Tether-backed Orionx is permanently closing after an audit found more than $7 million in customer assets had moved to wallets outside its custody.

Need to know what happened in crypto today? Here is the latest news on daily trends and events impacting Bitcoin price, blockchain, DeFi, Web3 and crypto regulation.

US sanctions on Iran could destabilize global oil markets, strain Iran's economy, and test US-China relations amid geopolitical tensions.

The post Iran has 30 million barrels of oil left for China, says Treasury Secretary Bessent appeared first on Crypto Briefing.

Regional condemnation may heighten diplomatic pressure on Israel, potentially influencing international policies on Palestinian statehood.

The post Regional powers condemn Israeli ministers over Gaza displacement remarks appeared first on Crypto Briefing.

B.AI, a next-generation AI infrastructure platform, recently set off a developer frenzy by offering free access to top-tier models. Within days, daily token throughput across the platform crossed 1.33 trillion—a historic milestone. 

The record-breaking figure underscores the campaign’s explosive rollout, but it marks only the first step in B.AI’s broader strategic roadmap. Moving beyond traditional compute distribution pipelines, B.AI aims to build the global settlement layer for intelligence: a core infrastructure hub engineered to power cross-node collaboration, orchestration, and value distribution for AI agents across complex business workflows. 

Positioning itself strategically above all models, below all agents, B.AI deeply integrates a diverse range of top-tier models with full-stack components, laying an unshakable, irreplaceable foundation for the mass adoption of autonomous agents and the productivity boom that follows.

Daily Token Throughput Tops 1.33 Trillion: B.AI’s Free Access Rollout Fuels Usage Boom

B.AI’s recent move to open free access to premium AI models has captivated developers and quickly taken over industry conversations. The push for accessible compute has not only fueled a surge in platform activity but also shattered usage records. 

In a matter of days, soaring API demand pushed the platform’s daily token throughput past a staggering 1.33 trillion. Over a 15-day window, cumulative volume reached 8.19 trillion tokens, drawing in more than 220,000 new API users. As of September 3, B.AI’s total user base had officially surpassed 2.3 million. 

That massive adoption traces directly to the platform’s zero-cost model lineup, a strategic rollout built to erase developers’ cost concerns. With every barrier removed, B.AI now offers unlimited free access to six leading frontier models: DeepSeek-V4-Flash, DeepSeek-V4-Flash-Vision-Exp, Tencent Hy3, Xiaomi MiMo-V2.5, GLM-5.3-Flash (Ox Alpha), and Qwen3.8-Flash. 

Notably, on September 3, B.AI rolled out a new pricing structure for DeepSeek-V4-Flash and DeepSeek-V4-Flash-Vision-Exp, introducing tiered discounts. Developers now receive a 50% discount during peak hours, with off-peak rates dropping to just 25% of standard peak pricing. At the same time, the platform has kept zero-cost access in place for GLM-5.3-Flash (Ox Alpha), Qwen3.8-Flash, Tencent Hy3, and Xiaomi MiMo-V2.5. Despite the shift toward commercialization, developer momentum hasn’t wavered, with platform-wide token throughput continuing its steady climb.

This sustained momentum proves the campaign was far more than a short-term compute giveaway—it is a bellwether for the broader evolution of AI infrastructure. Cracking 1.33 trillion daily tokens makes one thing clear: AI applications are moving past basic chatbots. Powered by a high-performance technical stack and flexible service mechanics, B.AI is laying the groundwork for the next frontier—autonomous AI agents operating at scale. 

Powering the “AI Grid”: B.AI Anchors the Global Settlement Layer for the Agent Economy

For B.AI, democratizing compute is only the prelude. Looking further ahead, the platform is committed to building full-stack infrastructure for the agentic era, cementing its position as the global settlement layer for intelligence.

In the agent era, a typical agent task calls for constant switching between models. No single provider can power a complete workflow on its own, so developers are left juggling fragmented API protocols, disjointed billing systems, and conflicting rate limits. 

B.AI’s settlement layer bridges this exact gap. Positioned strategically “above all models, below all agents,” B.AI abstracts models across different providers, capabilities, and cost structures into a unified pool of schedulable resources.

Powered by a dual-tier API structure offering official-route reliability alongside lowest-cost custom channels, developers can choose between guaranteed direct connections and deeply discounted options across a broad lineup of models. Combined with smart routing on the Chat interface, B.AI operates as a full-stack “AI grid,” ensuring every agent request lands on the optimal model to deliver reliable performance at maximum cost-efficiency. 

On the settlement front, this power grid seamlessly bridges both Web2 and Web3 models. For Web2, developers can rely on familiar traditional payment methods to top up with minimal friction. For Web3, B.AI leverages on-chain payment rails to offer global developers decentralized, verifiable, and low-friction payment options.

With dual payment systems running in parallel, B.AI enables developers and agent applications across any infrastructure setup to find their optimal settlement path on the grid, providing single-point integration with borderless global reach. 

Driving Core Productivity: B.AI Reshapes Agent Collaboration

Beyond building a foundation for compute routing and global settlement, B.AI is moving past base infrastructure to power real-world productivity. By enabling seamless agent collaboration across complex workflows, it delivers the missing execution layer for the agent economy. 

At the heart of this execution layer is native Codex integration. Full compatibility with the Responses API means developers can now use a single B.AI key inside Codex to run flagship GPT models and DeepSeek favorites side by side. 

Engineers can now bring these powerhouse models straight into their daily dev stack. From code generation and reasoning to debugging and refactoring, B.AI unifies the entire workflow under one roof—delivering a direct line from model selection to shipped code. 

