←
Dave Burrells8,949 posts
Dave Burrells avatar

Dave Burrells

@dburrells

Chief Commercial Officer @Alludium | ex-JPMorgan | Sharing insights on the Agentic Economy - where AI agents and blockchain rails meet | Views my own

◷ Joined December 2008▧ 697 Photos and videos
Dave Burrells avatar
Dave Burrells@dburrells·Nov 12

2.18 million AI agents are already registered on-chain. Licensing each one would kill the machine economy before it starts. Agents are unique and can be spun up in seconds. Requiring a licence for every single one is like demanding a separate ISO or SOC certificate for every software product a company releases. Regulatory approval cannot keep up 🥵 The workable design is different. Just as ISO and SOC certify the business, not each product, we license the creators, flag every agent, and let markets price the risk in real time. The licence can be revoked if standards are not maintained. Accountability remains. Experimentation continues. Check out my latest blog for the full framework 👇 medium.com/@dburrells/you-ca… #AIAgents #AgenticEconomy

Dave Burrells avatar
Dave Burrells@dburrells·14h

Gold fell nearly 2% overnight to a 1-year low. Bitcoin is stuck below $65,000. The dollar is near a 19-month high. This is not a normal market day. It is a live stress test for the next capital managers: AI agents. Storing wealth and managing liquidity are two different jobs. An agent that must pay for compute, data or services needs capital it can spend in seconds. It cannot liquidate longer-term holdings every time an invoice arrives. Parking everything in stablecoins solves the speed problem but creates issuer, counterparty and currency exposure. The logical architecture therefore separates three layers, each with different rules: 💰 Operating float: immediately accessible capital under strict, real-time limits 📈 Productive capital: yield-generating assets with controlled access and clear exit paths 🏦 Longer-term reserves: higher-duration holdings governed by stronger authority and slower decision rights Today’s moves show why the separation is non-negotiable. Even traditional stores of value can fall sharply when monetary conditions change. An agent that confuses operating cash with reserve wealth will eventually fail a liquidity test it did not see coming. Institutions are already building the rails. Yesterday BNY and Microsoft completed an industry-first pilot in which AI agents negotiated a US Treasury bill trade under deterministic rules, with humans retaining final authority. The opportunity is not teaching agents how to hold assets. It is building the financial infrastructure that lets them manage capital intelligently, without confusing money needed tomorrow with wealth intended for years. #AIAgents #AgenticEconomy

Dave Burrells avatar
Dave Burrells@dburrells·Nov 17

Institutional dollars just went 24/7 between Singapore and the US 🏦 DBS and JPMorgan connected their tokenised deposit networks. Ant International is the first client through the door. Traditional corporate treasury still stops for banking hours, cut-offs and fragmented accounts. Businesses run globally around the clock. Their money does not. Connecting bank deposit networks starts to remove that friction. Liquidity can move when the obligation appears, not when the banking system reopens. An AI agent can already monitor positions, forecast liquidity needs and decide when funds should move. That intelligence has limited value if the rails underneath cannot execute. This is not autonomous treasury yet. Access remains restricted to participating institutions. One important constraint has been removed. The agentic economy needs more than intelligent agents. It needs financial infrastructure that can keep up with them. Are banks quietly building the operating system agents will actually run on? 🤔 #TokenizedDeposits #AIAgents #RWA

Dave Burrells avatar
Dave Burrells@dburrells·Nov 16

A 60% drop in agent task cost changes the business case overnight 📉 Take a $3 job. At $4.40 in model costs it loses money before tools, verification or margin. At $1.70 it clears. Same buyer. Same price. A completely different unit economics. The agentic economy scales the moment the cost of doing the work falls below what someone will pay for it. Headline efficiency numbers are only the starting point. The real measure is total cost per accepted result, including retries, tool calls and checking the output 💡 Lower reliable delivery costs make smaller jobs worth selling, let agents buy specialist work from other agents, and open services that were previously too expensive to provide. Every durable reduction in that cost expands the set of work that can clear. The threshold is simple: a buyer willing to pay more than it costs an agent to deliver. #AI #AIAgents

