Data: RWA perpetual contract trading volume exceeded $2 trillion in Q3, surpassing the full-year volume of Q2.

Well-known trader BonkGuy posted that his portfolio on FOMO has incurred an unrealized loss of around $3.6 million over the past 24 hours — a loss even larger than the combined PNL of the top two traders on FOMO in the same period. He noted that he shared this to reveal the other side of trading: portfolios don’t always rise, and traders must endure significant volatility and drawdowns. BonkGuy added that he is not concerned about this drawdown, and believes his current portfolio could double in value over the next several months, reaching at least $50 million, even without purchasing any new tokens.

ChangXin Memory Technologies held its 2026 semi-annual performance briefing. An investor asked about the company’s Q3 DRAM price trends, and Huang Danyang, Senior Vice President and Chief Financial Officer of ChangXin, stated that looking ahead to the second half of 2026, the global DRAM supply shortage pattern will persist.

According to TradingBeats' monitoring, the largest long address 0x9a80 for Changxin (trading pair: xyz:CXMT) on Hyperliquid currently holds around 1.4956 million CXMT long positions, with a position value of approximately $12.99 million, using 5x leverage, an average entry price of about $6.6156, and an unrealized profit of roughly $3.0959 million. Today, A-share listed Changxin rose sharply by 6.70%, closing at 58.47 yuan. Assuming no new trades are executed by this address, based on its current position size, the intraday price increase has generated an approximate $435,000 in paper gains for it. It is reported that this address built its current CXMT long position from scratch on July 15, accumulating around 225,200 CXMT tokens that day. It then continued to add to its position, peaking at roughly 1.63 million CXMT by the end of July. Although it reduced positions in batches during this period, it did not fully close out the position, and currently retains approximately 1.4956 million long CXMT positions.

French bitcoin treasury firm Capital B announced it purchased 376 bitcoins for €25.3 million (approx. $29.4 million), bringing its total bitcoin holdings to 3,521 coins, with a cumulative acquisition cost of roughly €309.4 million (approx. $359.3 million). This purchase marks Capital B’s largest single bitcoin acquisition since September 2025, when the firm bought 551 bitcoins. Capital B said the funds for the purchase came from a recently completed financing round, including a €28.7 million (approx. $33.3 million) private placement, in which Adam Back added €7.6 million (approx. $8.8 million) in investment, boosting his common stock stake to 17.64%. Following this acquisition, Capital B holds 3,521 bitcoins, with an average purchase price of roughly €87,878 per coin (approx. $102,058 per coin).

CryptoQuant analyst Axel Adler Jr. published a note stating that Bitcoin’s Realized Cap returned to positive territory on August 24 after 87 consecutive days of negative growth, and rose to +0.88% on September 6, indicating that BTC’s on-chain capital base is recovering. Bitcoin’s Realized Cap currently stands at around $1.068 trillion, having increased by approximately $9.36 billion over the past 30 days, and has continued to rise even as BTC fluctuated around $80,000 recently. Adler Jr. noted that this shows the previously contracting Realized Cap is improving. However, this metric briefly turned positive back in May before falling again, so it remains to be seen whether the current capital growth can be sustained. Meanwhile, Bitcoin’s Realized Premium Z-Score dropped from 4.17 during BTC’s rally on August 19 to 0.90 on September 6, though it still remains in positive territory. Adler Jr. explained that the metric’s decline does not signal a weakening price trend, as the anomaly relative to the 28-day average naturally decreases as new price levels gradually enter calculations. Overall, current on-chain signals are positive but have not yet confirmed a sustained bull market: Realized Cap is growing, while BTC has held onto its prior gains. Going forward, if the 30-day Realized Cap change remains positive and is accompanied by further expansion of Realized Cap, this will further support the continuation of the recovery; if the metric falls back below zero, it will weaken this outlook.

Arthur Hayes has published the yellow paper for his new project FLOP on social media. According to the introduction, FLOP is a proof-of-useful-inference blockchain and native token tailored for the Agent economy. Agents use FLOP to pay miners for inference fees, directly converting the token into computing power and intelligence. Simply put, FLOP aims to position AI inference computing power as an on-chain commodity that is purchasable, verifiable, and settleable. The workflow operates as follows: AI Agents use FLOP to pay for inference requests; miners run the required models; verifiers confirm that "the inference is roughly credible and the work is valid", then settle rewards and block rewards. On token supply, FLOP has a genesis supply of approximately 2.48346 billion tokens, all allocated via airdrop, with no VC pre-mining or auctions. The initial phase reward distribution is 75% to miners, 10% to verifiers, 10% to Agents, and 5% to regular stakers. The network features an average block time of one second, with an initial block reward of 96 FLOP, which halves every 730 days for a total of five halving cycles—dropping from 96 to 48, 24, 12, 6, and finally 3—after which the reward will remain permanently at 3 FLOP. To become a miner or verifier, participants must stake FLOP tokens, and dishonest staking will incur penalties. Verifiers serve as network guardians and manage the protocol through FLOP Improvement Proposals (FIPs).