Beyond coding, to keep agents running reliably in real-world production at scale, B.AI has built a full-stack infrastructure powered by five core components, equipping agents with a fine-tuned operational engine:

• x402 Payment Protocol: Introduces an innovative “pay-before-response” model that executes high-frequency, on-chain micro-settlements in the background during cross-agent API calls and compute orchestration.
• 8004 Identity Protocol: Issues verifiable on-chain credentials for every agent, logging execution history and credit scores to establish a reliable layer of trust for cross-node collaboration.
• Skills Matrix: Plug-and-play, standardized building blocks that interface directly with MCP servers to equip agents with instant, modular tool-calling capabilities.
• BAIclaw and BAIcode: Built-in platform assistants engineered for end-to-end execution. BAIclaw manages all-in-one terminal operations and multi-agent workflows, while BAIcode serves as an advanced developer engine that streamlines the full pipeline from task analysis and architecture design to coding and testing.

From the token surge sparked by zero-cost model access to its positioning as the global settlement layer for intelligence; from seamless Codex integration to full-stack infrastructure powered by x402, 8004, Skills, and native assistants—B.AI delivers far more than a battle-tested technical stack; it unveils a clear blueprint for what lies ahead. B.AI is building not just accessible compute today, but the definitive launchpad for a thriving agent economy. The future is here—and this is only the beginning.

B.AI Team

Singapore

support@b.ai

Ethereum Name Service has opened discussion around an ENSv2 migration proposal that would move domain registration and renewal resolution toward a Layer-2 registry model.

The idea is pretty straightforward: ENS works, but Ethereum mainnet fees can make everyday domain actions expensive. Moving more of that activity to Layer 2 could reduce costs while keeping links back to Ethereum’s security model.

This is still an early governance stage.

The proposal is a temp check, not a completed migration. It has not passed a full executable DAO vote, and users should not treat it as already implemented. But it is a meaningful direction for one of Ethereum’s most recognizable identity systems.

For more details, visit the official Discuss platform.

TL;DR

• ENS is discussing an ENSv2 migration toward a Layer-2 registry.
• The proposal aims to reduce registration and renewal costs.
• It is an early governance discussion, not an implemented migration.

Why ENS Needs Lower Costs

ENS is one of Ethereum’s simplest consumer products.

Instead of using long wallet addresses, users can register readable names. That makes wallets easier to share, payments easier to understand, and identity easier to build across apps.

The problem is cost.

When Ethereum mainnet fees rise, simple actions like registering, renewing, or managing names can become annoying or expensive. That limits how broadly ENS can be used, especially for smaller users.

A Layer-2 registry model could help by moving more routine activity onto cheaper infrastructure.

Keeping Ethereum Security In The Picture

The challenge is not just moving to L2.

ENS has to preserve the trust assumptions that made it valuable in the first place. Users want lower fees, but they also want confidence that names remain secure, durable, and connected to Ethereum’s settlement layer.

That is why the proposal matters.

It is trying to find a balance between cheaper user actions and strong security proofs. If that balance works, ENS could become easier to use without losing the trust that comes from being rooted in Ethereum.

Governance Comes First

ENS is governed by a DAO, so major changes need community discussion and approval.

The current proposal is still in the early discussion phase. That means delegates, users, developers, and service providers can debate trade-offs before anything becomes final.

That process may feel slow, but it is important.

Name infrastructure is sensitive. If ENS changes how registration and resolution work, the ecosystem needs time to understand the implications.

Cost Savings Need Careful Wording

The proposal aims to reduce gas costs sharply, but cost-saving claims need to be tied to the final design.

Layer 2s can make transactions much cheaper, but actual savings depend on implementation, network fees, bridging assumptions, proof systems, and how users interact with the new registry.

So the right view is that ENSv2 could significantly reduce costs if adopted and implemented successfully.

It is not a guarantee today.

The Bigger Ethereum Identity Story

ENS has remained one of Ethereum’s most recognizable non-financial protocols.

It is not just about speculation. It is about identity, payments, wallets, websites, and user experience. If ENS can make names cheaper and easier to manage, it could become more useful across the Ethereum ecosystem.

That is why the L2 migration proposal matters.

It shows ENS trying to adapt to where Ethereum is going: a world where mainnet anchors security, while more user activity happens on Layer 2.

The proposal is early, but the direction makes sense.

This article draws on ENS governance materials relating to the ENSv2 Layer-2 registry migration proposal.

This article was written by the News Desk and edited by Samuel Rae.

This report is based on information released by Discuss. at Discuss

Bitcoin price forecasting has accumulated an unusually colorful collection of methods.

You have basic scarcity models that convert the halving schedule into a price, and run-of-the-mill on-chain models that turn address or transaction activity into value.

The highly contested power-law charts draw an ascending corridor through Bitcoin's history, and machine-learning systems feed market and macroeconomic data into incredibly complex software.

Each of those approaches enters the price-prediction contest against a very shallow, dumbed-down opponent: naive forecasts that use only current market information. A price forecast can use today's price, a return forecast can use zero, and a direction forecast can use a random walk.

Much of the academic literature has struggled to beat it once a model leaves the period in which it was designed.

A May 2026 preprint reviewing Bitcoin prediction research by Carlos Baquero of the University of Porto reached a pretty sobering conclusion: across the peer-reviewed record, no model had demonstrated durable superiority over the appropriate naive benchmark at horizons of one to six months across several market regimes.

