** Correction on key compromise **
A week ago, Drift moved to a new multisig, created by a signer from the old multisig. This signer did not add themselves to the new one.
The exploiter also initiated the proposal in the old multisig to hand over admin control to this new wallet.
Of the 5 signers on the new multisig, only 1 came from the previous setup; the other 4 were brand-new.
The wallet was set with a 2/5 threshold and a 0-second timelock.
~Five hours ago, that sole carryover signer used the new multisig to propose changing Drift’s admin.
One of the new signers co-signed a second later, instantly meeting the 2/5 threshold.
With no timelock in place, the transaction was executed immediately.
** Note **
Some of the relevant Solana programs are not verified, which limits full analysis.
We're continuing to dig into the onchain data and will publish a more thorough post-mortem covering the multisig migration, Solana DeFi contagion, and vault exposure in a follow-up.
1/ Drift's admin key was compromised.
$213M+ drained from @solana's largest DEX in under 10 seconds.
Unfortunately, we've seen similar patterns before:
- fake collateral market
- a manipulated oracle
- disabled circuit breakers
Let's break it down 👇
written w/ Chaos AI
Friday could have one of DeFi's worst black swan events.
On paper, over $19 billion in liquidations hit across CEXs and perp DEXs. The real figure was likely higher. Many were surprised that on-chain lending markets came out mostly untouched, but outcomes can be misleading.
Markets saw violent dislocations:
PAXG at $3,600 on @binance
EURC at $1.085 on @coinbase
USDE at $0.65 on Binance
These were not depegs. A mispricing on a single venue is not a structural loss of parity.
However, if those gaps had remained open, the price impact would have spilled over into DEXs and lending markets, draining liquidity and triggering liquidations faster than the system could absorb.
On @aave alone, $4.8 billion in positions were at risk, with $180 million in potential penalties. If those liquidations had cleared, assets would have gone to zero before stabilizers could react.
We avoided that outcome because of risk management deliberate design choices. One of them was the decision to price USDe 1:1 with USDT (s/o to @LlamaRisk too!).
The Limits of Traditional Oracles
Traditional price oracles are a pale version of what a true oracle should be. They aggregate quotes from secondary markets and assume those order books reflect fair value. That assumption works for liquid, fungible assets. Not for the new class of asset-backed and mechanism-dependent tokens that now dominate DeFi.
For tokens like:
- USDe
- LRTs
- PTs
- tokenized RWAs
Value depends on primary liquidity, reserve composition, and redemption mechanics. Treating all tokens as homogenous ignores these embedded dependencies and redemption paths that define their real economic worth.
A real oracle must go deeper. It must integrate every structural layer. Reserves, counterparties, redemption logic, protocol logic, and dependencies, to produce a risk-aware measure of value. Otherwise, you’re not pricing assets; you’re quoting volatility.
The Convergence of Price, Proof, and Risk
This is where price, proof of reserve, and risk oracles converge. They are not separate systems but interdependent layers of the same framework.
For asset-backed tokens, value cannot be defined solely by a price oracle. It emerges from the combination of market pricing and verified reserves. The interaction between what the market observes and what can be proven to exist.
Price oracles surface external market signals that reflect secondary-market behavior.
Proof-of-reserve oracles ground those signals in verifiable collateral, confirming the economic backing that gives the token meaning.
Risk oracles contextualize both, determining what the asset is worth to a specific system based on its structure, liquidity, and correlation to other exposures.
In other words, price without proof is speculation, and proof without context is static.
True valuation in DeFi depends on all three layers working together: market, backing, and risk, operating in real time.
Proof-of-Reserve oracles are progress, but we’re still far from the endgame. There’s no standard for attestation frequency or depth. Every gap between proofs is a blind spot where solvency is assumed, not verified. The ideal state is continuous, real-time attestation enabling live, observable reserves that update as conditions change. We’re moving in that direction, but we’re not there yet.
** Oracles, in a world of sophisticated DeFi, are application-specific instruments of risk, not universal truth machines. **
A lending market shouldn’t necessarily use the same oracle as a perps exchange.
Each system must decide what kind of volatility it wants to internalize and what kind it’s willing to pass downstream.
There’s no free lunch; pricing stability always transfers volatility somewhere else. The question is who absorbs it, and when.
Hard-coding 1:1 parity can be the right trade-off for lending systems that prioritize depositor safety, but it’s not universally optimal or endgame.
The underlying NAV risk doesn’t disappear; it just moves. Good risk design makes that transfer explicit, transparent, and priced.
A Call for Discipline
If you’re an asset issuer minting a tokenized RWA, stablecoin, or money market instrument, you must have a proof-of-reserve system in place and work with an oracle to provide a PoR oracle. Without it, integrators like lending markets, perp DEXs, and vault managers are blindly extending credit to your assumptions of solvency.
