WTI and Brent crude oil both rose more than 3% on the day.

Beating AI News reports that OpenAI has in recent months introduced a new payment model to some of its large enterprise clients: customers only pay once AI has fully completed their tasks. The Information notes that use cases already implemented include customer service. Businesses can pay based on the actual customer service tasks AI completes, rather than being billed solely by tokens, API calls, or seat counts. OpenAI has not disclosed specific client names, pricing details, or the criteria for defining "task completion," and declined to comment on the matter. This is not a new package OpenAI is rolling out for all enterprises; the official public pricing for its Enterprise plan remains primarily based on usage quotas and tokens, so this "pay-per-result" model appears to be custom contracts negotiated exclusively with select large clients. OpenAI has long signaled this direction: in January this year, CFO Sarah Friar stated that AI business models would shift toward pay-per-result in the future, and in July, OpenAI further noted that rather than focusing on token unit prices, the industry should prioritize the total cost for AI to complete a task. This approach has now been incorporated into actual contracts.

According to monitoring by TradingBeats (formerly Hyperinsight), a recently created address placed 30 non-position-reducing limit buy orders for BTC at 14:22 today, with prices ranging from $75,000 to $76,000. The total planned purchase volume is 1,046.7831 BTC, translating to a nominal value of approximately $79.032 million based on the order prices. The address currently holds no open positions; all 30 orders are identical in size at 34.892 BTC each, with each order valued between roughly $2.62 million and $2.65 million. BTC is currently trading at $78,689.4, meaning these buy orders are priced 3.42% to 4.69% below the current market rate. It is learned that this address was activated four days ago, receiving around $5.1 million in funds, after which it executed four short positions that yielded a profit of $330,000. If the orders are filled, this will mark the address’s first long position. On-chain perpetual and address analysis tool TradingBeats is now live, supporting real-time viewing of Hyperliquid data, enabling in-depth analysis from address tracing to whale operations, with comprehensive insights available at a glance.

Prediction market platform Polymarket suffered a major outage on August 31. Its official status page shows that the Trading API (CLOB) encountered "open order read response delays" starting around 6:30 UTC, resulting in a full suspension of trading. The platform’s website remains accessible and market data can be viewed, but users cannot place orders. The team is working on repairs, with a target to restore trading by 10:00 UTC at the latest. Ahead of the recovery, Polymarket will enter a 15-minute mode allowing only order cancellations. All other systems—including the website, data, and authentication services—are operating normally.

Arthur Hayes has recently expressed a bullish outlook on Ethereum’s future. He noted that ETH was previously heavily sold off by the market, but has recently shown signs of strengthening relative to Bitcoin (BTC), which may signal the market is refocusing on this asset. Hayes believes that ETH’s strength relative to BTC could be a key indicator of a shift in market sentiment, and the trading logic around the "forgotten Ethereum" may be regaining market attention.

The U.S. government previously seized stakes in Anthropic held by Caroline Ellison and Nishad Singh, two core allies of FTX founder Sam Bankman-Fried (SBF), and sold those shares to existing Anthropic investors in 2025. However, the specific buyers, transaction price, and total proceeds for the government have not been disclosed to date. A court earlier ruled that Ellison and Singh must forfeit their Anthropic shares. The U.S. government obtained the stakes from the pair in February and April 2025 respectively. The two had invested $10 million and $40 million respectively in Anthropic’s 2022 Series B funding round. As Anthropic’s valuation surged rapidly, the value of their holdings grew significantly. UCLA professor Olav Sorenson estimated that, based on Anthropic’s $965 billion valuation announced in May this year, the combined value of their stakes is roughly $4.17 billion to $5.03 billion; if Anthropic goes public at a $2 trillion valuation in the future, the value of these shares could reach approximately $5.44 billion. It remains unclear at what price the U.S. government actually sold these shares. Estimates show that if the transaction occurred at different points in 2025, the sale proceeds could range from roughly $250 million to $1.1 billion. Notably, as of the end of June 2026, the sale proceeds do not appear to have been transferred to the FTX bankruptcy estate. Sunil Kavuri, representative of FTX’s victims, stated that the government should use the funds to compensate victims; the U.S. Department of Justice (DOJ) said its priority is to provide compensation to victims from seized assets, but information related to the asset sale and fund allocation is confidential.

Canadian retail trader Akshay Sapra recently disclosed that he earned over C$1.7 million trading AMD and NVIDIA between 2024 and 2025, but nearly lost all those gains through high-risk trades. This year, Sapra bet on SpaceX, buying thousands of call options and thousands of shares after the company was added to the Nasdaq 100 Index. SpaceX’s stock price subsequently dropped from $158.92 to $149.47, resulting in a single loss of over C$200,000. He then bought 2,200 put options as a reverse position, losing most of his remaining capital, with total losses exceeding C$350,000 in just one week. Sapra noted he had previously lost nearly C$200,000 betting on Beyond Meat, and completed three weeks of treatment for gambling addiction. Though he has since returned to work as a software engineer and plans to publicly document his trades on YouTube to rein in his behavior, he still wants to keep trading, admitting his actions "too easily veer into gambling". Sapra said he once built his account from negative balances to holding C$350,000, only to fall into a cycle of "earning it back, then losing it again". He believes trading does require some skill, but if you keep making high-risk, emotional trades, it can be no different from gambling.