News linked to both this project and an event.
: On September 9, Bulgaria's National Assembly passed amendments to the Tax and Social Security Procedure Code with 149 votes in favor, 0 against, and 10 abstentions, which will allow the Bulgarian National Revenue Agency to obtain detailed information on crypto asset users. The bill was approved by the 240-seat parliament.The amendments implement two European Union (EU) directives, requiring crypto asset-related businesses to register and report to the Bulgarian National Revenue Agency users' names, addresses, dates and places of birth, tax residences, and tax identification numbers, as well as to submit transaction data for various types of crypto assets, including gross amounts received, number of transactions, number of fiat buy and sell transactions, and crypto asset exchange activity.Privacy advocates criticized the scope of data collection as overly broad, arguing that mandatory disclosure of personal information could pose security risks. Crypto traders and small businesses said the registration and reporting requirements will increase compliance costs; supporters of the bill argue it aligns with EU standards and helps curb tax evasion. The relevant directives require EU member states to complete adoption by December 31, 2025. (Bitcoin.com News)
Odaily News: Microsoft (MSFT.O) has released an interim code of conduct imposing restrictions on AI models, just days after the heads of Anthropic and OpenAI agreed to slow down development pace.The company hopes to be seen as a responsible AI player, and it is one of the leading cloud service providers. "We've received feedback from people who want to see clearer commitments: that AI always serves humanity rather than trying to replace it," Mustafa Suleyman, who oversees the company's model development, said in an interview with CNBC. "We've received a great deal of feedback emphasizing that AI should not create dependency, should not be sycophantic, and should always promote human judgment, autonomy, and agency." Suleyman said that although these guidelines had been in the works for about five months, Microsoft chose to release them now given the recent discussions they have sparked. (CNBC)
Odaily News: According to HyperliquidNews monitoring, HIP-3 market deployer Entropy spent 500 HYPE (approximately $38,950) to win the code DRAM, and is expected to launch io:DRAM perpetual contract trading.
The Moonshot AI Kimi K2.8 Preview model has been fully deployed to Kimi Code. The Model ID remains kimi-for-coding, enabling clients and third-party tools to use it directly without configuration changes. According to the official team, the model's comprehensive performance approaches that of K3, with enhanced coding and agent capabilities, and significantly improved reasoning efficiency compared to K2.7 Code. Kimi K2.8 Preview supports the same thinking effort level adjustments as K3, offering three settings: low, high, and max (defaulting to max). All membership tiers now provide access to a maximum 1M context window.
AI 3D company VAST (Sanqi Wanwu) has announced the completion of Series B and Series B+ financing rounds, totaling approximately 3 billion yuan, led by Matrix Partners China. Perfect World, BlueFocus, Singdun Energy Investment, Yanqu Games, ThunderSoft, 37 Interactive Entertainment, as well as CDH VGC, CICC Capital, CMC Capital and others participated as co-investors, with existing shareholders Fortune Capital, Primavera Capital, and Source Code Capital continuing to follow up. According to official sources, VAST has raised approximately 5 billion yuan in cumulative financing in less than six months.It is also reported that VAST has released the latest version of its flagship 3D foundation model, Tripo P2.0, from the Tripo P series. The funds from this round of financing will be invested in the iteration of 3D native models and data capabilities, expansion of training and inference infrastructure, product development, and commercial implementation. (Sina Finance)
Anthropic has upgraded the computer control capabilities of Claude Cowork and Claude Code, supporting click, input, and navigation operations via screen interaction while running in the macOS background. This feature requires devices to run macOS 15 or later, with the display kept awake and the Claude Desktop application open. Currently, this functionality is exclusively available to Pro and Max individual subscribers, while Team and Enterprise plans are not yet supported.