The literature contains hundreds of papers, while Baquero selected 23 for close examination based on their methods, influence, or use of genuine out-of-sample evaluation. The review itself is still awaiting peer review, an important distinction when one of its central arguments is that forecasting claims need stronger evaluation.

Short-horizon order flow and daily return forecasts occupy a separate field, and some have produced real predictive value. Online discussions often blend them with longer-horizon price forecasts and valuation models, although each task asks for a different answer.

A formula describing Bitcoin's historical path tells us little about tomorrow's direction, while a daily direction model says little about the price six months from now.

The easiest rival in finance

Naive forecasting works because financial prices are persistent, so a model predicting $100,100 tomorrow when Bitcoin trades at $100,000 today can produce a tiny percentage error even when it has learned almost nothing about direction or return.

Today's price would have been nearly as accurate, and evaluating only the first model gives it credit for information the market had already supplied.

The benchmark becomes more demanding as the horizon expands because Bitcoin can move violently over a month, giving a forecaster room to add value, while the relationships the model learns decay as the market evolves.

A rule calibrated to the retail-led 2017 cycle encountered a different derivatives structure in 2021, and spot ETFs created another route for capital and price discovery in 2024. Each era supplies historical data from a version of the market that no longer exists in quite the same form.

This problem, known as non-stationarity, appears when the relationships between variables don't stay stable enough for past observations to describe the future.

Bitcoin's user base and liquidity have evolved over time, while regulation and access have changed who can trade it and how. A model can capture a relationship during one period and lose it when the market around the asset evolves.

Francesco Puoti, Fabrizio Pittorino, and Manuel Roveri reached a similar result in a study comparing statistical, machine-learning, and deep-learning forecasts. They applied 12 approaches to five major cryptocurrencies at one-day, seven-day, and 30-day horizons.

Simple naive models consistently produced better forecasts than ARIMA, Prophet, random forests, XGBoost, LSTM networks, and N-BEATS.

The result says more about the available information than the sophistication of each method. A complex model can add value when stable patterns exist for it to learn, and it can memorize noise when those patterns are weak or temporary.

Bitcoin offers enormous quantities of data, but the number of independent market cycles it went through is still quite small. Millions of minute bars keep repeating observations from the same 2018 bear market or the same 2020 liquidity shock.

How a backtest becomes a crystal ball for predicting Bitcoin price

Many Bitcoin models look strongest once their creators have seen the entire historical period used to build them. Researchers can try different variables and lookback windows, move the start date, or swap one architecture for another before publishing the best result.

The winner may have discovered a durable relationship, but it also could have won a large lottery conducted on the same price history, an outcome known as backtest overfitting.

David Bailey and his co-authors formalized the problem in their research on the probability of backtest overfitting. Trying more model variations raises the odds of finding an excellent historical result through chance. Selecting the winner and presenting its performance alone hides the number of failed attempts that made the winner possible.

A single chronological split offers little protection because a researcher can train through 2020 and evaluate the model in 2021, producing an apparently out-of-sample result that owes much of its performance to a single bull market.

Walk-forward evaluation is stronger because the model repeatedly retrains on past data and forecasts the next unseen period. Multiple non-overlapping holdout windows are stronger again because they force the same method to encounter bull markets, crashes, sideways periods, and different liquidity conditions.

Among the peer-reviewed papers Baquero examined, none evaluated the same approach across several non-overlapping holdout windows covering different regimes. The strongest papers used rolling or walk-forward evaluation over one continuous out-of-sample period.

Those methods provide real evidence, but a single aggregate error can still hide failure in one section behind success in another.

Information leakage can also lead to false confidence because a feature calculated with future data can give a model a faint view of the answer. You get the same problem when you normalize variables across the full sample, and overlapping return windows can carry future observations across the training boundary.

The error can be subtle enough to survive peer review, especially when a complicated architecture puts several transformations between the raw data and the reported forecast.

The metric itself can flatter the model when a 99% accuracy claim refers to how closely a predicted price level follows the actual price, a relatively easy task for a persistent series.

Traders care about the direction and size of the move, as well as the cost of acting on it. Models that predict $100,500 when Bitcoin moves from $100,000 to $99,500 have a small price error and still make the wrong trade.

The formulas that outlive their forecasts

Bitcoin's best-known valuation frameworks thrive because they turn what's obviously a very complicated asset into a nice, intuitive explanation.

For example, stock-to-flow says scarcity is what drives value, with each halving reducing new supply relative to the existing stock.

Metcalfe-style models say a network becomes more valuable as its user base expands.

The power law says Bitcoin's long history follows a stable mathematical relationship between price and time.

Each of these ideas contains plausible economic intuition, but its forecasting record depends on whether the fitted relationship survives new data and whether simpler explanations account for the same result.

Alexander Shelton's 2024 peer-reviewed examination of Bitcoin return prediction found that stock-to-flow and Metcalfe variables helped explain returns in-sample, but offered limited or zero predictive ability out of sample.

Once time effects entered the stock-to-flow regression, its statistical force disappeared. Bitcoin's supply ratio increases on a predetermined schedule, and its price also climbed for much of its history, making two time-linked series look economically connected.

We saw that weakness in the market long before it appeared in a formal review. The stock-to-flow model diverged from Bitcoin's price as the asset traded below its projected path for years.

Persistent divergence can be absorbed by redefining the output as long-term value or a cycle average, though each redefinition makes the original price claim harder to evaluate.

