Chicago Federal Reserve President Austan Goolsbee has flagged high inflation as a major challenge for the US. Inflation remains well above the Federal Reserve’s 2% target.

The Fed held interest rates steady in July. However, three officials dissented and backed a 25-basis point rate hike.

Inflation Is The Biggest Problem

Goolsbee stated during an interview with Wired that rising prices are the biggest problem confronting the US, calling them more damaging than current labor-market conditions.

“The biggest problem facing our economy right now is not the collapse of industry and the collapse of jobs; it’s that the prices have been rising too fast. We have an inflation problem, and people hate inflation.”

Goolsbee also discussed employment and called the labor market “stable without being good,” highlighting the unemployment rate, hiring, and layoffs as key factors behind his reasoning. The Chicago Fed Chair suggested that market conditions have weakened but do not require the Federal Reserve’s immediate attention.

Inflation has remained higher than the Fed’s 2% target despite lower month-on-month price increases. June Consumer Price Index (CPI) fell 0.4%, while annual inflation dropped from 4.2% to 3.5%. Core CPI, which omits food and energy, remained unchanged in June but increased 2.6% from the previous year.

However, Goolsbee has not indicated whether he would support a rate hike at September’s Federal Open Market Committee (FOMC) meeting. While the Chicago Fed President is not voting on monetary policy, his observations could fuel an ongoing debate around rate hikes among regional Fed Chairs.

Policymakers Divided Over Rate Hike

Policymakers are deeply divided over interest rate hikes. The Federal Reserve left interest rates unchanged following the July FOMC meeting, with Beth Hammack, Neel Kashkari, and Lorie Logan dissenting in favor of a 25-basis point increase. Kashkari, the Minneapolis Fed President, urged the Fed to raise interest rates as high inflation, combined with the ongoing US-Iran conflict, has complicated the policy outlook. He added that the uncertainty prevents the Federal Reserve from issuing firm guidance or promising rate cuts.

Kashkari also warned that high oil prices could impact American households and businesses, adding that there was no certainty about when shipping routes through the Strait of Hormuz would return to normalcy.

St. Louis Federal Reserve President Alberto Musalem also supported a rate hike, arguing for pre-emptive measures before inflation pushes even higher. However, he did not cast a dissenting vote. San Francisco Fed President Mary Daly supported the Fed’s decision to leave interest rates unchanged, stating that more evidence was needed to determine if the increase was temporary or permanent.

Markets Look For Clues

The big question in the market is “will the Fed raise interest rates or leave them unchanged?” According to CME FedWatch, the odds are almost equal. Stubborn inflation and volatile job data have raised market uncertainty ahead of this week’s inflation report. Wednesday’s Consumer Price Report will give market watchers guidance on whether inflation is cooling. Traders expect higher interest rates by the end of the year thanks to US-Iran tensions. However, they are unsure when the Fed may raise rates.

Disclaimer: This article is provided for informational purposes only. It is not offered or intended to be used as legal, tax, investment, financial, or other advice.

Artificial intelligence is moving beyond chatbots and copilots. The next generation of AI systems is increasingly capable of acting on behalf of users—searching for information, purchasing services, managing workflows, executing trades, interacting with applications, and coordinating with other software agents.

But there is one major capability AI agents still need to operate effectively in an increasingly autonomous digital economy: money they can use programmatically.

This is where stablecoins could become especially important.

Unlike traditional bank-based payments, stablecoins can move value directly across blockchain networks, operate 24/7, and be integrated into smart contracts and software applications. For AI agents that need to make frequent, automated, and machine-to-machine payments, these characteristics could make stablecoins a natural financial layer.

AI Agents Are Becoming Economic Actors

An AI agent is more than a system that generates an answer. An agent can be designed to perceive information, make decisions, use tools, and execute actions with limited human intervention.

Imagine an AI agent managing an online business.

It could:

• Purchase computing resources when demand increases.
• Pay another AI agent for specialized data.
• Subscribe to an API.
• Purchase advertising services.
• Pay for storage.
• Execute transactions according to predefined rules.
• Receive payments for completing tasks.
• Exchange one digital asset for another.

Each of these activities requires some form of payment.

