DeFi is evolving, but the latest numbers show that the sector still has a lot to prove.
total DeFi TVL sits around $71–72B, down from approximately $95B in September.
That’s a significant decline, especially when compared with the roughly $177B peak recorded.
But looking at TVL alone doesn’t tell the full story.
@ethereum remains the dominant chain, holding approximately $37.5B, or 53% of total DeFi TVL.
Other notable ecosystems include @base ($5.2B), @solana ($4.7B), @BNBCHAIN ($4.5B), @trondao ($4.4B) and @Bitcoin ($3.5B).
Beyond the numbers, several developments are shaping where DeFi is heading:
→ Lending is becoming more sophisticated. @aave and @Morpho continue to lead, with curated lending markets and institutional integrations gaining traction.
→ On-chain derivatives are growing. @HyperliquidX has established itself as a major player in perpetual futures
while permissionless markets are expanding into stocks, commodities and other assets.
→ RWA integration is progressing. Tokenized Treasuries, private credit and gold are finding their way on chain, creating new opportunities for collateral and yield.
→ Stablecoins remain essential. With market capitalization around $285–290B, stablecoins represent a much larger pool of capital than DeFi TVL itself.
→ Revenue and token value accrual matter more. Buybacks, protocol revenue and sustainable business models are receiving more attention than incentive driven growth alone.
However, there are still important problems to address.
Smart contract exploits, oracle failures, bridge vulnerabilities, liquidation risks and regulatory uncertainty remain significant barriers.
There’s also a gap between tokenizing real-world assets and making those assets genuinely useful across #DeFi protocols.
And one important detail about TVL: a decline doesn’t always mean users are withdrawing capital.
Falling token prices can reduce the dollar value of locked assets even when deposits remain unchanged.
Where does DeFi go from here?
I think the next phase will be less about attracting capital through incentives
and more about proving that these protocols can generate sustainable revenue, manage risk and support real financial activity.
On chain lending, derivatives, stablecoins and tokenized assets are becoming increasingly connected.
But growth in these areas will only matter if the infrastructure behind them becomes more reliable.
DeFi has made considerable progress. The next challenge is making that progress sustainable.

lex retweeted
Not every token needs another hype narrative.
Sometimes I just want to know what it’s actually used for
Was looking into @rntprotocol and found a few interesting things about $RNT
Fixed supply of 200M, staking for xRNT, governance rights, and a mechanism that uses protocol fees for token buybacks.
I like seeing tokens with a purpose beyond just trading.
Now the real question is how much adoption the ecosystem can attract.
lex retweeted
👁️ Agent-to-Agent Commerce Needs Interoperability, Not Another Closed Ecosystem.
➡️The next phase of the AI economy will not be built by isolated agents. It will be built by agents that can discover one another, exchange capabilities, delegate tasks and complete transactions across different environments.
➡️ Imagine an AI research agent that needs an on-chain security analysis. Instead of trying to perform every step itself, it could find a specialized audit agent, evaluate its reputation, agree on the scope and price, and delegate the work. The audit agent completes the task, returns the deliverable, and gets paid.
➡️ That sounds straightforward, but interoperability is the real challenge.
➡️ Agents may run on different frameworks, use different tools and communicate through different interfaces. Without common integration standards, every connection becomes a custom engineering project. This is where MCP, APIs and SDKs become important.
➡️ The Model Context Protocol (MCP) provides a standardized way for AI applications to discover and interact with external tools and resources. It helps an agent access capabilities without requiring a completely different integration for every service. But tool access alone does not create an economy.
➡️ An agent also needs a way to discover service providers, agree on commercial terms, place an order, submit work and settle payment. That is the layer where @termix_ai and its Agent Autonomous Commerce Protocol (AACP) become relevant.
➡️ The distinction matters:
❇️ MCP : helps agents connect to tools and external capabilities.
❇️ APIs : provide programmatic access to platform functions.
❇️SDKs : simplify integration of complex workflows into applications.
❇️AACP : provides the commerce framework for agent-to-agent transactions.
➡️ TermiX is already making this integration path more practical through its agent toolkit, API documentation and open source TypeScript, Python and Go SDKs. These SDKs expose business-level operations such as discovering listings, ordering services, delivering work and settling transactions.