And if you’re an integrator:
- an exchange
- lending protocol
- vault manager
you need to work directly with asset issuers and risk partners to understand each asset’s structure, redemption logic, and reserve mechanics. Otherwise, your users will absorb risks that should have been managed at the protocol level.
The Path Forward
Yesterday could’ve been ** much worse ** for DeFi.
We’re building systems that now mirror the scale of traditional finance, but not yet its discipline. Real progress in financial infrastructure won’t come from speed . It will come from precision, transparency, and shared standards.
Every oracle, asset, and market design is a set of trade-offs.
The next phase of DeFi will belong to those who recognize that truth early and build with that rigor. Not those who move fastest, but those who move correctly.
Someone just open-sourced $10M worth of N8N automations for free.
2000+ workflows that power real businesses - everything from lead generation to AI agents.
The automations inside this vault:
→ Cold email systems generating millions in revenue
→ Lead scrapers pulling unlimited prospects
→ Content factories creating viral social posts
→ Sales funnels converting at 35%+ rates
→ Customer service bots handling thousands of tickets
→ Data extraction systems for competitive intel
→ CRM workflows that never drop a lead
→ E-commerce automations running 8-figure stores
→ AI agent systems replacing entire departments
These aren't basic templates. These are production-ready workflows that businesses pay $5K-$50K to have custom built.
Years of development. Thousands of hours of testing. Millions in proven results.
All organized in one repository.
While automation agencies charge enterprise rates for single workflows, you're getting the entire arsenal that took the community years to perfect.
No paywalls. No restrictions. Pure open-source gold.
Comment "VAULT" + RT + Like
I'll DM you the vault link
(Must be following so I can DM)
This just made every automation consultant obsolete.
There are multiple benefits but typically people fall under 2 option that can be used together.
Individuals (Chapter 2 - Investor Resident): 0% tax on capital gains for new residents.
Businesses (Chapter 3 - Export Services): 4% corporate tax rate for service businesses that export services from Puerto Rico to clients outside the island. 0% dividend.
For both cases to maximize opportunity you need to move to Puerto Rico and live there for minimum of 183 days.
1/ Everyone's talking about AI.
But few truly grasp what's coming.
This isn’t just “predicting the next word.”
It’s building machines that can plan, reason, and strategize faster than any human. 🧵
Imagine understanding exactly how to speak to Donald Trump's Followers to influence their buying process. What about speaking to Nike's followers? What about... Speaking in the voice of anyone at all?
We sit down with @BeelinesAi to understand how they leverage Ai to build out audiences and discover how to craft messages that resonate with any audience, from political followers to brand enthusiasts, in seconds.
This is a must watch for any marketer or anyone interested in truly innovative applications to use artificial intelligence vs just saying "We are Ai" with no use case.
00:00 – Exploring the Human Intelligence Model
01:14 – Innovative Applications of Beelines AI
02:31 – AI-Driven Customer Behavior Simulation and Enhancing AI Training through Human Interaction
06:23 – Leveraging Real-Time Data for AI Personalization
12:55 – Public Bees: Enhancing Discoverability in AI Networks
15:00 – Revolutionizing Advertising with AI Consumer Insights
16:09 – Utilizing AI for Market Research and Feedback Collection
19:17 – Influencer Marketing Strategies in the AI Era
21:08 – Upcoming AI Developments and Innovative Tools
23:54 – Final Thoughts and How to Connect
Guests: @gregarious and @IsaacLazoff
Host: @AviNMash
Imagine understanding exactly how to speak to Donald Trump's Followers to influence their buying process. What about speaking to Nike's followers? What about... Speaking in the voice of anyone at all?
We sit down with @BeelinesAi to understand how they leverage Ai to build out audiences and discover how to craft messages that resonate with any audience, from political followers to brand enthusiasts, in seconds.
This is a must watch for any marketer or anyone interested in truly innovative applications to use artificial intelligence vs just saying "We are Ai" with no use case.
00:00 – Exploring the Human Intelligence Model
01:14 – Innovative Applications of Beelines AI
02:31 – AI-Driven Customer Behavior Simulation and Enhancing AI Training through Human Interaction
06:23 – Leveraging Real-Time Data for AI Personalization
12:55 – Public Bees: Enhancing Discoverability in AI Networks
15:00 – Revolutionizing Advertising with AI Consumer Insights
16:09 – Utilizing AI for Market Research and Feedback Collection
19:17 – Influencer Marketing Strategies in the AI Era
21:08 – Upcoming AI Developments and Innovative Tools
23:54 – Final Thoughts and How to Connect
Guests: @gregarious and @IsaacLazoff
Host: @AviNMash
We're hosting our first Spaces event!
TLDR:
- This will be weekly
- It will be a community style discussion hosted by our leadership team
- We'll cover the intersection of AI x Marketing tools, trends, & best practices
- We'll share alpha on what we're building
RSVP 👇