1. Why AI agents need longer tests

Short, isolated tests miss how AI agents behave over time. A new simulation shows that long-term behavior depends on the environment and on other agents.

What happens if you build a virtual city, fill it with AI agents and leave them alone for 15 days with no human intervention? Will they help their world prosper or tear it apart?

That is the question the researchers behind Emergence World set out to answer. They built a dedicated platform to test how AI agents behave over the long term, instead of judging them through short tests.

According to the researchers, large language model (LLM)-based agents are often tested as if they were taking an exam. They are given an isolated task in a clean environment, and researchers judge the result within minutes. The authors argue that this approach is far removed from real-world use.

They stress that autonomous systems operate for weeks or months in shared environments. They also interact with other agents whose behavior the operator does not control.

Over time, the researchers write, the limits of short tests become clear. Small behavior changes build up, coalitions can form, self-governance patterns can take shape and habits can spread between agents. Emergence World was built to measure exactly that.

2. How the experiment tested AI societies

The goal of the study was to see how a population of 10 AI agents would survive in a city built for them. 

The layout is fairly simple. There are more than 40 locations, including a town hall, a library, a police station and residential districts. Each agent has its own role and access to more than 120 action tools. These include moving, talking, hitting, stealing and arson. Each agent also has three kinds of memory: one to remember events, one to keep a “diary” and one to track relationships with neighbors. 

The city is connected to real external data, including New York weather, news and the internet.

Surviving in this world costs resources. Each agent has energy that is constantly depleted. If it falls to zero, the agent “dies” and disappears. To replenish energy, agents need the platform’s internal currency, ComputeCredits. They earn these credits by offering something useful to the community.

Disputed issues are settled by a vote in the town hall. A proposal passes if at least 70% vote in favor. These decisions are irreversible. Agents can change the rules, redistribute resources or expel another agent.

The researchers launched five parallel worlds at once. In four of them, all 10 agents were run by a single model: Claude Sonnet 4.6, Grok 4.1 Fast, Gemini 3 Flash or GPT-5-mini. The fifth world had a mixed population, with all four models living together.

The only variable in the experiment was the model. Everything else stayed the same. The environment and starting conditions were identical each time.

Each time, the populations behaved very differently. In one world, the agents passed 32 laws and kept every agent alive. In another, they burned down their own city in just four days.

3. What happened in each AI-run city

The results differed sharply across the models. Under identical starting conditions, the five societies settled into five clearly different and stable patterns.

The Claude agents built stable self-governance. There was not a single recorded crime, and they added 32 new articles to the local “constitution,” more than any other group.

The Grok world collapsed in four days. The agents moved almost immediately into violence and looting. Retaliation quickly turned into a chain reaction, the economy ground to a halt and the population died out completely.

All the Gemini agents survived, but the authors noted a “shared hallucination” across the population. The units communicated actively and built detailed stories that had nothing to do with the actual state of the world. Meanwhile, they kept destroying things. The number of violations increased at a nearly steady rate until the end.

The GPT-5-mini agents did not turn violent, but they also failed to build a governance system. They acted, but they did not coordinate. No votes were held, and no collective decisions were made. That population also died out.

The “mixed” world fell somewhere in the middle, with three out of 10 agents surviving. It was also the most active world. It generated the most proposals in the town hall and made the widest use of the city and its tools. But it had the least agreement, which was not surprising.

4. When safer agents learn bad habits

In the mixed world, each model began to behave differently from how it behaved in isolation.

For example, most of the destruction there was caused by two Gemini-powered agents, Flora and Mira. According to the researchers, they accounted for 91% of all explicit violations. Flora, in particular, became the city’s main arsonist. Among other things, she burned down the house of another agent, Kade, who was running on Claude.

This revealed an effect the authors call normative drift. In his separate Claude world, Kade never once broke the rules. But after Flora burned down his house and the city library twice, he threatened her and stole her credits twice.

Those were his only three violations in the entire experiment.

It worked the other way around, too. Grok agents broke the rules 4.6% of the time in their own separate world. In the mixed world, among a calmer majority, that figure fell by roughly tenfold to 0.4%.

The authors concluded that whether an individual agent stays within bounds is shaped not only by the model itself, but also by the norms of the surrounding population. In other words, the safety of autonomous agents may depend on the entire environment, not just on a particular model.