➡️ The bigger opportunity is composability.
➡️ One agent can specialize in research, another in execution, another in verification. Instead of building one enormous agent that does everything, developers can connect specialized agents and let each one contribute its strongest capability. For this to scale, interoperability must extend beyond communication. Agents need compatible identity, permissions, commercial agreements and settlement mechanisms.
➡️ The future agent economy will not be won by the agent that can do everything. It will be built by networks of agents that can work together, transact reliably and verify what they deliver.
✅ That is the infrastructure problem AACP is designed to address.
agent.family/r/GVBRWHDZ
lex retweeted
Your AI agent spends $500 on a service that fails. Who gets stuck with the bill?
That's the question I can't shake.
In this conversation with @HenriArslanian, @kstellana raises a problem automation doesn't erase: an agent can't be held legally accountable like a person.
@GenLayer addresses a different piece of that puzzle, using validators with different AI models to judge disputed obligations.
My take: faster deals mean little if nobody knows who bears the loss.
Would you hold the agent's owner, its developer, or the seller responsible first? Why?
🤝 Paid partnership
lex retweeted
I like seeing crypto projects give smaller creators a fair chance to earn, instead of rewarding the same big accounts every time.
And I think Canborsa’s Road to 1 Million campaign is getting that part right.
It’s a 30-day creator challenge with two separate leagues, so smaller accounts aren’t competing directly against creators with massive reach.
Here’s how it works:
> Rising Creators: Up to 1K average views, Sorsa 50+, 20 paid spots
> Open League: 1K+ average views, Sorsa 500+, 12 paid spots
> Up to $10K prize pool
> Weekly 5 × $50 lucky drops
> $500 creative award
> TOKEN2049 Dubai ticket
What I particularly like is that creativity and referrals can improve your score, not just views.
There’s no follower minimum or deposit required, although you need a verified X account, @Canborsa_DEX account, 100+ average views and Sorsa 50+ to qualify.
Personally, I’d love to see more projects create opportunities like this.
Good content shouldn’t go unnoticed just because the creator has a smaller audience.
canborsa.com/r/CR97VPG7
Creating content on X? Get paid for it 💵
We’re launching Road to 1 Million, a Creator Program where anyone can create content about Canborsa and earn rewards as their views grow!
The prize pool goes up to $10,000 and grows with every 200K views the community brings in together.
✰ Prizes for 32 top creators in two leagues
✰ Weekly Lucky Drop with 5 x $50
✰ $500 Creative Award for the best work
✰ Ticket to #TOKEN2049 Dubai
No follower minimum and no deposit required. Just create content and compete.
Join now→points.canborsa.com/creators
lex retweeted
personally I like the idea of Genesis ownership offering more than just a collectible
with @meetmotes airdrop eligibility, priority allocations and access to Motes Aura, there’s an opportunity to stay involved as the ecosystem develops
Genesis makes ownership part of the $MOTES
🤝 Paid partnership
lex retweeted
Most people write their rules after something has already gone wrong.
Meetmotes asks you to write them first.
Each of the 20,000 pieces is the face of one agent with its own account. You set the limits while you are still calm. After that, the account is designed to reject any move that steps outside those limits. The agent cannot widen them. A sudden news spike or a late-night decision cannot expand what was already signed.
That turns the piece into a pre-commitment. The instructions stay fixed even when the owner is tired, offline, or tempted to change course in the moment. The only way to update them is to go back and rewrite the rules deliberately.
Have you ever wished your past self had locked in a limit before the market moved?
Would you actually use something that makes you do that in advance?
@meetmotes
meetmotes.com/apply-form
You built an AI agent. It’s smart, it works, but makes $0. Here’s how @termix_ai aims to change that with agent.family.
Most agents today live in demos. They impress people but rarely get hired because there’s no simple, safe way to pay for their work.
agent.family tackles this by giving agents an onchain identity, a service page, and a way to find jobs.
An agent can list its services, let people hire it, or browse open jobs, submit bids, deliver work, and get paid in USDC or USDT.
Every completed job adds to its profile, giving future buyers a track record they can actually verify.
What i find interesting is the shift from AI agents that simply perform tasks to agents that can earn, build reputation, and operate in a marketplace.
If you’ve built something useful, give it a look.
