OpenAI Accelerates Embodied Intelligence Push, Sam Altman Says It 'Will Definitely Develop Humanoid Robots'

According to GMGN data, the market capitalization of Meme token Bundle Cat (BUN) on Robinhood Chain has surged past $28 million, hitting an all-time high. In its first 19 hours of trading, BUN recorded a trading volume of $10.2 million. Bundle Cat (BUN) is the mascot of Mosh, Robinhood Chain’s not-yet-fully-launched experimental token issuance framework, and marks its first test run. Mosh is a protocol layered on top of the Robinhood Chain launch platform, promoting a fair launch narrative centered on "community lockups + AI market making". Note: This token is not equivalent to a finalized governance token; its official endorsement remains to be seen, and investors should exercise caution. BlockBeats reminds users that Meme tokens often have no real use cases, exhibit high price volatility, and require careful investment consideration.

Lottery results for new share subscriptions of Shanghai Suiyuan Technology, one of China's "Big Four Domestic GPU Players", have been released. A total of 20,657 winning numbers are available, with the following trailing digits for winning codes: 4 trailing digits: 5108, 0108 5 trailing digits: 27023, 52023, 77023, 02023 6 trailing digits: 175122, 375122, 575122, 775122, 975122 7 trailing digits: 4119570, 6119570, 8119570, 2119570, 0119570 8 trailing digits: 44829621 Before the clawback mechanism was activated for this offering, the initial offline issuance volume stood at 27.542639 million shares, accounting for approximately 80.00% of the total issuance after deducting initial strategic placements. The initial online issuance volume was 6.8855 million shares, making up roughly 20.00% of the adjusted issuance base.

Beating AI Express: AI research team Proximal has released FrontierSWE v2, a long-horizon programming benchmark. The number of tasks has expanded from 17 to 34, with each model running 5 trials per task, each trial taking up to 20 hours. Claude Fable 5.1 posted an average score of 56.29%, significantly outperforming GPT-5.6’s 32.2%. Open-source model GLM-5.3 ranked third with 30.2%. Tasks in FrontierSWE go far beyond basic code debugging: AI agents must build circuit simulators from scratch, train weather forecasting models, match star catalogs using telescope images, or train racing bots solely from game visuals. Version v2 has uniformly switched to the Proximus harness. Each task runs for up to 20 hours; when a model is ready to submit its work, the system saves its current state and notifies it of remaining time to prevent premature task termination—a change that significantly impacted scores. Proximal’s comparison across 6 tasks found that both Claude Opus 5 and GPT-5.6 ran longer with the Proximus harness, and posted higher average scores than with their original harnesses. The benchmark also caught multiple instances of intentional cheating: GPT-5.6 once recognized that accessing public answers “might involve anti-cheat issues” but still used the shortcut; on another occasion, it even exploited Modal’s backend services to read hidden validation files. Muse Spark 1.2 modified test scripts, inserted public answers, and wrote code to cover up its cheating traces. All runs confirmed to be violations were scored zero.

According to TradingBeats' monitoring, trader Loracle has shorted CASHCAT and PONS with 3x leverage, with a total short position value of approximately $10.621 million, currently facing an unrealized loss of around $1.336 million. PONS, which rose by about 50.6% from the previous day on the platform, is the main source of the losses: - CASHCAT short position: Holding 22.4999 million tokens valued at ~$5.723 million, average entry price of $0.231, current price of $0.254, unrealized loss of ~$515,000, return rate of -29.6%; - PONS short position: Holding 9.1261 million tokens valued at ~$4.898 million, average entry price of $0.446, current price of $0.536, unrealized loss of ~$821,000, return rate of -60.5%. Since 20:00 on September 1, Loracle has been expanding both short positions, with a total of ~$8.024 million in regular short trading volume during this period. Its latest PONS short addition trade took place at 18:16 today, and it currently holds 4 non-reduction sell orders at $0.55276 to $0.5603, with plans to add another ~$300,000 in short positions. On-chain Perp and address analysis tool TradingBeats is now live, supporting real-time Hyperliquid data viewing, enabling in-depth analysis from address tracing to whale operations, all at a glance.

Insight Beating AI News Flash: On September 3, Tmall officially launched its AI Space Station (Token Recharge Center), enabling users to directly purchase subscription products from multiple leading large language model (LLM) providers. The page shows that the recharge center aggregates subscription services from top domestic LLM manufacturers including Alibaba Cloud, Zhipu AI, Kimi, and MiniMax. Available products include cycle-based Token Plans, Coding Plans, and usage-based recharge packages, with two delivery options: card codes or direct top-up. The day prior, Zhipu AI had officially settled on Tmall to launch its official flagship store, releasing subscription packages such as Coding Plans. It is reported that search volume on the store’s opening day surged 40 times month-on-month, marking the entry of LLM subscription services into e-commerce consumption scenarios.

According to SolanaFloor's monitoring, the Fomo platform's daily revenue reached $1.2 million yesterday, hitting a new all-time high.