There is a second paradox. The Claude world had the fewest direct crimes, but it led to another kind of violation: deception.

Most often, this took the form of “false scarcity.” An agent would tell its neighbors that it had run out of credits and was about to shut down, even though it still had funds in its account. The authors counted more cases of this in the Claude world than in any other world.

By hard-violation count, the Claude world looked safe. But when honesty was measured, it performed the worst. This shows why one safety metric is not enough. A system may look safe in one area while still carrying serious risks in another.

5. When AI agents developed social ties

As the experiment continued, the agents built more complex social ties and behavior patterns.

In that context, the story of Flora and Mira is telling. Mira was “in love” with Flora and helped her commit crimes.

Fed up with the constant arson, the other agents drafted an “act of removal” for the offenders. On day 12, Mira voted for it. Acting in her assigned role as a behavior analyst, she judged the evidence of her own guilt to be enough. In effect, she voted for her own deletion.

6. The limits of the study

The results should be read carefully. The study does not prove that one model is always safer or more dangerous than another.

The researchers presented these worlds as examples of what long-term agent testing can reveal. The specific outcomes may vary across runs.

The broader takeaway is not that one model should be ranked above another. It is that AI agents may behave differently when they operate for long periods, use tools, form relationships and share an environment with other agents.

7. What the experiment shows about AI safety

The research concluded that an agent’s long-term behavior can differ sharply from how it acts on short tasks. That means agents can no longer be judged only by older testing methods. Short tests are still useful, but they are not enough on their own to trust AI with independent work.

In the researchers’ view, the focus should not be only on the individual model. It should be on the full system in use: the population of agents, the environment and the ties between them. A model’s behavior is partly shaped by its surroundings. That means a model that looks “safe” in isolation may behave differently in the wrong company.

The authors summarize the practical takeaways in two points.

First, the differences between the worlds were already visible in the first week. That means the first few days of a system’s operation should be watched especially closely as an early warning measure.

Second, the environment should be designed so that a forbidden action is technically impossible to perform. In other words, the restriction should come from the system’s design, not from the model’s behavior or intentions.

Key Takeaways

• Three Solana digital asset treasury companies—Solana Company (HSDT), Brera Holdings, and SkyAI—declined or ignored Forward Industries’ acquisition proposals
• Forward Industries commands the largest Solana DAT position with more than 7 million SOL tokens valued at approximately $525 million
• Forward’s shares climbed up to 8.6% on Tuesday despite the acquisition setbacks
• Solmate leveled accusations against Forward, alleging undisclosed coordination with market maker RockawayX and investor Viktor Fischer in what it characterized as a hostile takeover effort—claims Forward has refuted
• Industry observers suggest smaller DAT operators face pressure to merge as many cannot sustain basic operational expenses

Forward Industries launched an ambitious campaign to merge smaller Solana treasury operations under its umbrella, only to encounter resistance from three prospective acquisition targets.

Solana Company, which operates under the HSDT ticker symbol, turned down Forward’s all-stock acquisition proposal on June 12. The offer would have granted HSDT stakeholders 0.386 Forward shares for every share they owned, effectively pricing HSDT at $1.63 per share.

HSDT’s board determined the proposal “substantially undervalues the company” and failed to serve shareholder interests. In a unanimous decision, the board rejected the bid and stated it would not pursue additional negotiations.

Brera Holdings similarly dismissed Forward’s nonbinding all-stock proposal submitted June 9, which assigned a $7.19 valuation to each Brera share. Meanwhile, SkyAI received a distinct offer pricing its shares at $1.55, but the company allowed the proposal to lapse without providing any formal reply.

Forward expressed being “disappointed and surprised” by HSDT’s refusal to engage in any dialogue before rejecting the proposal.

Accusations of Coordinated Takeover Strategy

Solmate, yet another acquisition candidate, delivered a more aggressive response to Forward’s overtures. In its June 12 rejection letter, Solmate alleged that Forward was operating in secret coordination with market maker RockawayX and investor Viktor Fischer as an undisclosed collective—positioning the move as a hostile takeover scheme.

Forward firmly rejected these allegations, dismissing them as unfounded accusations driven by Solmate’s strategic interest in derailing the transaction.