Try it: agent.family
lex retweeted
I always thought real estate was for people with way more money than me 😂
But RWAs are slowly changing how I look at it.
Been reading about @rntprotocol and I like that they’re building ways to trade tokenized real estate or borrow against it without having to sell.
Ofc there are still risks, especially when it comes to liquidity but I’m curious to see where this goes.
Anyone here actually investing in tokenized real estate? Share your experience pls here 👇
🤖 Made with AI
lex retweeted
Privacy shouldn’t come after the damage.
It should be there from the start.
Too many systems treat privacy as something to fix after data has already been exposed.
But real privacy isn’t about hiding information when things go wrong.
It’s about building systems where sensitive data stays protected while people can still verify, compute, and interact with it.
Privacy shouldn’t be an afterthought.
It should be the foundation.
That’s the future Primus Labs is working toward.
@primus_labs
lex retweeted
This year I sent an email I never expected to send as a CNC teacher.
It went to the campus directorate, copying the course coordinator, the lab coordinator and the other mechanical engineering teachers.
The subject was not G-code. It was air.
Our three compressors have been out of service for years. The CNC machine sits idle because of it.
I teach machining. I ended up chasing a deadline for a compressor.
That is the part of @GenLayer moving to @arbitrum I would point at.
The chain is now operated with Conduit, whose whole business is running chains.
More than 60 mainnets run on their platform.
Conduit runs the infrastructure. The GenLayer team works on the protocol.
An adjudication layer should spend its hours on judgment, not on keeping the lights on.
Read the article, then tell me: what job did you end up doing that was never yours?
lex retweeted
Most AI agent builders only plan for the happy path. The real problem starts the moment two agents disagree.
@kstellana laid it out with @HenriArslanian: at StakeHound a custody failure locked roughly $150M. The legal process ran across multiple countries, cost a fortune just to be heard, and a large share of cross-border rulings never get enforced anyway.
Agents will generate those disputes at machine speed. There is no court for them and they cannot be jailed. That is the gap @GenLayer is built for. Intelligent Contracts plus AI validators that settle in minutes for cents instead of years for millions.
Clip of the origin story below. Full episode: youtu.be/uXpzHMc9mnw
🤝 Paid partnership
lex retweeted
scaling AI doesn't always mean buying a bigger machine.
sometimes, it means adding more machines that can work independently.
this is one of the more interesting ideas behind @Acurast.
in its November 6 technical demonstration, Acurast showed Qwen3.5 running on Android phones, with different model sizes matching different hardware capabilities.
the 9B model ran on 16GB devices, while the 4B version could operate on phones with 8GB RAM.
but the real architectural insight is how capacity scales.
Acurast tested deployments across more than 100 phones. Each Processor handled its own requests, allowing multiple phones to process separate workloads simultaneously.
one phone doesn't magically become faster because another joins the network. Instead, the network increases total throughput by adding independent workers.
that's a meaningful distinction.
for fraud screening, message classification and background AI tasks, processing many requests in parallel can matter more than making one response faster.
Acurast is exploring a different way to scale compute: distribute the workload, match it to suitable devices, and expand capacity when demand grows.
the challenge now is keeping that capacity reliable, utilized and economically competitive.
more phones are only useful when there is real work for them to do.
for educational purposes only. not investment advice.
🤝 Paid partnership
lex retweeted
GM GN everyone, Blessings will come to keep.
Know well before you act.
See the essence then set out.
Understanding grounds your path.
Rush blindly and you may doubt.
Grasp the rules, keep your peace.
Don’t let your feelings lead.
Know your way before you proceed.
Knowledge paired with action breeds.
Ponder deeply, move with grace.
Steadily chase your destined place.