If AI agents are expected to operate continuously and independently, relying exclusively on traditional payment systems could introduce significant friction.

Bank accounts often require identity verification, geographic availability, banking relationships, business accounts, payment processors, and human-controlled authentication. Those requirements make sense for people and companies, but they can become cumbersome when the payer itself is autonomous software.

Stablecoins offer a different model.

Stablecoins Give AI Agents Programmable Money

The defining feature of a stablecoin is relatively simple: it is a blockchain-based token designed to maintain a stable value, typically relative to a fiat currency such as the U.S. dollar.

For AI agents, the important part isn’t simply the stability.

It is the combination of stability + programmability + global accessibility.

An AI agent can interact with blockchain infrastructure through software. It can hold digital assets in a wallet, check balances, sign transactions according to its permissions, and interact with smart contracts.

That creates the possibility of a machine-controlled financial account.

Instead of an AI agent saying:

“I need a human to approve this $5 payment.”

the system could be designed to automatically execute the payment when predefined conditions are satisfied.

For example, an AI research agent might have a wallet funded with $100 in stablecoins. It could spend a maximum of $2 per API request, $10 per day on data, and $25 per week on specialized services.

These rules can potentially be enforced through smart contracts, wallet permissions, spending limits, and other programmable controls.

Machine-to-Machine Payments

One of the most interesting applications is machine-to-machine commerce.

The internet was originally designed primarily for humans to communicate and transact. AI agents introduce a new possibility: software communicating and transacting with other software.

Consider a network of specialized agents.

One agent performs market research.

Another analyzes financial data.

A third provides computational resources.

A fourth verifies information.

Instead of every transaction passing through a human-controlled billing process, agents could pay one another directly.

For example:

Agent A → pays stablecoins → Agent B → receives data → Agent A

The payment could happen automatically based on predefined conditions.

At large scale, this could create a new digital economy where tiny transactions occur continuously between autonomous software systems.

Why Stablecoins Instead of Volatile Crypto?

AI agents need predictable economics.

Imagine an autonomous agent with a budget of $1,000.

If it holds a highly volatile cryptocurrency, the purchasing power of that budget could change dramatically. A service that costs $20 today might effectively consume substantially more or less of the agent’s available capital tomorrow.

Stablecoins can reduce that problem.

A dollar-denominated stablecoin gives the agent a relatively predictable unit for budgeting, accounting, pricing, and payments.

That matters particularly for:

• API usage
• Cloud computing
• Data purchases
• Subscription services
• Digital labor
• Advertising
• Automated commerce
• Agent-to-agent payments

If AI agents are going to participate in real economic activity, predictability may be more valuable than speculation.

Stablecoins Could Enable Micropayments

Traditional payment infrastructure isn’t always optimized for extremely small, frequent transactions.

Blockchain-based stablecoin payments could potentially support smaller transactions with automated settlement, depending on the network and its transaction costs.

This opens the door to interesting business models.

An AI agent might pay:

• $0.01 for a data point
• $0.05 for a computation
• $0.10 for an API request
• $0.50 for a specialized analysis
• $2 for a completed task

Instead of purchasing a large subscription, an agent could potentially pay precisely for what it consumes.

This could transform the economics of digital services.

Rather than humans subscribing to software, software could dynamically purchase services from other software.

Stablecoins Could Give Agents Global Payment Rails

Another major advantage is geographic reach.

Traditional financial infrastructure remains fragmented across countries, banks, payment networks, currencies, and regulatory systems.

Stablecoins operate on blockchain networks that can be accessed globally.

For AI agents operating across borders, this could simplify settlement.

An AI company in one country could operate an agent that purchases computing services from another provider, while a third-party agent supplies specialized data from another region.

Stablecoins could provide a common settlement asset across these interactions.

The AI agent doesn’t necessarily need to understand banking systems in every country.

It simply needs to understand the payment rules of the digital network it operates on.

AI Agents Could Become Their Own Economic Identities

This leads to an even bigger concept.

Today, an AI agent usually operates under the identity and financial accounts of a person or company.

In the future, agents could potentially have their own cryptographic identities, wallets, permissions, and transaction histories.

That does not necessarily mean an AI becomes a legal person.

Instead, it could mean that an agent becomes a distinct economic software entity.