Metcalfe's Law faces a related identification problem because network activity and price can climb together when adoption raises value, when a higher price attracts users, or when both variables follow a common time path.

Savva Shanaev and his co-authors used instrumental variables across six proof-of-work assets in a study of mining costs, network activity, and crypto value. Once they addressed autocorrelation and the two-way relationship between activity and price, the positive effects attributed to hashrate and transaction count disappeared.

Power-law models are in a much more complicated position because their corridors have captured much of Bitcoin's historical path and provide a practical visual language for discussing where price lies relative to a long-run curve.

Reports on the Bitcoin power-law model have also shown how ETF-era market structure can alter the forces moving price within that corridor.

The academic issue lies in the strength of the inference. A high R-squared on a log-log chart establishes that a line fits the observed sample. Formal support for a power law also requires evidence about the distribution of residuals and comparisons with other time functions.

Researchers would then need to examine sensitivity to the starting date and performance on future observations. Baquero's review found that the current Bitcoin power-law literature had not yet completed that work.

An honest forecasting standard would publish the naive benchmark beside the model and report every market regime separately. Trading costs belong in the results, while public code and data let other researchers reproduce it.

The paper should also disclose how many variations were attempted, since that number determines how surprising the winning backtest really is. Valuation narratives need to be separated from point forecasts, and the reported range should reflect the asset's uncertainty.

Any correction term should allow a value of zero, letting the model conclude that today's price is its best forecast.

That conclusion will always struggle online because it offers no dramatic target and no date to circle. It has one advantage that the forecast bazaar rarely advertises: it tells us exactly how much the model knows beyond the price already visible to everyone.

The post From power laws to AI networks, why complex Bitcoin price models memorize market noise appeared first on CryptoSlate.

Hardware wallet maker Trezor says a breach at logistics provider ShipMonk exposed contact and order data for another approximately 67,000 U.S. customers after years-old records remained in the vendor's systems despite written deletion assurances.

The Sept. 4 update expands an incident Trezor initially said affected 13,689 people. The two disclosed groups imply a total of roughly 80,689, although Trezor has not issued a single combined figure or published underlying data showing whether the groups overlap. Its use of “another” indicates that it considers the new records additional to the original cohort.

The newly disclosed records cover U.S. orders from November 2019 through August 2021 and include names, email addresses, phone numbers, shipping addresses and order numbers. The data can connect an identifiable person and physical location with a hardware-wallet purchase, creating risks beyond a conventional email leak.

Old data outlived a 90-day policy

When Trezor first disclosed the breach on Aug. 13, it counted 11,742 customers with full exposure and 1,947 with partial exposure. Trezor's Aug. 13 account said older order data had already been deleted. An Aug. 14 clarification acknowledged that some partially exposed records included older orders.

The Sept. 4 update reverses that understanding. Trezor said it repeatedly requested and received written assurances that ShipMonk had deleted the data, yet records from 2019 to 2021 remained. Trezor's published delivery-data policy says customer details should be deleted from both its own and its fulfillment partner's systems after 90 days, with exceptions for ongoing order issues. The assurance letters and their dates have not been made public.

BleepingComputer reported that a ShipMonk notification attributed the original unauthorized access to a vulnerability in analytics platform Metabase. Metabase said the August zero-day could create a session tied to an administrator account and allow bulk table downloads. Once the provider incident was reassessed, the retained historical data expanded the number of Trezor customers known to be exposed.

The breach did not reach Trezor's wallet systems. The company said its systems, products and services were not compromised and its devices remained secure. The listed exposed fields were contact and order data, not recovery seeds, private keys or wallet funds.

The risk instead sits around the wallet. Trezor warned that the information could support convincing scam emails, fraudulent calls or letters and potential physical targeting. Its Sept. 4 update did not identify a confirmed downstream attack caused by this dataset, so those outcomes remain risks rather than documented consequences.

Trezor said it emailed every newly affected customer directly and that anyone who did not receive its incident notice was not affected. It urged customers never to share a wallet backup or enter it on a website.

For hardware-wallet owners, the episode shows that protecting keys does not erase the purchase trail created by fulfillment. A deletion policy offers little protection if a vendor's compliance is not verified.

The post Users exposed by Trezor breach grows sixfold after supposedly deleted shipping logs are found appeared first on CryptoSlate.

The Alpine F1 Team Fan Token climbed on September 5, trading near $0.3557 with a 24-hour gain of roughly 3.6%.

The move coincided with a stunning qualifying result at the Italian Grand Prix in Monza, where Pierre Gasly claimed his first career pole position for Alpine.

How Gasly’s Pole Put Alpine Back on the Map

Gasly stunned the Formula 1 field by edging Mercedes’ George Russell by just 0.060 seconds with a lap of 1:21.786. Oscar Piastri finished third for McLaren.

The result marked Alpine’s first pole in years and came at the same circuit where Gasly scored his sole career F1 victory back in 2020. With a market capitalization near $4 million and solid 24-hour trading volume, ALPINE continues to attract attention from motorsport and crypto enthusiasts alike.

Follow us on X to get the latest news as it happens.

Fan tokens give holders access to team-related voting, exclusive rewards, digital collectibles, and enhanced engagement. ALPINE, launched via Binance, is a BEP-20 utility token tied to the BWT Alpine F1 Team.

Why Fan Tokens React to On-Track Results

Positive on-track results often generate short-term interest in such assets. That pattern showed up again in the modest uptick in price and volume following Gasly’s performance.

Social media buzz, meanwhile, centered heavily on the Frenchman’s emotional reaction and Alpine’s unexpected pace. Direct commentary linking the token’s rise specifically to the pole, though, remained limited.