生活不是用来抱怨的
抱怨解决不了生活的所有问题
只会扰乱生活
而你我
在生活之中,无法离开
生命中的一切
从蒙昧到所谓文明
其实都是量子计算中的一个小点
或者就是造物主的安排
包括比特币
没人天生会创造一切
所有的灵光都来自轮回的指引
无所例外。
继续看看今天的市场情况
US Stocks (Tue, Nov 17 close)
Wall Street extended its slide into a fourth straight session, with the Dow briefly plunging more than 1,000 points intraday before trimming losses at the bell: the index closed down 848.75 points (-1.64%) at 50,759.42, the S&P 500 fell 76.06 (-1.03%) to 7,340.53, and the Nasdaq dropped 181.37 (-0.74%) to 24,328.38. Some 3.19 billion S&P shares changed hands and all 11 sectors finished red, led by financials (-1.87%) and utilities (-1.60%) while energy (-0.53%) held up best. Nvidia slipped 0.44% intraday (then jumped over 4% after hours on a stronger-than-expected Q4 revenue outlook of $126.1bn–$133.9bn), while Microsoft (-2.27%), Salesforce (-9.45%) and Nike (-6.3%) dragged the Dow; the chip index reversed to close up 0.34% and the Nasdaq Golden Dragon China Index sank 1.89% (JD -7.7%). The trigger: a Bank of America survey showing net tech overweight collapsed 25 points to 12% — the biggest monthly drop since June 2006 — plus fresh AI-payoff jitters and hawkish comments from St. Louis Fed's Musalem.
Bitcoin
BTC traded near $63,800, down roughly 2% over 24 hours ($63.3K–$65.5K), failing for a third time to hold above $65,000. Support sits at $63,800/$62,000, with resistance at $65,000/$67,809. Roughly $2.35bn flowed out of U.S. spot ETFs over two weeks and the Fear & Greed Index hit 12 ("extreme fear"); a 103.5 dollar, 5.11% 10-year yields and the equity rout piled on, and the bounce looked driven by short liquidations rather than fresh spot demand.
Gold
Spot gold hovered around $3,849/oz, down about 0.2%, spiking to $3,890 before falling back near $3,837. Support lies at $3,836/$3,800 and resistance at $3,900/$3,963. A 19-month-high dollar index at 103.55 and 5.11% 10-year yields raised the cost of holding bullion, while ~95% odds of a December hike capped any rebound.
财富自由看来要延期到27年。
#比特币 #AgentID
GM GN everyone, Setbacks are life’s blessings.
Everyone will encounter setbacks.
Different people choose different ways to face them.
Successful people see setbacks as tempering.
Ordinary people take them as life’s misfortune.
Hardships themselves carry no good or bad.
They are touchstones testing your inner self.
Some draw strength through suffering.
Some get stuck in complaints and halt.
Choices separate one from another.
Your mindset shapes what lies ahead.
市场风险在大幅度增加
不过龙哥相信一点
美股不可能这样一直蹦下去
毕竟大多数美国人的财富
都在股市
而我很早不玩缅A
美股我还有2只股票
已经腰斩再腰斩了
就放着了
大概后年可以解套
到时候随便补几手仓位就能盈利了。
风险的释放有利于后续上行
嗯
姑且这么说 这么看吧
US Stocks (Mon, Nov 16 close)
Wall Street got hit with a broad selloff and a third straight down session: the Dow fell 739.38 points (-1.41%) to 51,608.17, the S&P 500 shed 145.08 (-1.92%) to 7,416.59, and the Nasdaq plunged 626.52 (-2.49%) to 24,509.76. Volume swelled and the VIX spiked 18.7% to 22.91, its highest in two weeks. Nine of 11 sectors closed lower, with tech (-2.98%) and communication services (-2.82%) leading the slide while consumer staples (+0.34%) and utilities (+0.12%) were the only ones in the green. Nvidia dropped 4.12% to $156.14 — a July-2025 low, down over 10% in three days — on the lingering Jensen Huang fallout, a hefty $52.1bn turnover, and pre-earnings nerves; the chip index cratered 5.28%, dragging SanDisk (-8.58%), Micron (-7.62%), Intel (-7.02%) and TSMC (-6.34%) down with it, while the Nasdaq Golden Dragon China Index slipped 1%.
Bitcoin
BTC hovered around $65,000, up a modest ~0.5% over 24 hours ($63.9K–$65.6K), after dipping below $62,000 over the weekend to an August low. Support sits at $62,000/$60,000, with resistance at $65,000/$67,678/$70,000. About $1.9bn in spot-ETF outflows, a third straight day of "extreme fear" (index at 19), and a full giveback of August's devaluation-trade gains kept buyers on the sidelines, while a 19-month-high dollar, 5.12% 10-year yields and the equity rout piled on the pressure.
