Bubblemaps: Attacker exploited a vulnerability to generate $50 million in NES, netting only $60,000 in profit.

Beating AI News: GLM-5.3 Flash has just been officially open-sourced, and Zhipu AI is offering a limited 50% discount on its API services for a two-week period. The discounted rates are $0.075 per million input tokens, $0.25 per million output tokens, and just $0.015 per million cached input tokens. This price is an order of magnitude lower than Zhipu’s own GLM-5.2 model, which charges $1.4 per million input tokens and $4.4 per million output tokens. The discounted GLM-5.3 Flash’s input and output prices are roughly 1/18 of GLM-5.2’s. The 50% discount applies not only to Z.ai’s official API but also syncs to third-party model aggregation platforms. After the two-week promotion ends, prices will revert to $0.15 per million input tokens, $0.5 per million output tokens, and $0.03 per million cached input tokens.

Bloomberg reporter Emily Nicolle wrote that Bitcoin has rallied 23% in a single week recently, marking its largest weekly gain in over three years and ending its sideways trend since summer. US Treasury Secretary Scott Bessent’s proposal to expand the scale of long-term Treasury repurchases sparked market concerns over US debt and dollar depreciation, pushing funds toward alternative assets like Bitcoin. However, after Bitcoin broke above $80,000 on Tuesday, it has stabilized again, and this catalyst alone is insufficient to drive the market back to a genuine bull run. Nicolle notes that Bitcoin’s narrative as a hedge against the U.S. dollar and inflation remains unsustainable. After Trump reintroduced threats of tariffs on China last October, Bitcoin dropped more than 12% in 24 hours, while gold hit a record high in the same period. Year-to-date in 2026, gold has risen over 7% cumulatively, and Bitcoin, even with its recent rebound, is still down nearly 10%. She argues that the simultaneous rise of gold and Bitcoin last week does not prove they share the same safe-haven attributes, as this crypto rally is largely fueled by short covering. Strategy chairman Michael Saylor called on traders to keep buying Bitcoin during the rally, but his company did not add to its holdings accordingly. Separately, the CLARITY Crypto Market Structure Bill remains stalled due to disagreements over ethical provisions and other issues; the Senate is expected to resume deliberations only in mid-September, leaving limited time ahead of the November midterm elections. Bitcoin has yet to establish a stable, compelling value narrative, and for daily payments, users still prefer stablecoins or cash.

According to Bitget data, the US Dollar Index (DXY) rose 0.3%, erasing half of the decline driven by the repo program.

According to BIT (bit.com) market data, NVIDIA’s stock fell 1.26% to $210.4, with a total market cap of $5.09 trillion. NVIDIA will release its second-quarter fiscal 2027 results after today’s U.S. market close. This earnings report and its future performance guidance could serve as a key catalyst for the next phase of the AI sector’s trend.

Beating AI Express News: After Ox Alpha was announced during the day, Zhipu AI officially open-sourced GLM-5.3-Flash in the evening. The model weights are now available on Hugging Face under the MIT license. With a total of 320 billion parameters, only 18 billion are activated per forward pass. It natively supports text, images, and video, making it the first native multimodal model in the GLM-5 series. According to the official, GLM-5.3-Flash outperforms GLM-5.2 overall, with its programming and agent evaluation scores approaching Claude Opus 4.8, while costing only one-tenth of GLM-5.2’s price. It also features a new base model, marking the first time a hybrid sparse and linear attention architecture has been introduced in the GLM main series, and is pre-trained on 30 trillion tokens of multimodal data. Earlier, the community’s small-sample DeepSWE test on Ox Alpha once yielded an 80% score, but this was based on only 10 questions. After expanding the test sample, the score dropped to around 63%, and the tester voluntarily corrected their initial claim. This score still places it in the first tier, though it is not as high as the initial 80% figure. With the model weights now open-sourced, developers can deploy it directly using frameworks like vLLM, SGLang, and KTransformers, eliminating the need to use APIs from anonymous models.

Aave Labs announced that its Aave mobile application will launch early access on August 26, with Ghost Passes being distributed to select users. These passes are invitation credentials, and their holders can gift them to other users, allowing recipients to bypass the waitlist and gain immediate access to the Aave App.