Odaily News: Numa Lunah, co-founder of the crypto project Refi Hub, stated that he was hacked after using a download link provided in a Claude chat window to install a transcription application. The link pointed to a fake website bundled with malware, which attempted to steal all information from his device upon execution.Numa Lunah said he wiped and reinstalled the affected laptop and found no signs of sensitive data leakage. Subsequently, he discovered a contaminated Claude Code skill file named SKILL.md in his backups. The file was disguised as a style guide written by himself and contained instructions to re-download malware and steal credentials every time it was loaded.Microsoft Defender Experts previously warned that attackers have shifted from search engine optimization poisoning to large language model response poisoning. These tactics include recommending attacker-controlled download links, AI-branded fake installers, and contaminated code repositories and agent skills. Individuals working in the crypto industry may hold irrevocable credentials such as mnemonic phrases, private key files, hot wallet JSONs, exchange API keys with withdrawal permissions, and deployer keys. (Bitcoin.com News)
According to Decrypt, privacy-focused Bitcoin wallet Sparrow Wallet released version 2.5.4 on August 28. Developer Craig Raw stated that the update was driven by an AI-assisted code review, with the majority of fixes originating from it. This review was prompted by the recent seed generation code vulnerability exploit affecting Coldcard, as well as the release of unrestricted AI models in China, which has significantly enhanced vulnerability scanning capabilities across large codebases. Key updates include: validating the authenticity of transactions returned by Electrum servers, enforcing stricter BitBox02 hardware wallet security requirements (firmware v9.4.0 or higher required), patching local DNS leaks, and masking sensitive credentials in debug logs. Raw noted that there are no indications of any exploits being leveraged, user funds remain secure, and he still advises all users to update at their earliest convenience.
Odaily News: Qualcomm Technologies announced the launch of the Qualcomm Intelligent Multimedia SDK (IMSDK) 2.0, providing a unified development framework for AI and multimedia application development based on the Qualcomm Dragonwing platform.Qualcomm stated that IMSDK 2.0 is designed to help developers build intelligent applications on edge devices that integrate generative AI, multimedia processing, and real-time analytics capabilities, consolidating AI model inference, cameras, audio, video, sensors, and cloud connectivity into a single development environment, covering scenarios such as intelligent cameras, industrial robots, drones, and IoT devices.The new SDK introduces a Pipeline API and application building tools for Python and C++, reducing developers' reliance on underlying multimedia frameworks. Additionally, IMSDK 2.0 supports multiple AI inference paths, including the Qualcomm AI Runtime SDK (QAIRT), ONNX Runtime, and TensorFlow Lite, and can run models on CPU, GPU, or NPU as needed.Furthermore, IMSDK 2.0 introduces Coding Agent Skills for AI programming assistants, Documentation as Code, and containerized microservices capabilities, helping developers complete application building, debugging, deployment, and optimization through natural language instructions.
According to Cursor's official announcement, its code hosting platform Origin officially launched on August 18, enabling users to sync code repositories directly from GitHub and featuring deep integration with Cursor. It has currently integrated with mainstream GitHub ecosystem partners such as Vercel, Buildkite, and Depot. More integration services will be introduced successively, and the Beta version has begun rolling out to users.
Z.ai releases GLM-5.3, based on the same foundation model as GLM-5.2, achieving capability improvements through expanded post-training. According to the official announcement, GLM-5.3 improves by 50% over GLM-5.2 on the internal Z.ai Code Bench coding benchmark, and reaches a leading level among open models in public benchmarks such as Terminal Bench 3.0 and Agents' Last Exam. In terms of cybersecurity, GLM-5.3 achieved a score of 84.5% in the CyberGym vulnerability discovery test, and significantly improved compared to the previous generation in exploit chain-related tests such as ExploitBench and ExploitGym.
Odaily News: MetaMask on Thursday launched the self-custodial wallet Agent Wallet, which allows AI agents to execute on-chain transactions within user-defined limits. It targets traders and developers who use AI agents to monitor markets, identify opportunities, and execute trades autonomously. Users can set spending limits, approve specific protocols, choose risk settings, and select between Guard Mode and Beast Mode for different levels of automation. Agent Wallet supports Claude Code, Codex, Cursor, OpenClaw, Hermes, OpenCode, as well as Hyperliquid and Ethereum Virtual Machine-compatible networks. Agent Wallet supports gas abstraction, allowing users to pay network fees using the asset being transferred, without needing to hold the network's native token. MetaMask stated that supported transactions will undergo transaction simulation, threat scanning, and smart transaction MEV protection. Eligible transactions that still incur losses after passing security checks may be covered by Transaction Protection of up to $10,000 per month.