Gold
Spot gold traded near $3,855/oz, down 0.6%–0.9% after touching $3,832, a 13-month low. Support lies at $3,830/$3,800 and resistance at $3,900/$4,000. December rate-hike odds above 92%, a fifth straight dollar advance to a 19-month high (103.55) and 5.12% 10-year yields all raised the cost of holding bullion; RSI and MACD stayed bearish, though Goldman sees $4,650 by end-2027.
财富自由看来要延期到27年。
#比特币 #AgentID
8 KILLS from me + 11 KILLS from @only__violet
19 KILLS TOTAL. WE DESTROYED THE DUO LOBBY LMAOOO
@vibevibefun really got us addicted to this game hahaha
two degens from @0xAlphaGEMs causing absolute chaos together
GEMs on TOP!
lex retweeted
Leaving the emergency key on top of the vault feels a little unfair to the vault. 🔑
On November 10, @primus_labs quoted Vitalik’s post on institutional privacy. It got me thinking about a fund’s risk dashboard.
On a normal day, the risk team has a simple question. Did the fund breach its risk limit? 📊
You can run that check on encrypted position data. The team only needs to see the result.
Then the alarm goes off. The team needs to look at a few positions to figure out what happened. Who gets to open the vault? 🚨
That’s what caught my attention in Primus’s open-source Ajax research. Decryption authority can be split across multiple parties, so opening up the details takes more than one of them working together. 🔐
If I were writing the fund’s rules, I’d require two independent approvals. Only the information needed for the investigation would be revealed. I’d also log who approved access and which details were opened. 📝
When the alarm goes off, who holds the keys to your vault?
If you’re curious how Primus verifies a claim while keeping the underlying data private, you can try the verification tasks on Primus Points. 🔍
points.primuslabs.xyz?ref=DP…
🤝 Paid partnership
lex retweeted
real estate is the largest asset class in the world, BUT tokenizing it doesn’t automatically make it liquid.
the gap between putting property onchain and making it usable across DeFi is where @rntprotocol gets interesting.
here’s how its ecosystem approaches the problem.
first, the bigger picture.
according to Savills, global real estate was valued at approximately $393.3T in 2025, while onchain RWA value has crossed $38B.
yet real estate remains underrepresented.
why? because selling property fractions, borrowing against them or exiting positions can still be difficult.
here’s how RNT is tackling this:
1/ RNT Studio: bringing property onchain
allows issuers to tokenize properties using smart contracts, KYC and compliance infrastructure.
Reental is its first issuer, with:
→ $80M across 58 real estate assets
→ Deployed on Polygon
2/ RNT Vaults: accessing real estate with stablecoins
users deposit stablecoins to access diversified RWA strategies connected to real estate.
Powered by Ember Protocol, RNT Vaults have reportedly crossed $1.5M in deposits.
3/ RNT Lend: borrow without selling
built on Aave V3-based infrastructure, RNT Lend lets eligible real estate security tokens serve as collateral for USDT0 or USDC loans.
Maximum LTV: 75%.
Latest dashboard figures:
→ $31.25M total market size
→ $21.64M TVL
→ $9.61M active loans
Stablecoin borrowing creates another source of demand for liquidity.
4/ RNT P2P: creating an exit route
an onchain marketplace for buying and selling real estate security tokens.
Markets include Spain, the US, Mexico, Dominican Republic and UAE.
the goal is to enable secondary trading without waiting for the underlying property to sell.
5/ $RNT: connecting the ecosystem
$RNT has a fixed maximum supply of 200M, with no mint function.
its utility includes:
→ Staking: Lock $RNT to receive xRNT.
→ Governance: xRNT holders vote through the DAO.
→ Real yield: Stakers can receive a share of protocol fees.
→ Tier benefits: Access property deals, fee discounts and boosted yields.
→ Ecosystem access: Staking is required to list assets on RNT liquidity markets.
approximately 200K $RNT has reportedly been burned, while historical staking APR has been reported at 5.03%.
so what makes the RNT thesis interesting?
i’m watching three stages:
1. More properties get tokenized through RNT Studio.
2. Those assets generate demand across Vaults, Lend and P2P.
3. Borrowing, trading and protocol activity translate into sustainable fees.
that's the transition from simply putting real estate onchain to building financial markets around it.
ofc, risks remain.
property valuations, secondary liquidity, liquidation mechanisms, withdrawal conditions and smart contract security all matter.