Dave Burrells avatar
Dave Burrells@dburrells·Nov 15

Your AI agent just paid a sanctioned entity. “I didn’t know that restriction applied” is already a weak defence for a human. For software that can encode every sanctions list, spending limit and licensing rule, it is indefensible. The law belongs in the guardrails ⚖️ Restrictions must not live only in PDFs that an agent is hoped to remember. They must exist as enforceable operating boundaries. This creates Compliance-as-code: jurisdiction-aware rule libraries that businesses update the way they update software dependencies. Publish a rule. Deploy the guardrail. Compliance propagates across thousands of agents at machine speed ✅ The first legal system with a deployment pipeline. #AIAgents #AgenticAI #ComplianceAsCode

Dave Burrells avatar
Dave Burrells@dburrells·Nov 15

Bitcoin traded toward $62,000 this weekend 📉 $1.91 billion left US spot Bitcoin ETFs across seven straight sessions. Roughly $2 billion in leveraged crypto positions were liquidated in 24 hours 🏦 That is two independent sources of selling hitting the same market at once. ETF redemptions force funds to sell underlying Bitcoin. Leveraged longs are closed by the exchange, not by the trader. Once the first wave of liquidations hits, price drops further and triggers the next. The mechanism is self-reinforcing 🔄 Higher-rate expectations and the sell-off in AI equities supplied the initial pressure. The weekend simply made it worse. Spot ETFs are closed. Crypto markets stay open. Thin liquidity means the same dollar of forced selling moves the price further. This is where the long-term treasury argument becomes operational. An asset held for multi-year conviction can still become a forced sale if the cash required to meet obligations sits inside the same balance sheet. Being directionally right does not protect you from being liquidity-wrong. You can hold the correct long-term view on Bitcoin and still run out of cash before the thesis is proven. #Bitcoin #BTC #Crypto

Dave Burrells avatar
Dave Burrells@dburrells·Nov 14

ChatGPT can now find a hotel, book it and charge your card. You still approve every spend. That is the last human gate before agents start spending inside limits you set once. @OpenAI’s Wallet keeps your rewards and dispute rights intact. The agent never sees the real card number. A payment token does the work. The approval step is temporary. OpenAI has already said it wants users to pre-authorise spending within limits they set in advance. The payment rail is the easy part. The hard part is deciding how much authority an agent can hold before something goes wrong and you still have a clear path back to your money 🤔 If the agent books the wrong hotel, “it followed the instructions” will not settle the dispute. The authority boundary has to be explicit and machine-checkable. This is the shift from agents that recommend to agents that commit. Authority is moving from design documents into production systems. How much of your monthly spend would you let an agent decide without asking? #AIAgents #ChatGPT

Dave Burrells avatar
Dave Burrells@dburrells·Nov 14

If your AI agent signs a contract tomorrow and it fails, who pays? 🤔 Companies are legal persons because capital and liability sit behind the name. AI agents are not. Grant them personhood today and the accountability chain snaps. The human or company behind the agent gains a liability shield. The AI becomes the defendant. The responsible balance sheet disappears. That is not progress. It is a structural hole in the system. 🤌 The only test that matters is simple: AI personhood becomes interesting when it adds accountability rather than removes it. Until an agent can own assets that actually absorb loss, sign contracts that bind a real balance sheet, and face claims without evaporating the principal, the answer is not yet. We are building agents that act. We have not yet built the architecture that makes those actions accountable. What has to exist before legal personhood strengthens the chain instead of breaking it? #AIAgents #AI