For example:

Agent ID: ResearchAgent-204
Wallet: Dedicated blockchain address
Budget: $500/month
Spending limit: $20/transaction
Allowed services: Data + computing
Approval threshold: Human authorization above $20

This structure could make autonomous systems easier to monitor and control.

Blockchain transactions could also provide an auditable record of what the agent spent and where the funds went.

The Combination of AI + Smart Contracts Is Powerful

AI agents are good at making decisions.

Blockchains and smart contracts are good at executing deterministic rules.

Stablecoins connect the two through money.

That creates a potentially powerful architecture:

AI → Decision

Smart Contract → Rules

Stablecoin → Value

Blockchain → Settlement

Consider an autonomous procurement agent.

The AI determines that a company needs additional computing capacity. It compares providers, selects one based on price and performance, and initiates the purchase.

A smart contract could enforce the agreed conditions.

The stablecoin payment could be released when those conditions are satisfied.

The blockchain records the transaction.

In this model, AI handles the intelligence while blockchain handles coordination, ownership, and settlement.

The Challenges Are Just as Important

Stablecoins are not a magic solution.

AI agents managing money introduce serious risks.

Security

If an AI-controlled wallet is compromised, attackers could potentially gain access to its funds.

Agents therefore need strong wallet security, permission systems, spending limits, and transaction controls.

Hallucinations and Bad Decisions

An AI agent can make incorrect decisions.

If an agent is allowed to spend money autonomously, an incorrect assumption could become a financial loss.

This makes human oversight and programmable constraints extremely important.

Smart Contract Risk

Smart contracts can contain vulnerabilities.

An AI agent interacting with poorly designed contracts could potentially expose its funds to unnecessary risks.

Regulatory Uncertainty

Stablecoins operate within an evolving regulatory environment.

Different jurisdictions may impose different requirements on issuers, users, payment providers, and businesses.

AI agents participating in financial transactions could introduce additional compliance questions.

Privacy

Blockchain transactions can be transparent.

That can be useful for auditing, but it may also expose information about an agent’s activities, counterparties, and spending patterns.

Future systems may therefore need privacy-preserving technologies alongside transparent settlement.

The Bigger Picture: An Economy of Autonomous Agents

The most important idea isn’t simply that AI agents could use stablecoins.

It is that AI agents could become participants in digital markets.

Imagine millions of specialized agents operating simultaneously.

Some agents generate content.

Others analyze data.

Some manage logistics.

Others provide computing power.

Some negotiate prices.

Others verify information.

They could continuously interact, purchase services, sell capabilities, and exchange value.

Humans would still define objectives, budgets, permissions, and constraints—but machines could handle much of the execution.

Stablecoins could serve as one of the financial primitives that makes this economy possible.

Stablecoins May Become the Financial Language of AI

The next phase of AI may not be defined solely by how intelligent models become.

It could also be defined by what those models are allowed to do.

An AI that can only generate text is powerful.

An AI that can use tools is more capable.

An AI that can independently coordinate resources, purchase services, and receive payments becomes something fundamentally different: an economic actor operating in the digital world.

Stablecoins could provide the predictable, programmable settlement layer required for that transition.

The combination of AI agents, blockchain networks, smart contracts, and stablecoins could therefore create an entirely new category of machine-driven commerce.

The future internet may not just connect people.

It may connect agents that work, negotiate, transact, and pay each other around the clock.

And when machines start doing business with machines, they will need money that machines can actually use.

REQUEST AN ARTICLE

Key takeaways

• Uniswap falls nearly 6% on Wednesday after declining 5% in the previous session.
• Uniswap has launched Continuous Clearing Auctions on Avalanche, allowing teams to conduct on-chain token sales and bootstrap liquidity.
• UNI’s social dominance and volume have fallen sharply, signaling weaker retail attention.

Uniswap (UNI) faces intense selling pressure on Wednesday, falling nearly 9% after recording a 5% decline the previous day.

The pullback comes despite Uniswap’s continued product expansion, including the introduction of Continuous Clearing Auctions on Avalanche. The feature allows blockchain projects to conduct fully on-chain token auctions and establish initial liquidity through Uniswap v4.

