Kaff

► Top 5 categories by revenue in the last 30days
🥇 Stablecoin: $701.25M
> dominated by: Tether, Circle, @Paxos
🥈 Derivatives (Perps): $67.98M
> dominated by: $hype, $Edge, $Lit
🥉 Launchpad: $34.75M
> dominated by: Pumpfun, @risedotrich, Four_meme
4️⃣ Dexs: $30.44M
> dominated by: Pumpswap, Aerodrome, Uniswap
5️⃣ Prediction Market: $23.75M
> dominated by: Polumarket, Kalshi, @trylimitless | $LMTS

In case you missed why $HYPE adding ~2B value in a single day.
> #Hyperliquid partnered with #Coinbase, USDC will become the main stablecoin replacing USDH, sharing 90% revenue for HL
> Coinbase + Circle stake HYPE + share yield
HL has ~ $5B USDC in its eco, 90% yield revenue = $160M+ for HL revenue = ~ $150M buy back $HYPE.
> Coinbase already bought $20m of $HYPE over the last week, and bought $5m more today and staked all of it
> HL revenue increases 22-26%/year → $HYPE increases 20% in price today.

Elsa is one of the few AI agent projects with real, verifiable metrics that most ppl are missing
@HeyElsaAI has been shipping a live DeFi product for months, one that’s learned what agents actually need through 18.7 million real user interactions.
- $540M+ real onchain volume
- 250K daily active agents
All extracted from a production system already running at scale.
x402 + Aegis forms the trust layer that enables agent-to-API payments at scale.
$ELSA staking delivers real utility:
– Priority access to new agent launches
– Governance over Aegis parameters
– Fee share from AetherForge execution

► Where Account Abstraction Actually Failed
Last cycle, I used to expect account abstraction was the tech that could onboard the next billion users.
Yet, I still just open rawdog EOAs everyday and click confirm like nothing changed.
Looking at it cleanly, AA didn’t actually fail technically. By May 2026:
– 62M smart accounts deployed
– 2.4B cumulative UserOps
– $180M gas sponsored by paymasters
– 14M+ EIP-7702 delegations since Pectra
Problem is users never cared. AA became middleware for developers instead of a real consumer narrative. Most users didn’t even know they were using it.
Farcaster

Talking about Privacy-Preserving in AI, I think the current two biggest tech are #TEE and #ZK, but their approaches are quite different.
1/ #TEE is like a “hardware secure room” (special CPU/GPU): data runs inside, and no one outside can see it. It’s fast, practical, and suitable for all situations.
– Privacy projects on TEE: @iEx_ec, @SecretNetwork, @OasisProtocol, @PhalaNetwork.
2/ #ZK (or ZKML) uses cryptographic technology that allows proving AI results are correct without revealing the data or the model.
– Privacy projects on ZK: @AleoHQ, @Zcash, @RAILGUN_Project, @CantonNetwork
So, which

#RWA is a strong narrative with impressive growth from $5.4B in 2025 to ~$20B in May 2026.
However, this is not a story of the entire industry exploding b/c more than 95% of this added value came from only two categories:
– Tokenized Treasuries
– Commodities
The remaining categories such as private credit, real estate, and equities have either grown very slowly or are still at the “pilot” stage.
► Here are the stage of 5 categories in #RWA
1/ Tokenized Treasuries: Bonds & Money Market Funds
Market cap: ~$12.97B USD (67.2% of the total market).
Growth: +$9.00B USD (+225.5%) in 15 months.
Top pl

For a while, I just thought @Uniswap V4 was another good tech upgrade but nothing around it would really change that much.
I was literally wrong when I realized the upgrade turned the biggest DEX into programmable liquidity infra.
V3 optimized where liquidity sits. V4 optimizes how liquidity behaves.
All pools now live inside one contract instead of deploying separate pool contracts every time.
Pool deployment cost dropped 99%, V4 tracks deltas through the transaction and settles only at the end.
Multi-hop swaps, LP changes, complex routes all get cleaner because the system doesn’t need to kee

#Hyperliquid hip-4 hits ATH volume on Sunday, ~6M from just one outcome contract
Hip-3 also usually sees higher trading volume on weekends.
It is clear that HL is becoming the go-to trading venue for tradfi on weekends and holidays, when tradfi markets are closed.
Bull on $hype~

Wallet address alone is not enough once AI agents start moving real money.
I’ve been reading back into @HeyElsaAI’s new Aegis layer.
x402 already gives agents a way to pay. but payment is only half the problem.
If I’m opening an endpoint to autonomous software, I don’t just want to know it can pay me 2 cents.
I want to know if this agent is legit, so reputation starts looking like the new credit score for software entities.
250K+ onchain agents are already active daily, and x402 has processed 165M+ txn. That’s enough activity to start building portable track records from real agent behavior.
T

► RWA × AI Agents: When Portfolios Start Managing Themselves
AI agents need blockchain rails to actually do stuff. And honestly it kinda gives #DeFi capital a 🧠
Tokenized #RWAs already crossed ~$30.9B excluding stables, with ~759k holders across 170+ platforms.
What got me thinking lately is whether agents can autonomously manage tokenized portfolios.
Sounds weird but it’s already happening despite how small it still is. By most generous counts, less than 2% of tokenized RWA TVL is directly managed by autonomous agents.
RWAs are almost perfect for agents b/c they’re yield streams, collateral,

Crypto already financialized humans pretty well.
Identity, reputation, income, credit, capital markets. The entire stack exists for humans coordinating capital onchain.
Now software is starting to need the same thing. once agents become real economic actors, they also need working capital.
What I find interesting about @HeyElsaAI is the direction they’re pushing with the infra stack around this thesis.
Aegis gives agents identity and reputation.
AetherForge gives them execution and deployment rails.
I especially like Elsa Forge because it’s the capital formation layer sitting on top of the sta

Starting to think @NEARProtocol might be the first chain actually designing for agents instead of just throwing AI words around for marketing.
$NEAR Intents already processed $19.7B+ crosschain volume across 32 chains.
– 24.3M swaps
– 580k monthly active addresses
– $77M average daily routed volume
– $37.8M cumulative fees
Users don’t even need to map out the execution anymore. Just say the outcome , then solvers fight to fill it.
This is quietly becoming agent-native behavior.
Agents probably gonna love their Chain Signatures design too.
It lets a #NEAR account control native accounts on B

► DeFAI: How AI Agents Are Replacing Traders.
I believe that the emerge of AI agents will probably become the main factor for the comeback of DeFi.
1 human can trade on dex with only 10-20 txn max but with 1 agent can literally do 1000-10,000 txn per day
That is what "AI agents are becoming the primary DEX users" actually means in practice.
– currently Agents account for >19% of on-chain activities
– made 60-80% of the global crypto trading volume.
– by 2030: >80% of DeFi TVL and >80% of DeFi transactions will be executed by AI agents; “manual swap” on DEX will become obsolete.
#DeFai protocol