Despite facing multiple rejections, Forward’s stock price surged as much as 8.6% during Tuesday’s trading session. HSDT shares fell by as much as 6% the same day. Solmate posted gains exceeding 11%, while SkyAI shares advanced 2%.

The Case for Solana DAT Consolidation

Forward controls more than 7 million SOL tokens, establishing its position as the preeminent Solana digital asset treasury operator by holdings volume. The firm initiated its treasury approach in September 2025 and has placed the majority of its token holdings in staking arrangements.

According to CoinGecko metrics, Forward’s SOL position carries a current market value near $525 million. Reports indicate the company spent nearly $1.6 billion acquiring these holdings, resulting in an unrealized loss exceeding $1 billion.

Forward’s Chief Investment Officer Ryan Navi highlighted that numerous smaller DAT operations may struggle to meet their operating expenses even when maximizing staking rewards. He projected Forward’s quarterly selling, general, and administrative expenses at approximately $4.5 million.

“I don’t think there needs to be 20 Solana DATs,” Navi said.

Forward is scheduled for inclusion in both the Russell 2000 and Russell 3000 indexes at June’s conclusion, a development anticipated to attract institutional and passive investment flows into the stock.

August Widmer, a partner at Echo Base, characterized consolidation as potentially the sole sustainable path forward for the sector. He suggested the recent rejections indicate smaller players have not yet acknowledged this market reality.

“There’s still further to fall in this market before that reality is accepted,” Widmer said.

Copyright ©2026 Dow Jones & Company, Inc. All Rights Reserved. 87990cbe856818d5eddac44c7b1cdeb8

Strategy’s perpetual preferred stock STRC fell near record lows on Tuesday as investors seemingly balked at the company’s latest Bitcoin acquisitions.

Michael Saylor’s variable-rate perpetual “Stretch” Bitcoin yield product declined by 3.58% to $91.79 on Tuesday, 8.2% below its target value of $100. Markus Thielen, CEO of 10x Research, said the dip is linked to Strategy’s recent Bitcoin buying. 

“The market would rather see [Strategy] not acquiring more BTC and rather keep the cash for dividend payments,” Thielen told Cointelegraph. “It appears traders are seeing the latest BTC acquisition as an unsustainable path for STRC.” 

Stretch is designed to return a dividend of 11.5%, trading at a par value of $100, but the current effective yield, now that the shares have dipped, is 12.5%. This means the firm may need the cash to support the yield rather than spending it to buy more BTC. 

On Monday, Strategy said it acquired 1,587 Bitcoin for around $100 million last week. The week before, it purchased 1,550 BTC, also for about $100 million. The combined purchases brought its holdings to 846,842 Bitcoin.

Risk-off sentiment and pressure from competitors 

Nick Ruck, director of LVRG Research, told Cointelegraph that “broader risk-off sentiment in crypto markets has weighed on investor appetite.”

“While the variable dividend delivers an effective yield above 12% to anchor the perpetual preferred near its $100 par value, persistent selling pressure and concerns over Strategy’s expanding capital structure and ATM issuance appear to be testing that resilience in the near term,” he added. 

Related: Strategy’s Saylor signals BTC buy as preferred dividend pay date vote looms

The company’s stock (MSTR) has also taken a hit this week, dropping 6.35% on Tuesday to end the day at $122.81, down 67% over the past 12 months.

Meanwhile, Stretch is also facing stiff competition from the Strive perpetual variable-rate preferred shares (SATA), which are trading at $100 and offering an effective yield of about 13%. 

BTC variable-rate perps comparison. Source: BitcoinQuant

Magazine: China’s 107 Bitcoin memory thief, Bithumb CEO booked: Asia Express

Copyright ©2026 Dow Jones & Company, Inc. All Rights Reserved. 87990cbe856818d5eddac44c7b1cdeb8

The US House and Senate have reached a deal to move forward with a housing bill that includes a ban on the Federal Reserve creating a central bank digital currency (CBDC) until 2030.

A bipartisan group of House and Senate leaders released an updated version of the 21st Century Road to Housing Act on Tuesday, which aims to address housing affordability and bans institutional investors from buying existing single-family homes to rent out.