However, declining social activity and derivatives demand suggest the launch has not been enough to offset the cryptocurrency market’s broader risk-averse mood.

Continuous clearing auctions launch on Avalanche

Uniswap’s Continuous Clearing Auctions provide Avalanche developers with a new mechanism for launching tokens and bootstrapping liquidity onchain.

The model is designed to reduce friction during token distribution by allowing teams to conduct auctions transparently through smart contracts. Projects can then connect their newly distributed tokens with Uniswap v4 liquidity.

The launch expands Uniswap’s presence on Avalanche and strengthens its role as infrastructure for token issuance, trading and liquidity management.

It follows the recent launch of the TradePools platform on Robinhood, which allows users to deposit USDC, USDT or ETH in pursuit of yield.

While these developments may support Uniswap’s long-term utility, they have yet to produce a meaningful improvement in near-term demand for UNI.

Retail interest in Uniswap is weakening as traders prepare for the release of July’s US Consumer Price Index report, scheduled for Wednesday at approximately 12:30 GMT.

The CPI reading could influence the Federal Reserve’s next interest-rate decision and affect demand for risk assets. A hotter-than-expected report could strengthen expectations for tighter monetary policy, while softer inflation could improve sentiment across cryptocurrency markets.

Santiment data shows Uniswap’s social dominance fell to 0.08% on Tuesday from 0.19%. Social volume also declined to 40 from 152.

The sharp contraction indicates that UNI accounts for a smaller share of cryptocurrency discussions and is attracting less attention from retail traders.

Uniswap’s derivatives market reinforces the decline in retail participation.

CoinGlass data shows UNI futures open interest fell more than 3% over the past 24 hours to $261.60 million. The decline indicates traders are closing positions and reducing their leveraged exposure.

Long liquidations reached $2.88 million during the same period, significantly exceeding short liquidations of just $1,950. The imbalance shows that falling prices have disproportionately forced bullish traders out of their positions.

However, UNI’s open-interest-weighted funding rate improved to 0.0016% from negative 0.0054% the previous day.

The return to positive funding indicates that the remaining leveraged market carries a slight bullish bias. Still, falling open interest and heavy long liquidations suggest overall sentiment remains fragile.

Uniswap Technical outlook: UNI tests 100-day EMA

Uniswap is testing its 100-day Exponential Moving Average at $3.55, an important near-term support level.

UNI remains below the 50-day EMA at $3.65 and the 200-day EMA at $3.93. These moving averages create overhead resistance and reinforce the prevailing bearish structure.

The Relative Strength Index has declined to 40, placing it below the neutral midpoint of 50 and indicating growing selling momentum. However, the indicator remains above the oversold threshold of 30.

The Moving Average Convergence Divergence indicator has also fallen below its signal line, while its expanding bearish profile suggests downside momentum is strengthening.

A decisive daily close below the 100-day EMA at $3.55 could extend Uniswap’s decline toward the 50% Fibonacci retracement level at $3.25. This level is measured from UNI’s advance between $2.31 and $4.57.

A successful defense of $3.55 could allow buyers to attempt a recovery. However, UNI must reclaim the 50-day EMA at $3.65 to ease immediate selling pressure.

Above that level, the 23.6% Fibonacci retracement at $3.89 and the 200-day EMA at $3.93 form a significant resistance cluster.

Until Uniswap recovers above these moving averages with stronger trading activity, the short-term outlook is likely to remain bearish.

An attacker drained 99.7% of the XRP reserve backing the Coreum cross-blockchain bridge on August 9 by conjuring fake evidence of deposits that fooled the bridge’s own operators into authorizing real withdrawals.

Nobody stole anyone’s private key to execute the clever hack. Instead, Coreum’s bridge liquidity account paid out 199,916 XRP worth over $200,000 across 94 transactions, each one carrying a majority of signatures from its own relayers. 

Hours later, Coreum was shocked to discover it held just 493 XRP worth roughly $500.

A memo was all Coreum’s bridge required

The relayer software watched the bridge account’s history for payments carrying a Coreum recipient memo.

It never verified that payment destination. A seemingly valid memo existed, yet the destination was the hacker’s wallet.

Ostensibly independent operators reached identical conclusions and signed off on all of the withdrawals because they were all running the same buggy code.

