Shield: The KiloEx Attacker Tagged Address has returned approximately 1.4 million USDT to KiloEx

According to Lookonchain's monitoring, prior to this morning's market plunge, four newly created addresses deposited 1 million USDC into Hyperliquid, opening a 40x short position on Bitcoin, with the position valued at $12.5 million.

Beating AI News Flash: OpenAI Product and Platform Lead Thibault Sottiaux announced Day 2 of the company’s 28-day product sprint, rolling out four updates in one go. He later noted that Day 2 could end up bringing either four updates, a quota reset, or both, potentially resulting in four improvements plus one quota reset. For Codex users, the most direct update is Auto-review, which is now free for users logging in via ChatGPT accounts and no longer consumes subscription usage. Auto-review automatically determines whether to allow agents to run operations outside the sandbox, reducing the need for repeated user confirmations. Sottiaux added that this feature previously accounted for 2% to 10% of planned usage. The other three updates are: OpenAI streamlined its API’s five paid tiers into three: Build, Launch, and Grow; ChatGPT launched the Meetings plugin, which automatically records meetings and generates summaries and to-do lists; and the Decisions API is now in open beta, allowing apps to quickly select models, tools, or next steps.

Bitcoin has dropped below $84,000. Just before the market dropped, 4 newly created wallets deposited 1M $USDC into #Hyperliquid and opened 40x shorts on 148.49 $BTC ($12.5M). Insiders?

Beating AI News Flash: Google officially launched Nano Banana 2.1, an upgraded iteration of Nano Banana 2 that remains positioned as a high-efficiency image generation and editing model. Key enhancements span image quality, prompt adherence, text generation, and multi-turn character consistency. Support for the model is now live across Google products including Gemini, AI Mode, AI Studio, Flow, Stitch, and Google Ads, with its API identifier set as gemini-nano-banana-2.1. Image editing is a core focus of this update: Version 2.1 supports 1K, 2K, and 4K resolutions, allows up to 14 reference images at once, and maintains consistency for up to 4 characters and 10 objects. Google has also improved mask editing (modifying only specified areas), plus enhanced infographic layout, image text, and ultra-wide aspect ratio generation. The model now offers three thinking intensity tiers: minimal, medium, and high. In Google’s internal model benchmarks, the thinking-enabled Nano Banana 2.1 outperforms Nano Banana 2 and Nano Banana Pro across all 10 evaluated metrics, including text-to-image, general editing, multi-character consistency, and mask editing. Independent Arena rankings confirm notable gains: 2.1 currently ranks 5th in text-to-image and 6th in single-image editing, the highest among Google’s models, though still trailing GPT Image 2.5 and GPT Image 2. Pricing has shifted overall: Standard API image output costs are $0.0336 (1K), $0.0504 (2K), and $0.0756 (4K), roughly half the price of Nano Banana 2. However, input pricing has risen from $0.50 per million tokens to $1.50, while text and thinking output pricing has increased from $3 to $7.50. The more reference images used and the higher the thinking intensity, the smaller the cost advantage from the lower image pricing.

As Bitcoin experiences a sharp pullback, the crypto market appears to be gripped by panic once more. Over the past half hour, prominent trader and DeGods founder @frankdegods has rapidly offloaded altcoins worth roughly $1 million on-chain, including $1 million worth of BP, $250,000 worth of EDEL, $70,000 worth of GP, and other tokens.

Beating AI Express: Google DeepMind has open-sourced EmbeddingGemma 2. The model has just 740 million parameters, capable of directly processing text, code, images, video, and audio, and embedding all of them into a shared vector space for search and Retrieval-Augmented Generation (RAG). The previous-generation EmbeddingGemma only supported text processing. This new model allows partial loading as needed: 270 million parameters for text and code only, 440 million when adding vision capabilities, and a full 740 million parameters when all components are loaded. Google tested the model on the Pixel 11 Pro; the quantized text-only version uses as little as ~191MB of active memory, while the full version takes around 567MB. Its context window has been expanded from 2K in the prior generation to 8K, supporting processing of up to ~5.5 minutes of audio, 29 images, or 58 video frames in a single pass. The most notable improvement is in code retrieval: its MTEB Code score rose from 68.76 in the previous version to 78.68, while multilingual text performance remained largely stable, increasing slightly from 61.15 to 61.36. The model also supports reducing the default 768-dimensional vectors to 512, 256, or 128 dimensions, cutting vector storage requirements by up to one-sixth. The license has also been relaxed: while the prior EmbeddingGemma used Google’s proprietary Gemma terms, EmbeddingGemma 2 is released under the Apache 2.0 license. Gemma 4, launched by Google earlier this year, also uses Apache 2.0, reflecting a clear trend of recent open models being more business-friendly and accessible for secondary development.