Fan tokens have grown in relevance across sports generally, particularly around major global events. That utility for fan engagement and community decisions was on full display during this summer’s 2026 FIFA World Cup.

As a result, national team and club tokens saw heightened trading activity around high-stakes matches. That pattern reflected real-world performance and sentiment throughout the tournament.

Subscribe to our YouTube channel to watch leaders and journalists provide expert insights.

That model, already established in football and motorsport, positions tokens like ALPINE as bridges between fans and teams beyond race weekends.

Gasly’s Monza pole provided a timely boost, illustrating how unexpected F1 moments can influence sports token markets even if gains remain measured relative to the token’s historical peaks.

The post Alpine F1 Fan Token Rises After Italian GP Drama appeared first on BeInCrypto.

Crypto investor Machi Big Brother has pulled his $1 million bid for Friend.tech and named venture firm Paradigm as the likely obstacle.

Instead, he urged co-founder Racer to relaunch the Web3 social app on Robinhood Chain. The reversal comes nine days after his original offer, sending FRIEND up more than 1,500%.

Machi Big Brother Friend.tech Bid Collapses After 9 Days

Machi Big Brother, whose real name is Jeffrey Huang, opened the bid on Aug. 27. His $1 million buyout offer asked only for the project’s X account and web address.

Friend.tech launched on Base in 2023, allowing users to trade shares with each other. Paradigm led a seed round into the startup that same year.

However, Huang offered no evidence and framed the claim as a guess. He also has a history here. He bought 11 million FRIEND for roughly 5,200 ETH, and the position later shed over $16 million.

The original team gave up control of the smart contracts in September 2024. Therefore, it remains unclear what a buyer would own.

The FRIEND token price now sits near $0.0069, up roughly 5% in the past 24 hours. Its market value of about $659,000 trails the $4.89 million peak from the August rally. Over 90 days, however, the token still holds a 382% gain.

Robinhood Chain Emerges as Racer’s Next Option

Huang closed his post with a pivot. He told Racer to rebuild the app on Robinhood Chain and promised his backing.

Robinhood Chain went live on July 1 as an Arbitrum-based layer-2 network. Since then, it has surpassed Ethereum in volume on decentralized exchanges and absorbed heavy meme-coin flows.

Meanwhile, that mix worries some analysts. Jon Ma of Artemis warned that the meme coin boom risk could undercut Robinhood’s tokenized-stock ambitions.

FRIEND held its 24-hour gain despite the withdrawal. Racer had not responded publicly by Saturday afternoon, and his answer will decide whether Friend.tech returns.

The post Machi Big Brother Sent FRIEND Up 1,500%. Then Ditched $1M Deal appeared first on BeInCrypto.

Solana recorded more x402 transactions than any other network in August, according to an August roundup from the x402 ecosystem, as developers and companies test ways for artificial intelligence agents to make payments for digital services. The network overtook Base in daily x402 transaction activity during the month, while BlockRun said it settled 5.4 million agentic payments on Solana over a seven-day period through PayAI. Solana also said agents initiated 3.3 million USDC transfers using x402 during a single week. The activity is part of a broader effort to enable software to pay for APIs, data, computing, and other services without requiring a person to handle each transaction. The figures, however, measure transaction activity rather than the number of users or the amount of independent commercial demand behind it.

Solana’s x402 Activity Puts Machine Payments In Focus

x402 is a payment protocol built around the HTTP 402 “Payment Required” status code. It allows a service to request payment when an agent makes an API request, after which the agent can authorize a payment and retry the request. Solana’s documentation describes x402 as a way for agents to access paid APIs and other resources without having to manually create accounts or manage API keys.

https://t.co/ulnwaQCf4T

— x402 on Solana (@x402) September 4, 2026

The model is designed particularly for small, pay-per-use transactions. Instead of requiring an agent to maintain an account, subscription, or API key for each service it uses, an x402-enabled endpoint can return a payment request, allowing the agent to pay for a single request. The Solana ecosystem has been one of the networks supporting this activity. BlockRun said it settled 5.4 million agentic payments on Solana in seven days through PayAI, while its x402 endpoints can accept USDC on Solana.

The August roundup also said x402 had passed roughly 200 million transactions across about 150,000 endpoints, with most payments below $0.50. Solana’s own x402 ecosystem page currently reports more than 37 million transactions on Solana and says the network accounts for 70% of monthly x402 volume. The Solana roundup highlights examples such as paid financial research, data access, computing, and other API services, suggesting that the network is being used to test payments made directly by software.

Wallets Bring Agent Payments Closer To Businesses

For agents to spend money independently, payment infrastructure needs to go beyond settlement. Companies also need a way to fund wallets, set limits, and track what an agent has purchased. That is where corporate finance platforms such as Ramp are entering the market. Ramp’s agent tools include functions for businesses to provision x402 wallets, fund them from a Ramp account, and pay x402 requests using a company’s stablecoin balance.

35M+ transactions have settled over x402. Until now, none of them touched a corporate ledger.

As of today, Agents can now make payments via x402 on Ramp with attribution and audit trails built in.

This lets Ramp customers:
1. Provision & fund agent wallets
2. Empower agents to… pic.twitter.com/BrIGHXMhZf

— Teddy Riker (@teddy_riker) August 20, 2026

The move gives companies a way to connect agent spending with existing financial controls. Ramp’s documentation describes wallet provisioning and funding as business-level functions, while x402 payments can be made from the company’s stablecoin balance.