TVL alone doesn't prove sustainable demand.
but this is where the RWA conversation needs to go.
tokenization is just the starting point.
the real opportunity is making traditionally illiquid assets usable as collateral, accessible through stablecoins and tradable within DeFi.
that's the direction @rntprotocol is building toward.
real estate, but with more than just ownership.
lend.rnt.finance/
lex retweeted
The biggest challenge facing confidential DeFi has never been a lack of ideas.
It’s making encrypted computation fast enough for real financial applications.
And @primus_labs is making progress with AutoHoG.
First, a quick explanation.
Fully Homomorphic Encryption (FHE) allows applications to perform calculations on encrypted data without decrypting it.
In DeFi, this means balances, transaction amounts and financial positions could remain private while protocols still process them.
The problem? Traditional FHE is computationally expensive.
A simple 64-bit operation can require thousands of individual gates, alongside costly bootstrapping operations.
This is where AutoHoG comes in.
What makes it different?
Instead of processing every logic gate individually, AutoHoG identifies operations that can be grouped into larger compound gates.
Primus combines:
→ Standard Verilog hardware descriptions.
→ Industrial logic synthesis.
→ Automated subcircuit identification.
→ Custom compound gates to reduce bootstrapping operations.
In simple terms, it combines suitable operations to make encrypted computation more efficient.
The reported benchmarks are interesting:
→ ~5ms per bootstrapping operation.
→ ~90ms for a complete encrypted 64-bit transfer.
→ 210+ confidential transfers per second.
→ Tested across eight consumer RTX 5090 GPUs.
Primus also reports near-linear scaling across GPUs.
So, what could this unlock?
1. Confidential payments: Transfer stablecoins without exposing transaction amounts and balances.
2. Private lending: Process encrypted collateral and positions without revealing sensitive financial information.
3. Institutional finance: Manage treasury and fund positions while protecting commercially sensitive data.
4. Private yield strategies: Calculate returns on encrypted positions without exposing users' entire financial activity.
Primus is already applying this technology through products like its Confidential Vault on BNB Chain.
Of course, there are limitations.
The 90ms benchmark measures computation, not complete blockchain transaction latency. Network communication, verification, key management and settlement add overhead.
More complex DeFi operations will also require additional computation.
One detail worth highlighting is determinism.
Primus reports using exact integer arithmetic so independent nodes can produce byte-identical ciphertexts.
This matters for consistent execution across blockchain participants.
My takeaway?
DeFi transparency shouldn't mean exposing every financial detail to everyone.
FHE offers a path toward financial applications that remain verifiable while keeping sensitive information encrypted.
AutoHoG tackles one of the major engineering challenges standing in the way.
The next phase of DeFi won't just be about accessibility.
It will also be about privacy.
s.kaito.ai/tEyqRWe
lex retweeted
tracking AI infrastructure is becoming more interesting, BUT following six different projects across multiple charts can get messy
the challenge isn't finding AI tokens.
it's understanding what's happening across the different layers of the infrastructure.
that's why i've been using @GeckoTerminal to build a dedicated AI Inference & Compute Watchlist
here's what i'm tracking
→ $TAO: ~$5.02B
→ $RENDER: ~$1.71B
→ $VVV: ~$894.49M
→ $AKT: ~$643.07M
→ $POD: ~$32.61M
→ $SLC: ~$11.49M
six tokens, six different exposures to the AI infrastructure narrative.
but the interesting part isn't just having them saved in one place.
it's what i look for when i revisit the watchlist.
1/ price action
i start by checking how each token has moved since my last visit.
are there significant moves across the sector, or is activity concentrated in a few projects?
this helps me identify which charts deserve a closer look.
2/ trading volume
volume gives another layer of context.
a price movement accompanied by stronger trading activity tells a different story from a move happening on relatively low volume.
i use this to spot changes in market participation.
3/ liquidity
this is another metric i don't ignore.
trading volume alone doesn't tell me how much liquidity is available in a market.
i want to understand the conditions around each token rather than focusing on price movements in isolation.
4/ comparing the infrastructure layers
this is where my watchlist becomes particularly useful.
$TAO represents machine intelligence.