Russia’s central bank will allow retail investors to only trade bitcoin BTC$64,148.66, ether ETH$1,912.59 and USDT on regulated exchanges, making Tether’s dollar-linked token the only stablecoin on the initial list.

The draft rules would limit non-qualified investors to 300,000 rubles (around $3,600) of crypto purchases per year at each intermediary. Qualified investors wouldn’t face the cap.

The whitelist adds detail to legislation passed in July that opens regulated crypto trading from Sept. 1 but did not specify which assets retail investors could buy. Crypto payments inside Russia remain prohibited.

The wording sets the 300,000-ruble limit per intermediary rather than across an investor’s total purchases, potentially allowing larger aggregate exposure through multiple brokers or exchanges.

It isn’t often that a reporter gets asked to pose as a venture capitalist to fool suspected North Korean IT workers.

But in June, I found myself joining a Zoom call as “Aelin Ashriver,” an investor from the fictitious Definitive Communications, to meet the development team of crypto startup Ballena Azul.

The IT workers on the call believed they were pitching for VC backing for their startup. In reality they had spent weeks working inside a fake crypto company set up purely to study their methods and infrastructure by Mauro Eldritch, founder of cybersecurity firm BCA LTD, and Heiner García, a cyber threat intelligence analyst at Telefónica Tech and founder of NorthScan.

Cointelegraph tagged along for one stage of the investigation.

During the call, I played up the ruse by suggesting I might even be able to land Ballena Azul some coverage in Cointelegraph.

So at least someone was telling the truth.

Suspected DPRK IT workers pitch for venture capital backing from the fictitious Definitive Communications, played by Cointelegraph. Source: ANY.RUN

Building a company for suspected North Korean IT workers

Eldritch and García built the fictitious Ballena Azul with infrastructure provided by cybersecurity platform ANY.RUN. An existing UK registration for an unrelated company of the same name, which was dissolved in 2022, added legitimacy to the project.

Eldritch assumed the identity of co-founder “Leonardo Nelson,” while García took on the alias “Andy Jones” and posed as the company’s team lead.

Related: North Korean cyber spies are no longer just remote threats

One of the most valuable pieces of intel that the five-week ruse exposed were the external servers the workers used as intermediary points before connecting to Ballena Azul’s controlled virtual desktops.

Exposed servers were particularly valuable because such infrastructure is often recycled across operations and can remain active for long periods.

García tells Magazine the servers were associated with malware families linked to North Korean campaigns that steal credentials, crypto wallet data and other sensitive information.

“Some of the servers we found were tied back to distributing InvisibleFerret and BeaverTail/OtterCookie in prior years and were active to this day,” he says.

But some others were totally new and had zero intelligence about them, looking clean and keeping outside of mainstream block lists or threat feeds.”

He adds that the infrastructure could serve multiple purposes, with servers previously used for malware distribution also acting as command-and-control infrastructure, and as proxies for operators carrying out their day-to-day work.

The suspected workers do not need to deploy malware to pose a threat, according to the researchers. Once hired, they can gain legitimate access to a company’s internal systems, source code and other sensitive information. The longer they remain undetected, the longer they can continue drawing salaries that researchers say ultimately help fund the North Korean regime.

The operation also showed the group relied on artificial intelligence tools to help compensate for gaps in their technical knowledge. They used ChatGPT for writing and coding, including to answer basic questions and complete assignments they struggled with themselves. They preferred Google Gemini for image alteration and document forgery.

A suspected DPRK IT worker and ChatGPT team up in an attempt to obtain testnet crypto during the Ballena Azul operation. Source: ANY.RUN

Other tools employed included remote desktop software, crypto wallets and a service for sharing two-factor authentication codes.

North Korean IT workers have become a growing cybersecurity threat to the cryptocurrency industry. Consensys said in July that it had engaged a North Korea-linked developer through a third-party service provider before identifying the threat and cutting off access.

In another case, US prosecutors charged four North Korean nationals in 2025 with using false identities to obtain remote IT jobs and allegedly stealing more than $900,000 in cryptocurrency from two companies, including a US blockchain research and development firm.