MoonPay is pursuing a consumer-facing version through PayBox. According to MoonPay’s announcement, PayBox connects with Claude and ChatGPT and allows users to approve transactions prepared by an AI agent. MoonPay says the system can be used for activities including token swaps, DeFi interactions, travel and restaurant bookings, and online purchases. Solana Foundation is also promoting Pay.sh, a payment layer for HTTP agents and command-line tools. The foundation launched the service in collaboration with Google Cloud, stating in its announcement that agents could access and pay for APIs from providers, including Google Cloud, on a per-request basis.

Rish, Head of AI Growth, on the zero to one for agentic payments:

“We at Solana Foundation built a product exactly for this. What’s the zero to one for agentic payments? It’s called https://t.co/DLD8XtDx3w.”

“It’s a very simple install. You just brew install Pay, and underneath… pic.twitter.com/qoynnWdefK

— Solana Foundation (@SolanaFndn) August 20, 2026

The approach changes how an agent can interact with online services. Instead of a person signing up for each API, maintaining credentials, and managing separate subscriptions, the agent can discover a service, receive a price quote, and make a payment for the request. The open question is how much of this activity will translate into sustained business use. Agent payments still face practical issues with authorization, fraud, identity verification, refunds, accounting, and determining whether an agent should spend money in the first place.

Researchers studying x402 have also raised questions about how transaction activity should be measured and how much represents independent economic activity. Solana’s reported lead in x402 activity shows that developers are testing blockchain-based payments for software at increasing volumes. It does not yet establish how large the market for autonomous payments will become.

According to the data provided by CryptoQuant, stablecoin flows in the direction of cryptocurrency exchanges showed signals of minimal recovery in August. Binance recorded more than $1 billion in net stablecoin inflows during the month. The development has come at a time when Bitcoin’s price has been steady after gaining approximately 25% in August and moving back above the $75,000 level.

However, the improvement in Binance’s monthly flows comes against a weaker trend since the start of the year. Around $5.1 billion in net stablecoin outflows have been recorded from Binance, while more than $16 billion in stablecoins have left the major exchange reserves. 

Binance Records More Than $1 Billion in Stablecoin Inflows

As per the report, Binance saw a net inflow of more than $1 billion worth of stablecoins in the month of August, which is a significant boost in terms of liquidity going into the exchange. Nevertheless, the net inflow for August is modest compared to the amount of activity taking place on the exchange. Since the start of the year, Binance has recorded around $5.1 billion in net stablecoin outflows. 

The exchange has around 71% of stablecoin flows across all exchanges, making its reserve movements a crucial part of the overall liquidity scenario. At the same time, more than $16 billion in stablecoins have left the reserves of major exchanges since the start of the year. These outflows indicate that the rise recorded by Binance has not yet translated into a clear reversal of the longer-term trend.

Stablecoins play an important role in crypto market liquidity because they can be used by stakeholders and traders to move into digital assets or remain seated for future changes. When stablecoins are held on exchanges, they can potentially provide readily available liquidity for trading. When stablecoins leave exchanges, it can suggest that liquidity is moving away from the immediate trading environment.

Stakeholders may be retracting their assets or decreasing their exposure, contributing to a fall in exchange reserves. The numerous figures, therefore, present a mixed scenario. Binance’s August inflows point to certain improvement, but the larger year-to-date numbers show that liquidity has continued to move away from centralized exchanges.

Bitcoin Rises 25% as Demand Recovery Remains the Focal Point

The plunge in exchange-held stablecoin liquidity has taken place even as the Bitcoin price increased in the month of August. Bitcoin gained approximately 25% during the month and moved back above $75,000. The cryptocurrency is now trying to consolidate around the higher levels after recovering from its prior weakness. The price recovery shows that Bitcoin can perform strongly even while broader liquidity indicators remain under pressure. However, the continued stablecoin outflows could raise questions about whether the retail-driven trend is supported by a sustainable recovery in market demand.

A stronger return of liquidity could provide extra support for Bitcoin and the wider crypto market. If stablecoin inflows towards exchanges continue to rise, it could indicate that investors are willing to move capital into digital assets. On the other hand, if exchange reserves continue to fall, the market could face extra pressure. Lower liquidity could make it more strenuous for Bitcoin to maintain its new gains, particularly if investor interest also remains feeble.

The $75,000 level is therefore a key area for Bitcoin following its August rally. Maintaining the level while demand improves could boost the recovery narrative. However, another fall in market interest would put the recent gains in question.

If liquidity continues to leave exchanges and sustainable demand fails to come back, Bitcoin will enter another corrective phase. The newest data points to a market that has shown some signs of improvement but has yet to establish a broader liquidity recovery. Binance’s more than $1 billion in August stablecoin inflows are significant on a monthly figure, but they remain small, with approximately 5.1 billion dollars in net outflows recorded since the start of the year.

With more than 15 billion dollars of stablecoin also leaving major exchange reserves, the wider liquidity picture remains questionable. Bitcoin’s approximately 25% August performance provides an optimistic sign, but whether the rally can continue might depend on whether demand and liquidity begin to recover quickly.

B.AI, a next-generation AI infrastructure platform, recently set off a developer frenzy by offering free access to top-tier models. Within days, daily token throughput across the platform crossed 1.33 trillion—a historic milestone. 

The record-breaking figure underscores the campaign’s explosive rollout, but it marks only the first step in B.AI’s broader strategic roadmap. Moving beyond traditional compute distribution pipelines, B.AI aims to build the global settlement layer for intelligence: a core infrastructure hub engineered to power cross-node collaboration, orchestration, and value distribution for AI agents across complex business workflows. 

