According to BlockBeats, on September 2nd, CryptoQuant analyst Darkfost published an article stating that demand in the Bitcoin futures market is showing signs of weakening. Data shows that the 30-day average net active buying volume has begun to decline after a period of relatively strong growth, a trend similar to that of May 2026. When Bitcoin broke through $65,000 on August 19th, the Taker Buy/Sell Ratio rose to 1.21, but market optimism is waning as investors gradually increase their short positions.

Currently, the 30-day average net active buying volume has dropped from $213.7 billion to $97.8 billion in just a few days, a decrease of over 50%; meanwhile, the Taker Buy/Sell Ratio has remained negative since August 30. This indicator compares the volume of active buy and sell orders in the futures market, reflecting changes in investor sentiment and positions.

Analysis indicates that trading volume in the futures market remains significantly higher than in the spot and ETF markets, thus its capital flows have a strong impact on overall market trends. In the short term, the Bitcoin market is showing signs of deterioration, with weakening futures demand and increased short positions potentially further suppressing price performance.

According to BlockBeats, on September 2nd, based on BIT (bit.com) market data, cryptocurrency stocks generally fell in pre-market trading in the US, including: CRCL down 2.48%, MSTR down 2.08%, MARA down 2.44%, COIN down 1.83%, BMNR down 2.57%, and HOOD down 1.28%.

Beating AI News Flash: Meta Superintelligence Labs has released Muse Voice Transcribe. It integrates real-time speech-to-text conversion, speaker differentiation, and user completion timing into a single model. It can process over one hour of audio and differentiate between more than 20 speakers simultaneously.

In Artificial Analysis's real-time English transcription test, its WER (word error rate, the lower the better) was 3.1%. GPT Live Transcribe had a WER of 3.9%, and Gemini 3.5 Transcribe Live had a WER of 4.0%. Muse Voice Transcribe produced the final text approximately 0.16 seconds after the speaker finished speaking.

It doesn't assign a fixed waiting time to all words. The model reads 80ms of audio each time and decides whether to continue listening or start outputting. Simple words can be written out faster, while difficult words are written out after listening to the context for a longer time. Meta uses reinforcement learning to optimize both accuracy and latency.

The model is trained in over 70 languages, with 25 of them having been validated. It can also directly recognize mixed Chinese and English speech in a sentence. It is currently integrated with Meta AI for Mac and Muse Code, and the Meta Model API is also available. The price is $3 per 1000 minutes, or $0.18 per hour.

According to an official announcement released by BlockBeats on September 2nd, Binance's cross margin, unified account mode, and unified account professional version will add four bStocks tokens—Seagate (STXB), ProShares UltraPro Short QQQ (SQQQB), Moderna (MRNAB), and CrowdStrike (CRWDB)—as eligible collateral assets on September 2nd, 2026 at 20:00 (UTC+8).

The corresponding bStocks trading pairs will also open for leveraged trading. Eligible users will then be able to use these bStocks tokens as margin collateral to expand the collateral options for margin trading.