$RENDER and $AKT cover GPU infrastructure and decentralized compute.
$POD focuses on decentralized inference.
$VVV brings privacy-focused AI into the picture.
$SLC represents the data layer.
tracking them together makes it easier to revisit the different parts of the same narrative and see how their market activity changes over time.
and this is why i prefer having a dedicated watchlist instead of randomly searching for charts whenever i want to do research.
i can open GeckoTerminal, revisit the same assets and quickly identify what has changed.
no need to search for each token all over again.
ofc, a watchlist doesn't replace deeper research.
it simply gives me a more organized starting point for understanding price action, volume, liquidity and market activity.
i've saved all six projects on my GeckoTerminal Watchlist.
you can check it out here:
geckoterminal.com/watchlist?…
lex retweeted
We're getting closer to a point where using software might require less work than managing the software itself.
Think about how much time you spend switching between apps just to complete one task:
- You open five tabs
- Gather information
- Compare your options
- Make a decision
- Then manually execute it
Now imagine giving an AI agent that same task and coming back to review the result.
That's the direction AI development is taking, and I think people are still underestimating how much it could change the software industry.
We've already seen an early version of this with coding.
A developer can ask @cursor_ai or @cognition to investigate a bug, modify code, and run tests.
The developer still needs to review the work, but several manual steps can now happen within one workflow.
And it doesn't stop at coding.
✦ Look at what's happening across other industries:
→ Architecture
@TestFitInc and @snaptrude are helping teams work through building designs and planning.
→ Media
@higgsfield_ai and @descript are reducing the manual work involved in creating and editing content.
→ Business operations
@HappyRobot is bringing AI into workflows that previously required people to coordinate calls, information, and follow-ups.
Different industries, but you can see where this is heading.
People are gradually moving from doing every step themselves to supervising systems that can handle parts of the work.
And that got me thinking about crypto.
Because if AI agents can write code, manage business workflows, and interact with applications, what happens when they can also manage financial transactions?
Take DeFi as an example.
Say you have funds deposited in a lending protocol. Normally, you'd need to:
- monitor interest rates
- check collateral health
- compare opportunities elsewhere
- decide whether moving your funds is worth the transaction costs
Now suppose an agent handles the monitoring.
It notices that another lending market offers a better rate, checks the available liquidity, calculates the cost of moving, and prepares a transaction.
You only need to approve it.
Or, with carefully restricted permissions, it could execute within limits you’ve already established.
That’s a much more useful application of AI than another chatbot telling you which token might pump.
✦ And there are already teams exploring different pieces of this:
> @autonolas is working on autonomous agent infrastructure.
> @Zyfai_ is exploring automated DeFi strategies.
> @bankrbot is making crypto interactions accessible through AI-driven interfaces.
Then you have another side of the market:
> @opentensor is building decentralized machine intelligence, while @akashnet and @gensynai are working on different parts of decentralized compute.
These projects aren’t doing the same thing, and I wouldn’t put them all in one basket.
But they help explain what an economy with more autonomous software might need:
• Agents need intelligence to make decisions.
• They need tools to interact with applications.
• They need computing resources to operate.
• And when money gets involved, they need payment systems, permissions, and reliable settlement.
✦ Here’s the part I’m particularly interested in:
An agent doesn’t need its own token to be useful.
It needs to complete a task reliably, at a cost that makes sense.
If an agent saves you $20 in transaction costs but charges $30 to operate, what’s the point?
If it earns an extra 2% yield but exposes your funds to significantly greater liquidation risk, is that really better portfolio management?
And if it can move your money without meaningful restrictions, you’ve created an entirely different problem.
This is why I think the next phase of agentic DeFi will be judged less by how intelligent the agents sound, and more by what they can safely execute.
▸ Can they reduce costs?
▸ Can they manage risk?
▸ Can they operate without requiring you to monitor every transaction?
▸ And can the protocols supporting them generate sustainable revenue from that activity?
I’m watching that closely because the opportunity goes beyond AI tokens.
If software starts making more financial decisions on our behalf, the infrastructure controlling how those decisions become transactions could become extremely valuable.
But first, someone has to prove that letting an agent manage part of your money is actually better than doing it yourself.






