Dave Burrells avatar
Dave Burrells@dburrells·Nov 13

One commercial transaction may soon run through a ten-layer chain of AI agents spawning agents. If accountability stops at the first layer, the entire system fails the moment something goes wrong. There is no one left to hold responsible, and the damage becomes untraceable 🔍 The chain cannot break. Agent ↓ Operator ↓ Principal ↓ Legal person Every time an agent commissions or spawns another, the chain simply lengthens. Pseudonymous operation in the market is useful. Anonymous status in law is not. However deep the software stack becomes, responsibility must still reach a balance sheet that can be sued, fined or wound up. Unbroken. Every time. @brian_armstrong just noted agents are forecast to process $3-5 trillion in transactions by 2030. What happens to that volume if the liability chain snaps at layer four? 🤔 #AgenticAI #AIAgents

Dave Burrells avatar
Dave Burrells@dburrells·Nov 13

Bitcoin has broken below $69,000. Nvidia is down sharply after Jensen Huang’s resignation. Over $1 billion in crypto positions have been liquidated in 24 hours 📉 The price action is the signal. The system architecture underneath is the story. Huang’s exit has compressed confidence across the AI trade. Bitcoin was already under pressure from ETF outflows and tighter liquidity conditions. The two assets moved together because the same institutional capital, risk appetite, and leverage now dominate both markets. Once key levels broke, leverage amplified the move in the way it always does 🔄 This creates a structural tension. Bitcoin’s long-term thesis rests on independence from companies, governments, and monetary policy. Its near-term price, however, is increasingly a function of the same liquidity and investor base that drives AI equities. As AI and crypto institutionalise, financial-market linkages can matter as much as technological differences 🏦 Decentralised infrastructure does not automatically produce independent price behaviour. How should we model cross-asset risk when the institutional plumbing connects assets that were once treated as separate systems? 🤔 #Bitcoin #AI #Nvidia

Dave Burrells avatar
Dave Burrells@dburrells·Nov 12

The NYSE’s parent just put traditional securities on Uniswap v4 rails. OKX and ICE (owner of the NYSE) are launching OKXICE on 23 November. A tokenised securities venue under the SEC’s Innovation Exemption 🏦 70+ assets, including 59 Direxion ETFs. Trading 24/7 through permissioned Uniswap v4 pools on XLayer. This is not mainly about 24/7 stock trading. It is the moment traditional financial assets become programmable. When shares and ETFs live as tokens with real shareholder rights, they can be held, collateralised, rebalanced and settled by code 🤌 AI agents can then manage portfolios, allocate capital, post collateral and settle within defined mandates, without a human in the loop for every step. We spent years making money programmable. We are now making the assets themselves programmable. That is when the agentic economy plugs directly into real capital markets. What happens to portfolio management once agents can hold and rebalance real securities the same way they already move stablecoins? 🧐 #AIAgents #RWA #Tokenization

Dave Burrells avatar
Dave Burrells@dburrells·Nov 12

Tokenising an asset does not make its value available to spend. @StanChart and Ant International have just piloted a facility that lets a corporate treasury access cash before its tokenised money market fund redemption settles 💡 This is the distinction that matters. Blockchain makes ownership transferable around the clock. The underlying assets, settlement processes and banking arrangements still operate on traditional cycles. The pilot bridges that gap with financing. Now scale that problem to autonomous treasuries. An AI agent managing capital cannot simply select investments and execute payments. It must know, in real time, how much liquidity is actually available, when it can access it, and what that access will cost 🤔 That requirement opens the door to automated liquidity management, collateralised financing, and working capital that responds continuously to economic activity. The next stage of tokenisation is not simply putting more assets on-chain. It is building the financial infrastructure that makes those assets useful in an economy that never stops operating. What liquidity constraints will your agent hit first? #RWA #AgenticEconomy

Dave Burrells avatar
Dave Burrells@dburrells·Nov 11

An AI agent that can change its own risk limits isn’t really managing capital. It’s just hoping nothing goes wrong 🙏 The useful part of what Coinbase shipped today isn’t the specific caps or auto-stops. It’s that the agent can’t touch them. The rules live outside the agent. Once agents start handling real money, stop-losses and position limits stop being enough if the agent can rewrite them. You need hard boundaries on what it can trade, how much, how fast, and when it has to stop, and those boundaries have to be enforced somewhere the agent doesn’t control. Agent makes the calls. Platform holds the mandate. That’s the difference between an agent that can trade and one you can actually trust with capital. How many agents out there still have the keys to their own risk settings? 🤔 #AIAgents #AgenticAI

Your agent, your rules. Today we’re shipping new controls for Coinbase for Agents that your agent can’t override. Set a max trade size, a 24-hour volume limit, an auto-stop threshold for your portfolio, and more.

