In 2 days from now, a big safety upgrade will be executed on @aave
- A more resilient safety model for Aave
- Possibly 12% APY on stablecoin staking, 6% APY on ETH
8 things you should know about Aave Umbrella 👻🧵
1️⃣ What's Aave Umbrella?
A new system that lets you stake aTokens - the interest-bearing tokens you get when lending on Aave (aUSDC, aGHO, aWETH) to earn rewards
In return, you take on risk: if borrowers don’t repay and Aave ends up with bad debt (deficit), your staked aTokens can be slashed to help cover the loss
2️⃣ Why is Aave Umbrella better than Safety Module (the current system)?
Safety Module: Users stake $AAVE or LP tokens
=> Not ideal because if Aave has a shortfall in stablecoins like USDC, it has to sell the staked $AAVE to cover it, causing sell pressure and delays.
Umbrella:
- Users stake aTokens, which are already the right asset type. If there's a loss in USDC, aUSDC can just be burned. No need to sell tokens
- Slashing is automatic and based on smart contract data. No need to wait for a governance vote
=> A faster and more resilient safeguard for Aave
3️⃣ How Aave Umbrella works:
1/ Users stake wrapped aTokens (like wrapped aUSDC) into Umbrella 2/ Earn staking rewards + regular Aave aToken yield (in either GHO, AAVE, USDC, etc.) 3/ Stakers accept slashing risk, only on the asset and network you staked 4/ Cooldown required before withdrawing (20 days + 2-day window)
For example, if you stake your aUSDC on Aave v3 Ethereum, you'll earn the base yield + extra Umbrella aUSDC rewards
If USDC deficit appears on the Aave pool, you are only accepting slashing risk for borrowed USDC on that pool in Arbitrum
4️⃣ How will stakers be protected?
Aave introduced a “deficit offset” for each asset. It’s a buffer.
The Aave DAO treasury covers first, before stakers are touched.
Example: If USDT’s offset is set to $100k, then no aUSDT staker gets slashed unless the bad debt on USDT exceeds $100k.
So, it’s a shared defense: DAO takes risks first, and rewards stakers for being next in line.
5️⃣ How are rewards calculated?
Umbrella uses a smart reward curve, in which:
- A target liquidity and a maximum emission rate of rewards are set
- Rewards increase if the staked amount is below the target to encourage more staking
- Rewards softly decrease if staked capital goes above the target
- The curve will smooth out reward emission, avoiding crazy APYs early on
This makes the system more predictable and sustainable.
6️⃣ Is there a risk?
Yes, there’s slashing risk, but it's designed to be rare and small
Example: In a recent month, Aave only had $400 in deficit, across $9.5B in borrows. Umbrella will handle these kinds of tiny deficits automatically.
Plus, rewards + lending yield usually far outweigh any potential loss
7️⃣ Who controls Umbrella?
- DAO decides which assets and networks Umbrella supports
- DAO sets the reward parameters and manages funds
- Slashing happens automatically, based on pool data
- UI is open source - anyone can host and run their own interface
8️⃣ I'm a current stakers, what should I do?
- stkAAVE/stkABPT: you can chill now. But over time, they will be phased out once Umbrella reaches the target liquidity
- stkGHO: will move to the new Umbrella version from day 1. The switch should be simple
As a delegate on Aave DAO, we fully support the Umbrella upgrade.
The proposal passed with unanimous approval, 100% in favor. Execution is set for June 5th.
Want to amplify your voice in future proposals?
👉 We’d appreciate your delegation to our address (represented by @DefiIgnas): 0x3ddc7d25c7a1dc381443e491bbf1caa8928a05b0
From data and training infrastructure to frameworks and launchpads, these projects are cultivating a bullish ecosystem for Crypto AI agents.
Let's dive into details 🧵
We spent 5 hours scanning and studying each project building in Crypto AI and Agentic AI so you don't have to.