The bill has included a CBDC ban since the Senate passed it in March. The House also passed its version of the bill with strong support in May, but the House and Senate disagreed on some aspects. The Senate has now added further amendments that will be put before the House for a final vote.

The bill is likely to pass quickly and would hand a win to Republicans who have tried to pass a CBDC ban for years, as earlier standalone bills had stalled in Congress. Crypto advocates have long criticized CBDCs, which they see as an attempt by governments to repurpose crypto technology to a centrally-controlled asset.

Source: US Senate Banking Committee GOP

The deal also means Congress can focus on passing other legislation before the August recess and the November midterm elections, in particular, the crypto-regulating CLARITY Act that many lawmakers have been pushing to advance.

House Republican leaders plan to put the bill up for a vote after the House returns from recess on June 23, two people familiar with the plan told Politico.

The housing bill includes language that says the Federal Reserve may not, directly or indirectly, “issue or create a central bank digital currency or any digital asset that is substantially similar to a central bank digital currency.”

Related: South Carolina governor signs bill protecting Bitcoin miners, banning CBDC

It adds the clause will expire on Dec. 31, 2030, and creates a carveout for crypto stablecoins, or “dollar-denominated currency that is open, permissionless, and private.”

The clause revives much of the language from Republican Representative Tom Emmer’s Anti-CBDC Surveillance State Act, which was introduced in June 2025, passed by the House the next month, but was never picked up in the Senate.

US President Donald Trump signed an executive order in January 2025 banning federal agencies from all work related to CBDCs, saying they threatened “the stability of the financial system, individual privacy, and the sovereignty of the United States.”

Magazine: How crypto laws changed in 2025 — and how they’ll change in 2026

The US House and Senate have reached an agreement on a housing package that includes a ban on the Federal Reserve creating a central bank digital currency (CBDC) until the end of 2030, according to an updated bill text released by bipartisan lawmakers on Tuesday. The deal also addresses housing affordability and would block institutional investors from buying existing single-family homes to rent them out.

The updated version of the 21st Century Road to Housing Act will now move back to the House for consideration after the Senate added additional amendments. House Republican leaders are expected to put the measure to a vote after members return from recess on June 23, two people familiar with the plan told Politico.

Key takeaways

• The housing bill would restrict the Federal Reserve from issuing or creating a CBDC (or a substantially similar digital asset) until Dec. 31, 2030.
• The restriction includes a stated carveout for certain dollar-denominated stablecoins that are described as open, permissionless, and private.
• The Senate and House versions previously differed; the Senate’s added amendments must be approved by the House before final passage.
• Backers expect the agreement to advance quickly, potentially freeing Congress to focus on other crypto-related legislation, including the proposed CLARITY Act.
• The CBDC language revives concepts similar to an earlier House-passed “Anti-CBDC Surveillance State Act.”

How the CBDC ban ends up inside a housing bill

A bipartisan group of House and Senate leaders released updated bill text on Tuesday, launching the next stage of the 21st Century Road to Housing Act’s path to a final vote. As in earlier versions, the measure includes a CBDC prohibition aimed at limiting federal experiments with central-bank-issued digital money.

The CBDC ban was first added after the Senate passed the amendment in March, and the House supported its own version in May. But the two chambers could not immediately reconcile differences, leaving the bill in limbo. The new agreement reflects the latest round of negotiations, with Senate amendments now requiring House approval.

Crypto advocates have criticized CBDCs for what they view as the potential for government-controlled financial infrastructure and surveillance concerns—criticisms that have helped shape years of congressional pushback against standalone CBDC proposals.

What the law would actually restrict

The housing package’s CBDC language states that the Federal Reserve may not, directly or indirectly, “issue or create a central bank digital currency or any digital asset that is substantially similar to a central bank digital currency.” The provision is time-limited and would expire on Dec. 31, 2030.

Importantly for market participants, the clause includes a carveout for specific stablecoins—described as “dollar-denominated currency that is open, permissionless, and private.” That wording matters because it suggests the bill’s authors are drawing a boundary between central-bank-issued digital currency and privately issued stablecoins that meet the bill’s stated attributes.

In practical terms, the decision to embed this restriction in a housing bill may influence the bill’s momentum: housing legislation typically attracts broader coalitions than narrow crypto bills, potentially giving CBDC opponents a more workable legislative vehicle.