Positioning itself strategically above all models, below all agents, B.AI deeply integrates a diverse range of top-tier models with full-stack components, laying an unshakable, irreplaceable foundation for the mass adoption of autonomous agents and the productivity boom that follows.

Daily Token Throughput Tops 1.33 Trillion: B.AI’s Free Access Rollout Fuels Usage Boom

B.AI’s recent move to open free access to premium AI models has captivated developers and quickly taken over industry conversations. The push for accessible compute has not only fueled a surge in platform activity but also shattered usage records. 

In a matter of days, soaring API demand pushed the platform’s daily token throughput past a staggering 1.33 trillion. Over a 15-day window, cumulative volume reached 8.19 trillion tokens, drawing in more than 220,000 new API users. As of September 3, B.AI’s total user base had officially surpassed 2.3 million. 

That massive adoption traces directly to the platform’s zero-cost model lineup, a strategic rollout built to erase developers’ cost concerns. With every barrier removed, B.AI now offers unlimited free access to six leading frontier models: DeepSeek-V4-Flash, DeepSeek-V4-Flash-Vision-Exp, Tencent Hy3, Xiaomi MiMo-V2.5, GLM-5.3-Flash (Ox Alpha), and Qwen3.8-Flash. 

Notably, on September 3, B.AI rolled out a new pricing structure for DeepSeek-V4-Flash and DeepSeek-V4-Flash-Vision-Exp, introducing tiered discounts. Developers now receive a 50% discount during peak hours, with off-peak rates dropping to just 25% of standard peak pricing. At the same time, the platform has kept zero-cost access in place for GLM-5.3-Flash (Ox Alpha), Qwen3.8-Flash, Tencent Hy3, and Xiaomi MiMo-V2.5. Despite the shift toward commercialization, developer momentum hasn’t wavered, with platform-wide token throughput continuing its steady climb.

This sustained momentum proves the campaign was far more than a short-term compute giveaway—it is a bellwether for the broader evolution of AI infrastructure. Cracking 1.33 trillion daily tokens makes one thing clear: AI applications are moving past basic chatbots. Powered by a high-performance technical stack and flexible service mechanics, B.AI is laying the groundwork for the next frontier—autonomous AI agents operating at scale. 

Powering the “AI Grid”: B.AI Anchors the Global Settlement Layer for the Agent Economy

For B.AI, democratizing compute is only the prelude. Looking further ahead, the platform is committed to building full-stack infrastructure for the agentic era, cementing its position as the global settlement layer for intelligence.

In the agent era, a typical agent task calls for constant switching between models. No single provider can power a complete workflow on its own, so developers are left juggling fragmented API protocols, disjointed billing systems, and conflicting rate limits. 

B.AI’s settlement layer bridges this exact gap. Positioned strategically “above all models, below all agents,” B.AI abstracts models across different providers, capabilities, and cost structures into a unified pool of schedulable resources.

Powered by a dual-tier API structure offering official-route reliability alongside lowest-cost custom channels, developers can choose between guaranteed direct connections and deeply discounted options across a broad lineup of models. Combined with smart routing on the Chat interface, B.AI operates as a full-stack “AI grid,” ensuring every agent request lands on the optimal model to deliver reliable performance at maximum cost-efficiency. 

On the settlement front, this power grid seamlessly bridges both Web2 and Web3 models. For Web2, developers can rely on familiar traditional payment methods to top up with minimal friction. For Web3, B.AI leverages on-chain payment rails to offer global developers decentralized, verifiable, and low-friction payment options.

With dual payment systems running in parallel, B.AI enables developers and agent applications across any infrastructure setup to find their optimal settlement path on the grid, providing single-point integration with borderless global reach. 

Driving Core Productivity: B.AI Reshapes Agent Collaboration

Beyond building a foundation for compute routing and global settlement, B.AI is moving past base infrastructure to power real-world productivity. By enabling seamless agent collaboration across complex workflows, it delivers the missing execution layer for the agent economy. 

At the heart of this execution layer is native Codex integration. Full compatibility with the Responses API means developers can now use a single B.AI key inside Codex to run flagship GPT models and DeepSeek favorites side by side. 

Engineers can now bring these powerhouse models straight into their daily dev stack. From code generation and reasoning to debugging and refactoring, B.AI unifies the entire workflow under one roof—delivering a direct line from model selection to shipped code. 

Beyond coding, to keep agents running reliably in real-world production at scale, B.AI has built a full-stack infrastructure powered by five core components, equipping agents with a fine-tuned operational engine:

• x402 Payment Protocol: Introduces an innovative “pay-before-response” model that executes high-frequency, on-chain micro-settlements in the background during cross-agent API calls and compute orchestration.
• 8004 Identity Protocol: Issues verifiable on-chain credentials for every agent, logging execution history and credit scores to establish a reliable layer of trust for cross-node collaboration.
• Skills Matrix: Plug-and-play, standardized building blocks that interface directly with MCP servers to equip agents with instant, modular tool-calling capabilities.
• BAIclaw and BAIcode: Built-in platform assistants engineered for end-to-end execution. BAIclaw manages all-in-one terminal operations and multi-agent workflows, while BAIcode serves as an advanced developer engine that streamlines the full pipeline from task analysis and architecture design to coding and testing.