Don't forget to bookmark this thread or RT for future reference! 🤌
🧠 AI Agent Frameworks:
- @ai16zdao's ElizaOS: multi-agent simulation framework based on Typescript
- @GAME_Virtuals: low-code toolkit, modular agentic framework by Virtual team
- @0xzerebro's Zerepy: open-sourced agent framework based on Python
- @arcdotfun: rust-based rig framework for AI agents
- @aetherdotfun's AetherOS: a framework to build agentic ecosystems and swarms that can work, learn and adapt to each other.
- @ReiNetwork0x: dual-layer architecture to connect AI and blockchain
ETH mindshare hit its rock bottom in 12 months (by @_kaitoai)
CT's attention has been dominated by "AI agents," which account for over 50%.
ETH hasn't broken its ATH for 3 years.
Is this the bottom signal? Will 2025 be the pivotal year for ETH? 🧠🧵
First up, let's wrap up why ETH so underperformed in the last year:
1. Competition from other blockchains like Solana and Base 2. L2 explosion and blob space with its Dencun upgrade, making ETH less deflationary. 3. Staking mindshare fades away with the point meta 4. The launch of Ethereum ETFs in the U.S. was initially expected to boost ETH's price, but ended up a disappointment. 5. Market sentiment changed quickly, Ethereum is experiencing an "identity crisis," with doubts about its potential as a store of value or internet money.
All led to a dead end: Lower network activity
Ethereum has struggled to regain its peak activity levels. Weekly transactions, which peaked at 11M in May 2021, have plateaued at around 9M.
Similarly, active weekly addresses have remained range-bound between 300K and 400K.
- Bitcoin passed $100K
- Total crypto market cap hit $3.91T
- Stablecoins' supply surged to $204B
- Spot BTC and ETH ETFs reeled in +$100B in under a year
- The U.S. ushered in its most pro-crypto Congress yet
Your 2024 Crypto Roundup 🧠
1/30
Let's go through some top topics around 2024:
- Macro Trends
- Crypto Policy
- Institutional Adoption
- Rise of Solana
- Memecoins
- Fundraising
- User Adoption
- DeFi
- DePIN
- Crypto x AI
Before we dive in, don't forget to bookmark this thread for further reference.
2/30
🧠 Macro Trends
- Legacy Markets: Despite fears of recession, inflation moderated, and the U.S. economy showed resilience.
The Fed began cutting rates, and the S&P 500 delivered a remarkable ~27% YTD return.
AI-driven productivity and a surge in gold demand became key macro themes, while energy prices remained stable despite geopolitical tensions.
$PENGU is the official token of the Pudgy Penguins NFT collection, featuring 8,888 adorable penguins.
Launched in 2021 on Ethereum, Pudgy Penguins initially faced challenges, but eventually have become the biggest wave of crypto adoption to the world.
• Acquired by @LucaNetz in 2022 for 750 ETH
• Brand revitalized → Pudgy Toys introduced at Target, Walmart, and Amazon
• Now the 2nd largest NFT project by market cap, just behind CryptoPunks
AI swarms—autonomous networks of agents working together across decentralized systems—are quickly becoming the next major shift in Crypto x AI.
What should you know about them? 🧠🧵
2/ Why AI swarms?
When AI agents work solo, their abilities are limited—they can't share resources, adapt, or tackle complex tasks.
Instead, picture a team of smaller specialized agents working together.
Each one handles a specific task, and as a team, they solve problems way more efficiently.
3/ How do AI swarms work?
These agents share data, learn from each other, and even adjust roles depending on the situation. It’s like a group project, where everyone actually pulls their weight.
Here are some interesting developments around AI Agents. The next wave is getting bigger 🧵
1/10
Solana-based AI Agents are bouncing back to their peers on Base.
The first AI Agent created by @ai16zdao, @degenspartanai is about to launch, and it is considered as @aixbt_agent on Base
The bullish point is that 8% of profits generated from ai16z fund will be used for $degenai buybacks, creating a flywheel effect as the ai16z ecosystem grows.
2/10
Fully Autonomous DeFi Agents driven by @modenetwork
Mode is the first L2 focused on AIFi with their AI Agent Stack, bringing AI agents into DeFi to autonomously trade, yield-farm, lend, borrow, and perform other complex actions.
Their AI Agent App Store is worth checking with some agents like ARMA (by @gizatechxyz) and Oladius (by @autonolas)