Dave Burrells avatar
Dave Burrells@dburrells·Nov 11

7.12 million BTC or 35.4% of Bitcoin’s supply sits behind publicly visible keys that a sufficiently powerful future quantum computer could target. 🧐 Think of the public key as a lock. Today, inspecting it does not let you make the key. A powerful enough quantum computer could change that 🔓 This is potential exposure, not evidence those coins can be stolen today. Address reuse accounts for exposure affecting 25.8% of total supply. Much of the exposure is shaped by everyday custody decisions. Moving coins to fresh addresses can reduce reuse-related exposure. Longer term, Bitcoin would need protocol support for quantum-resistant signatures and a plan for migrating existing holdings. For funds and treasuries, that makes custody quality part of the reserve calculation. The same asset can carry different exposure depending on how it is held. Better wallet practices can reduce exposure now. A protocol upgrade and migration would address the deeper problem. Long-term holders need both on their roadmap. Is the exposure in your holdings already measured, or still assumed away? 🤔 #Bitcoin #BTC

Dave Burrells avatar
Dave Burrells@dburrells·Nov 10

Tokenised stocks just crossed $3.2 billion. Agents are already settling hundreds of millions of payments. Yet most of those tokens still cannot fund a Sunday compute bill. That is the binding constraint on the agentic treasury. Stablecoins cover the operating account: compute, data, APIs, services. Tokenised equities can sit in productive capital, next to a separate liquid float and longer-term reserves. Two different jobs. @sytaylor’s four-model distinction decides whether the second job is real. A treasury must know exactly what it owns, which rights travel with the token, who holds the underlying asset, and how (or whether) it can turn the holding back into spendable cash on demand. A token that settles on a Sunday is only useful if the cash leg settles with it. Otherwise the agent still needs excess liquid float just to bridge the gap. Ownership, liquidity and settlement have to work under one owner-controlled mandate. That is the programmable balance sheet. Without it, agents keep renting capacity instead of compounding capital. #AIAgents #RWA

Every stablecoin-linked card program could soon offer stocks. But right now the headlines are about companies opting out of tokenization. That's why I think: Stock tokens are where stablecoins were in 2023.

Dave Burrells avatar
Dave Burrells@dburrells·Nov 10

Your agent doesn’t care about your loyalty status. It cares about the total outcome. 59% of shoppers would abandon their preferred brand for an unfamiliar competitor if an AI agent recommended it (Adobe). In hospitality, researchers now call agents the “gatekeepers of loyalty.” The membership that once settled the decision is becoming one input among many. Hotel status that delivers breakfast, an upgrade and late checkout is real value. An agent still weighs it against location, cancellation terms, total cost and what you actually need from the trip. The cheapest room is not automatically the best deal. The loyalty room is not automatically the best deal either. Think of the agent as a personal buyer with your preferences and status loaded, optimising the whole basket rather than protecting a relationship. Brands that keep the perks meaningful and machine-readable will still win repeat bookings. Brands that treat status as a lock-in will watch the agent book elsewhere. Loyalty is not disappearing. It is becoming conditional, earned on every transaction, not assumed from the last one. What happens to brands whose value lives only in the human relationship, not in the data the agent can read? #AIAgents #AgenticCommerce

AI agents could shift travel bookings toward brands and away from third-party sites. Stephen Grambling explains the potential implications: mgstn.ly/4kKqc1y

0:46