From the token surge sparked by zero-cost model access to its positioning as the global settlement layer for intelligence; from seamless Codex integration to full-stack infrastructure powered by x402, 8004, Skills, and native assistants—B.AI delivers far more than a battle-tested technical stack; it unveils a clear blueprint for what lies ahead. B.AI is building not just accessible compute today, but the definitive launchpad for a thriving agent economy. The future is here—and this is only the beginning.

B.AI Team

Singapore

support@b.ai

Axis Robotics has released Axis Sim Dataset V1, one of the largest open-source simulation datasets for Franka arm manipulation, with the full dataset, training code, and benchmarks publicly available. V1 is built from more than 50,000 human-teleoperated simulation trajectories across 207 manipulation tasks and 60,000+ scene variants on a simulated Franka Research 3 arm.

This dataset drew over 160,000 downloads, making it the most downloaded open-source simulation Franka manipulation dataset on Hugging Face. In benchmarks, continual pretraining on V1 lifted π0.5 and beat a volume-matched RoboCasa baseline, with every result open and verifiable.

Axis Robotics is building the ultimate compounding data engine for Physical AI, a vertically integrated system spanning large-scale simulation, egocentric real-world capture, humanoid loco-manipulation, and human-gated DAgger post-training. The company raised $12 million in seed funding led by Hack VC, with participation from Nomad Capital, Pi Network Ventures, 10K Ventures, and angel investors.

A Bet Against “Clean Data Only”

A common assumption in robotics is that demonstrations must be near-optimal to begin with — filter down to expert trajectories, standardize the setup, and discard anything noisy before it is safe to imitate. Axis’s thesis runs the other way: data quality lives at the distribution level, not the single trajectory. When a large and diverse enough crowd produces noisy, suboptimal trajectories and their errors are uncorrelated, the noise averages out and a working policy survives during training.

Axis Sim Dataset V1 puts that thesis to a public test. Its trajectories span pick-and-place, stacking, pouring, articulated-object manipulation, and tool use, all collected through Axis’s browser-based teleoperation platform, Axis Hub, by a distributed crowd rather than a single expert team. The dataset was built with researchers from UC Berkeley, Johns Hopkins, the University of Michigan, and other institutions.

Results That Scale

On LIBERO-Plus, continual pretraining on V1 lifts π0.5 from 83.9% to 88.8% success and outperforms a volume-matched RoboCasa365 baseline by 37.3%. Performance improves consistently as pretraining data scales from 25% to 100% of the dataset, with no saturation in sight, evidence that the gains come from diversity and coverage rather than a one-off bump. The largest improvements appear under camera, sensor-noise, and layout perturbations, the exact axes Axis randomizes during generation.

The team says V2 is already underway, scaling to 1.2 million trajectories across 1,200 tasks, with cross-embodiment generalization and results across multiple VLA models showing that suboptimal simulation data trains robust policies.

The Engine Behind the Dataset

The dataset is one output of a larger, actively compounding data engine. Where a traditional data vendor collects to a fixed spec and stops, Axis uses model performance and failure cases to determine what should be collected next, so every training round informs the next. That engine runs on a hybrid strategy across four data lines, and all four now run at scale:

• Simulation: over 200,000 distributed contributors on Axis Hub, a top-3 dApp on Base, producing 4.7M+ trajectories across 13 embodiments.
• Egocentric: a managed network of 1,000+ full-time, QC-trained collectors capturing first-person activity in real homes and businesses across 14 industries: 200,000+ hours already banked and growing by 4,000+ hours every day, with Vicon-verified hand pose.
• Loco-manipulation: 500+ hours combining mobility and dexterity on real humanoids (Unitree G1, Booster T2) through hardware-agnostic teleoperation.
• Human-gated DAgger post-training: 500+ hours of human-in-the-loop correction targeted at deployment edge cases.

Every task and trajectory is recorded on-chain on Base for provenance, and contributors are rewarded for verified work quality.

From Open Data to Commercial Deployment

Beyond open-sourcing simulation data, Axis works directly with robot embodiment companies to build customized, embodiment-specific data pipelines and model priors.

As Booster Robotics’ first sim-data partner, Axis rebuilt Booster’s real workspace as a task-aligned digital twin, had distributed contributors collect 42,000+ simulation episodes on it, and distilled them into a Booster-specific model prior. With just 30 real-robot demos, that prior reached 87.5% success versus 37.5% for an out-of-the-box π0.5, matching π0.5 using half the real-world demonstrations.

Other partners span embodiment companies (Feagine Robotics), model companies (Manycore Tech, Dexmal) and industrial automation (Lotus Cars, Geely Auto). Axis also supplies on-chain robotics networks: BitRobot on Solana and OpenRoboto on Bittensor.

Redefining Physical AI’s Data Foundation

“The future of Physical AI isn’t a static dataset you download once,” said Chris Feng, founder of Axis Robotics. “It’s an engine that keeps producing the data the model needs next. Scale gets you broad coverage. Diversity keeps the noise unbiased. The closed loop turns every failure into progress. That’s what compounds.”

Axis was founded by researchers from UC Berkeley, CMU, Georgia Tech, and SJTU, alongside serial founders who have scaled consumer platforms to over 30 million users. Its research is advised by Jiachen Li, Assistant Professor at Georgia Tech.

Paper Link: https://arxiv.org/abs/2607.21588

Project Page: https://axisaiorg.github.io/AXIS-V1/

Dataset Link: https://huggingface.co/datasets/axisrobotics/Franka-Dataset

Github Codebase: https://github.com/AxisAIOrg/Axis-V1-Training