at the same time 4 years ago when vaults became standardized as capital allocation tools we didn’t have mature smart account infrastructure and AI There should be no question if something is non-custodial if the management of something is all purely code (whether public onchain, or verifiable offchain) The sooner we embrace this future the better. Cypherpunk vaults
3
1
4
692
gabidev retweeted
There is a lot I disagree with in this piece. Vaults are not a “necessary complement” to lending markets. There are many ways to coordinate and aggregate liquidity without giving discretionary allocation power to curators. The Midnight example especially doesn’t make sense to me. Fixed rate markets can be built in many different ways. If a market needs a curator to decide which rates, maturities and risks users should be exposed to, while the curator has little skin in the game and depositors ultimately bear the losses, I would argue that’s a bad market structure, not a reason why vaults are necessary. A vault here doesn’t solve the problem. It just abstracts it away from the user. I’m not saying vaults themselves are bad. They can be useful for passive users, especially when allocation follows predefined rules and can be automated without requiring discretionary risk decisions from a curator. But discretionary vaults are not necessary for coordinating lending markets, fixed-rate or otherwise.
5
2
39
4,411
Hey @StaniKulechov, you should check this out. This is one of the reasons I’m building Kompose. You create a strategy as a smart account with permissioned allowances to perform very specific actions. Other users can then deposit into and use your strategy while keeping custody of their own assets. If you modify the strategy, all depositors are notified in real time. They can either accept the new version, continue earning with the previous version, or exit. There’s no need to deposit assets into a vault contract. Each depositor keeps ownership of their assets, while the strategy only has permission to perform the actions it was explicitly allowed to perform. Strategies can also be fully automated using session keys and policies, while keeping the permissions tightly scoped. @aave is one of the main protocols already integrated. The goal is to make DeFi strategies composable, automated, and transparent without giving up custody or control.
This categorisation doesn’t make sense and is pretty much self-serving. First of all, arguing that a vault where a curator has discretion over how capital is allocated across markets, and can even expand into new markets beyond the user’s initial mandate, which, btw, is a known Morpho drawback (very non-LP friendly), should be considered non-custodial simply because it has a timelock is about as strong an argument as a wet European paper straw. Especially the part of implicit approvals that simply changes the whole allocation mandate and users don’t even have the proper tools to monitor these changes. Also the part on relying roles, doesn't really solve much, simply creates a blame game and relocates potentially liability but doesn't solve the actual problem. Vaults that could reasonably be considered non-custodial are those without a manager. For example, vaults that simply wrap deposits into a lending protocol, or the original Yearn vaults. These are make sense to be categorized as non-custodial vaults. There’s nothing inherently wrong with discretionary vaults, as long as the regulatory path is figured out. I’m all for developing industry standards, but let’s at least do it in a way that serves the broader industry rather than your own interests. 😂
1
1
12
433
Here is another use case for Jev in DeFi. Vault risk classification. I pulled the signals that matter for Morpho vault risk (allocations, LLTV, utilization, idle, oracle, curator, APY sanity, liquidity and more) across 50 @Morpho vaults on Base, then asked Jev to score each one LOW / MEDIUM / HIGH / EXTREME. 50 judgments in 22.6s, ~452ms avg, ~60% median confidence, around $0.006 in costs Result mix: 30 MEDIUM - 12 HIGH - 8 EXTREME - 0 LOW. Jev also ranked curators and vaults by trust, who looks safest, who looks riskiest. Claude, Codex and other models can do this as well but it takes much more time and money.
4
3
31
2,886
Would be cool to know when a curator does a change in a vault I'm invested in, weird that this is not a thing right now.
83
It's a good thing that they are now focusing on making models cheaper rather than "smarter".
Introducing Claude Opus 5.5, the first model in our new Claude 5.5 family. It performs at the level of Claude Fable 5.1 for most tasks, and costs 40% less to run than Opus 5.
1
216
Hey @zerodev, running into a slow typecheck with the SDK. createKernelAccountClient takes about 100 seconds when getPaymasterData calls sponsorUserOperation inline. The same functions take about a second on their own. @zerodev/sdk 5.5.10, viem 2.56.0, TypeScript 5.9.3. createKernelAccountClient is generic over the transport, the chain, the account, the client, and the RPC schema. With getPaymasterData written inline, TypeScript infers all five of those from the callback, and the callback calls sponsorUserOperation. That function's client argument is not the named ZeroDevPaymasterClient type. It is a full copy of a viem Client, including a recursive extend over every public and wallet action. Walking that during inference takes about 100 seconds. The same functions typecheck in about 1.5 seconds when they are not inferred together. You have the callback in two files, so the build sits there for about 3.5 minutes.
1
156
I built a @Polymarket trading agent using @typesafeai's new model, Jev. Here is an example of it placing a profitable prediction on the 5 minute BTC UP/DOWN market. The code is open source: github.com/VGabriel45/polyma…
3
7
449
Hey @0xProject why can't I swap into tokenized stocks through your API ? I'm not in a geo-blocked location.
2
1
306
Morpho keeps cooking 👨‍🍳
Stock-backed loans are live on Morpho. Users can lend and borrow against @coinbase tokenized stocks at variable & fixed rates on @base.
207
what 🤯
Today we are announcing that S&P Global has entered an agreement to acquire OpenZeppelin. Onchain finance is growing from an emerging market into core financial infrastructure, and the standards and rails our team and community built are becoming the rails of global finance. OpenZeppelin smart contracts facilitated over $37 trillion in value transferred, with the vast majority of the largest DeFi protocols, blockchain networks, stablecoins and tokenized funds relying on them. With S&P Global, we expect to accelerate the impact of onchain finance, backed by more than a century of trust in global markets, benchmarks, and risk frameworks. To our clients and to all the users of OpenZeppelin open source tools: • OpenZeppelin Contracts and all our open source applications and tools remain open source, free, and publicly maintained on GitHub. Building open source standards stays a core priority. • Audits, engineering work, and ecosystem programs continue with the same team, brand, quality, and customer experience, with what will be the added benefit of S&P Global's research capacity, market data, and institutional reach. For the last decade, OpenZeppelin has set the security standard for onchain finance. Today begins a new chapter for that mission, together with one of the most trusted names in global markets. Read the full announcement: openzeppelin.com/news/spglob…
208
Everyone thinks AI is replacing engineers and causing lay-offs. I came to a conclusion that this isn't true. Ever heard about the Jevons Paradox? This is playing out in real-time with AI and coding, and it's not something new, it already happened multiple times in the past. It looks like AI is killing the programming job. Instead, it’s doing the exact opposite. Here is why the "death of the programmer" is a myth:
1
1
138
5. The Shift in Role AI isn't replacing programmers, it's upgrading them. Software engineers are shifting from "code plumbing" (writing boilerplate) to becoming System Architects and Product Directors. Yes everyone can vibe-code but you cannot trust a non-technical person to build highly scalable, available and secure systems. There is a big difference between vibe-coding and AI programming.
1
14
AI reduces the human effort required per line of code, but it explodes the total volume of software the world needs. If you're an engineer, don't panic. The pie just got much bigger.
10