Cobie aggregated addresses and transferred out $6.58 million worth of LDO, with multiple exchanges receiving large deposits

According to GMGN market data, EMBER, the native token of Solana-based platform Ember Curve, exceeded $37 million in market cap within its first day of launch. The token has rallied over 150% in the past hour, bringing its current market cap to $28 million, with a 24-hour gain of 460% and 24-hour trading volume of $20.4 million. Ember Curve is a new token launch platform built on Meteora for the Solana blockchain. BlockBeats reminds users that token prices are highly volatile, so investment requires caution.

Well-known trader Bonk Guy revealed he purchased EMBER, his first Solana ecosystem token in a long while, today. He spotted the token yesterday when its market capitalization hit $3 million but missed the entry opportunity; after conducting thorough research, he opted to chase the rally, noting that EMBER is one of the token launch platforms with the strongest fundamentals currently, backed by Meteora which boasts robust resources and endorsements. Bonk Guy forecasts EMBER’s market cap will surpass $100 million. On-chain data indicates Bonk Guy’s public wallet address holds EMBER valued at $780,000, with an unrealized profit of 93%. BlockBeats reminds users that token prices are highly volatile, so investment requires caution.

According to GMGN data, the on-chain meme coin flybrain on the Robinhood network has undergone a sharp correction, with its current market capitalization standing at $12 million. Earlier today, it briefly surged past $55 million, and has fallen more than 60% in the past six hours. Prior to this, Marc Andreessen, co-founder of a16z, followed the project's official X account. The project bills itself as the first meme coin developed by researchers using a fruit fly's brain—not "brain-inspired AI", but the actual neural connectome of a real fruit fly. This brain contains 165,122 neurons and 10,228,000 synaptic connections, all sourced from a single male fruit fly, traced individually via electron microscopy. BlockBeats reminds users that most meme coins lack practical use cases, are highly volatile, and require caution when investing.

According to GMGN market data, the market capitalization of STONK, the native token of Solana-based token launch platform StonkFun, has hit a new all-time high, surging past $280 million to stand at $286 million as of press time, with a 56% gain in the past 24 hours. StonkFun is an innovative launchpad on Solana, with a core narrative of "truly bridging memes and RWA (real-world assets)". The platform allows users to launch tokens with one click, and freely select paired assets including xStocks (tokenized equities like SPYx, which tracks the S&P 500), PreStocks (pre-IPO shares), fiat currencies, commodities, other mainstream cryptocurrencies, and even another StonkFun token. STONK acts as the equity voucher for "all leveraged launch platforms with liquidity pools available for pairing with real assets", designed to directly link Solana ecosystem meme culture to RWA liquidity. BlockBeats reminds users that most meme coins lack practical use cases, exhibit high price volatility, and require cautious investment.

Per TradingBeats monitoring, as of press time, SKHX trades at $1,359.6 and SKHY at $190.05 on Hyperliquid, with SKHY carrying a roughly 39.8% premium over SKHX. The spread has widened from 33.9% to around 40% over the past week. During this expansion, six key addresses simultaneously established paired positions: long SKHX and short SKHY. Over the past two days, the six addresses have cumulatively built ~$24.439 million in SKHX longs and ~$24.876 million in SKHY shorts, for a total notional trading volume of ~$49.315 million across both sides. The two largest positions among them are: 0x382d holds ~$8.225 million in SKHX longs and ~$8.407 million in SKHY shorts; 0x2ab3 holds ~$7.865 million in SKHX longs and ~$7.862 million in SKHY shorts. The remaining four addresses each have paired positions worth $1.5 million to $3.2 million. The six addresses entered positions at nearly identical times, with roughly equal notional values on both sides, forming a clear relative value trading structure: buying the relatively cheaper Korean common stock mapping, while selling the higher-premium US-listed ADR. As of now, the spread has not yet converged. The six addresses still hold ~$22.184 million in SKHX longs and ~$22.845 million in SKHY shorts. The SKHY shorts have an unrealized profit of ~$481,000, but the SKHX longs carry an unrealized loss of ~$581,000, resulting in a combined unrealized loss of ~$100,000. Since opening positions, they have paid a net ~$52,400 in funding fees, with ~$21,900 spent in the last 24 hours. However, as of 19:09, SKHX’s dynamic funding rate has turned negative. Based on current positions and rates, the portfolio is estimated to shift to a state of collecting ~$4,890 in net hourly funding fees. This change remains to be confirmed by subsequent hourly settlements; if sustained, it will reduce the cost for the whales waiting for the spread to converge.

Beating AI News (from Dongcha) reports that Nvidia’s NVLabs has open-sourced SoL-Pi, an efficiency optimization layer for the Pi coding agent harness. The research team built an automated workflow where multiple agents analyze Pi’s runtime trajectories, propose optimizations, modify code, and validate changes automatically. Out of 152 initial candidate optimization directions, only 4 were retained, all focused on cutting redundant overhead. These optimizations include: running tests immediately after code edits to eliminate one model call; storing read large files and long outputs locally for on-demand retrieval; deciding whether to compress context after subtask completion; and filtering key points from extremely long logs using cheaper models first, then cross-checking the content against the original logs. On EdgeBench, SoL-Pi uses 45%–49% fewer tokens than the original Pi, reducing costs by roughly a third. Compared to native harnesses like Codex and Claude Code, it cuts token usage by 35%–64% and lowers costs by 50%–54% based on official API pricing. That said, the cost savings come with a minor performance tradeoff: SoL-Pi retains approximately 94% of Pi’s average performance score. For example, in the 63 CPU tasks of Terminal-Bench 4, it solved 15 tasks, while both Codex and Pi solved 18 each.