The US Treasury said in March that North Korean IT worker schemes generated nearly $800 million in 2024 to help fund the Pyongyang regime’s weapons-of-mass-destruction programs.

Inside fake crypto company Ballena Azul

The ruse began when García connected with a recruiter via GitHub, who had been linked to Famous Chollima, a threat group associated with North Korean IT worker operations.

García said that Ballena Azul needed to hire software developers and the recruiter offered up “Jack Anderson,” “Angelo Espree” and “Lucas Theo.” At least two of them presented US identification.

The trio were given various programming assignments inside controlled virtual desktop environments, which allowed García and Eldritch to observe how they worked.

Angelo Espree was one of the developers onboarded through a recruiter associated with DPRK operations. Source: ANY.RUN

The researchers also deliberately introduced technical problems, including selective network outages and disappearing mouse cursors, to see how the suspected workers reacted and which tools they turned to when things went wrong.

“Honestly, the biggest surprise was how much of it ran on improvisation,” García says. “There was no rigid playbook, no polished corporate process behind them.”

During their many weeks working inside the controlled environments, the suspected North Koreans left behind a treasure trove for the researchers, including chat logs, AI conversations, crypto wallet information, VPN exit nodes and hours of live video footage. Their connections also exposed the servers that became one of the investigation’s most valuable findings.

To be sure, the heavy AI reliance isn’t unique to the workers hoodwinked in Ballena Azul’s operation. 

Ballena Azul workers generally used AI as a crutch for coding and technical tasks they struggled with. Reuters reported Monday that another North Korean hacking group, Kimsuky, was using AI for a more offensive purpose. The group was reportedly running AI tools locally to help automate cyberattacks, analyze stolen data and produce more convincing phishing campaigns.

Evolving playbook of remote DPRK IT workers

This was not the first time Cointelegraph has played a minor role in exposing suspected North Korean workers.

In February 2025, García and Cointelegraph conducted a job interview for a suspected operative calling himself “Motoki.” The developer claimed to be Japanese but ragequit the interview after being asked to introduce himself in his mother tongue.

Still, García kept communicating with him. Motoki eventually offered to send García money to buy a computer that he could access remotely, allowing him to work through a local machine instead of connecting through a VPN to bypass restrictions used by employers and freelance platforms.

Related: From Sony to Bybit: How Lazarus Group became crypto’s supervillain

García later documented suspected North Korean operatives recruiting freelancers to provide verified accounts, identities and remote access to their computers. In one version of the scheme, operatives could work through machines physically located in the US, making them appear to employers and freelance platforms as US-based contractors.

In May, two US “laptop farmers” — people who hosted a cluster of computers that North Koreans could remotely access — were sentenced to 18 months in prison for helping DPRK IT workers pose as US-based employees in schemes that generated more than $1.2 million and affected nearly 70 companies.

Taking Ballena Azul down

All fake things must come to an end, so the researchers introduced “Benito Camella,” Ballena Azul’s co-founder, who had supposedly been focused on other business in Milan while the company expanded.

When he returned, Camella confronted the workers over discrepancies in their identities and documents. The confrontation quickly began to clear the chat room. Espree left the video call first, while Anderson stayed longer before realizing the scheme was unraveling.

“Are you living two lives, Mr. Anderson?” Camella asks Jack Anderson during the confrontation. Source: ANY.RUN

But the researchers kept the deception going even after the meeting ended. In the company’s Telegram channel, the “CEO” accused “Andy Jones” of bringing in “illegal workers” and putting the company at risk. “Jones” responded that he had been under pressure to build a team quickly and was not being paid enough to do it. He maintained that he had done the best he could with what he had.

The staged argument ended with the fake CEO terminating both their working relationship and friendship, keeping up the appearance that Ballena Azul had collapsed because of a disastrous hiring decision.

One of the suspected North Koreans later contacted García privately to apologize for what had happened and ask whether he was all right.

According to the researchers, they never heard from the rest of the group again.

To this day, they say, the suspected workers do not know they wasted weeks working inside an environment built to extract intelligence from them.

Magazine: Do the Coldcard attacks mean all hardware wallets are now insecure?

Editor’s note: Cointelegraph could not independently confirm the nationality or affiliation of the suspected DPRK IT workers, and no government agency has publicly identified them.