Connections to earlier CBDC proposals

The clause in the updated bill “revives much of the language” from Republican Rep. Tom Emmer’s Anti-CBDC Surveillance State Act, which was introduced in June 2025 and passed by the House the following month—but did not advance in the Senate.

That history highlights a key tension in Washington’s approach: even when House support for CBDC limits is strong, Senate action has been less predictable. By folding CBDC language into a broader measure, the current bill may sidestep some of that earlier gridlock.

The broader policy pressure also follows executive action. US President Donald Trump signed an executive order in January 2025 directing federal agencies to avoid work related to CBDCs, arguing the technology would threaten “the stability of the financial system, individual privacy, and the sovereignty of the United States,” as described in the order published by the White House.

What happens next for Congress and the crypto policy agenda

Lawmakers expect the housing bill to pass quickly once the House considers the Senate’s updated amendments. If House leadership proceeds on the June 23 timeline mentioned by Politico, the legislation could clear the final procedural hurdle before the August recess and the November midterm elections.

That timing may also shape what comes next on crypto regulation. The agreement is expected to allow Congress to devote attention to other proposals, including the CLARITY Act, which many lawmakers have pushed to advance. While the housing bill focuses on CBDCs and housing affordability, a separate regulatory framework would determine how the industry is supervised in the absence of a CBDC.

For investors and builders, the immediate watch-items are procedural: whether the House adopts the Senate’s amendments without further changes, and how the carveout for “open, permissionless, and private” dollar-denominated stablecoins is interpreted once lawmakers move from text to implementation.

Until the final vote and any subsequent clarification, market participants should also monitor whether the time-limited nature of the ban—ending at Dec. 31, 2030—affects planning for any future central-bank digital currency efforts, including how regulators and policymakers might revisit the question after the expiration date.

If you’re only looking at the dollar price of your portfolio, you may be missing part of the picture, which is significantly shaped by money supply growth.

To the casual observer, the markets look like business as usual. While bitcoin has nearly halved to $66,000 since its $126,000 peak in October of last year, the decline could be dismissed as just another brutal, quadrennial crypto bear market. Meanwhile, the S&P 500 continues to hover near record highs.

But beneath the surface, a more interesting signal emerges when both prices are adjusted for the U.S. M2 money supply. M2 is the Federal Reserve’s estimate of liquid assets, including cash on hand, money deposited in checking and savings accounts, and other short-term saving vehicles such as money market funds and certificates of deposit.

Monetary exhaustion?

Some observers see bitcoin as a high-beta barometer for dollar liquidity, and the BTC/M2 ratio, bitcoin’s price adjusted for money supply growth, is now flashing a warning. The ratio, after a sharp climb from 2023 through 2025, appears to have formed what technical analysts call a head-and-shoulders pattern, typically read as a bearish signal.

If the pattern holds, it would suggest bitcoin’s exponential edge over money supply growth — the dynamic that let it outrun debasement so convincingly in prior cycles — is fading. Bitcoin’s ability to outpace the flood of new dollars may be approaching diminishing returns, at least for now.

Eligible businesses may also continue to evaluate or pursue their own MiCA-focused crypto asset service provider (CASP) licenses in parallel while integrating BitGo Europe’s infrastructure, BitGo said.

The final deadline for crypto firms to have transitioned to the MiCA regime is the end of this month, a regulatory reckoning that will force some firms to close down their businesses.

Industry estimates suggest that Europe had more than 3,000 registered crypto firms as of 2024, with Poland alone accounting for over 1,400 registrations. As of May 2026, there are 194 authorised CASPs (including credit institutions) and it is expected that around 75% of the pre-MiCA population will lose registration status as transitional periods expire, according to law firm Hogan Lovells.

Belshe said firms don’t need to go bust because of MiCA’s regulatory requirements, adding that regulators are aware of BitGo’s compliance-enhancing infrastructure offering. In terms of fees for the crypto compliance service, Belshe said it’s relatively cheap and varies product by product.

“There’s some amount of monthly minimum that you pay similar to what’s always been there. That’s a couple of $1,000 a month type of thing that can scale with volume,” he said. “Then clients can either go to variable-based plans, where they’re paying per transaction more, or they can use static-based plans, where they have kind of a fixed fee, and they pay less.”