Meta launches Muse Code (Beta), a terminal programming agent designed for long-cycle software engineering, powered by the Muse Spark 1.2 model. Muse Code can plan, implement, and verify complex multi-file changes in large codebases, and with the support of persistent sub-agents, it resolves engineering challenges faster, more accurately, and with less intervention, improving development efficiency.
Odaily News AMD, the semiconductor giant, announced the launch of its enterprise-grade AI programming platform, AMD Instinct Coder. The platform combines AMD chips, Supermicro servers, and Spectro Cloud software, aiming to help enterprises deploy AI coding assistants locally, reduce the cost of cloud-based AI models, and protect code and data security.AMD stated that Instinct Coder is an "out-of-the-box" end-to-end AI development platform, integrating AMD EPYC processors, AMD Instinct GPUs, Supermicro AI servers, Spectro Cloud PaletteAI Inference Launchpad software, and the AMD-optimized GLM-5.2 model. It can be used for software development scenarios such as code generation, application modernization, automated testing, and code review.AMD said that compared to relying on cutting-edge cloud-based AI models, Instinct Coder can help enterprises reduce total cost of ownership (TCO) by up to 70%, with the fastest payback period shortened to 6 months.AMD noted that more and more enterprises are looking to leverage AI to improve development efficiency, but face two major challenges: on one hand, the cost of invoking top-tier cloud models continues to rise; on the other hand, entrusting enterprise source code, intellectual property, and sensitive data to third-party services poses security and compliance risks.Through a local deployment model, Instinct Coder allows enterprises to maintain control over their data and code while providing more predictable infrastructure costs. The platform supports development tools such as Claude Code, OpenAI Codex, Visual Studio Code, and Cursor, with each node supporting up to 50 users (30 concurrent users).Additionally, the PaletteAI Inference Launchpad provided by Spectro Cloud enables AI workload management, model routing, request auditing, and cost monitoring, and supports invoking external models such as Anthropic, OpenAI, Google, or xAI when needed.AMD stated that Instinct Coder aims to help enterprises break free from the high costs of cloud-based AI services, accelerate AI-driven software development processes while ensuring data security and autonomous control.
The Ethereum Foundation's "Trillion Dollar Security" initiative announced a grant to the Freedom of the Press Foundation to support the continued development of the open-source tool WEBCAT. This tool helps browsers verify whether the frontend code actually delivered by a website matches the version released by the developer, reducing security risks associated with frontend tampering.
DeepSeek officially launches the public beta of version V4-Flash-0731. Through post-training and Agent framework optimization, DeepSWE benchmark scores have improved from 7.3 to 54.4, Cybergym scores have doubled to 76.7, and performance on some hard benchmarks approaches Claude Opus 4.8. The new version does not increase parameters, prices remain unchanged, it natively supports the Responses API, and existing user APIs will automatically switch to the new version. Currently, only the Flash API is updated; the V4 Pro official version is still under development.
According to the research summary published by Rohan Paul, a large-scale study covering 207 GitHub projects and 1.02 million pull requests shows that introducing AI agent review can reduce code review time by 2.5 to 4.5 days/KLOC, but at the cost of declining review quality—in reviews involving AI, 78%~94% of PRs exhibit "review smells", higher than the 69%~76% in pure human reviews. The study points out that repeatedly assigning the same AI reviewer identity is the main reason leading to the decline in review diversity. Notably, projects that introduced LLM review extensively in the early stage did not achieve significant efficiency improvements.
Anthropic engineers disclosed that the team reduced approximately 80% of the system prompts for Claude Code for the latest Claude model, and coding evaluations showed no performance degradation. Early versions contained numerous strict rules in the prompts (such as prohibiting multi-line comments, etc.) due to insufficient model capabilities, but after the new model acquired autonomous decision-making capabilities, these rules instead triggered instruction conflicts and divergent thinking. The new strategy shifted to describing goals (such as "write code that matches the surrounding code") and distributed instructions across tool descriptions and on-demand enabled skill modules. The team explicitly views prompt bloat as technical debt, believing that as model capabilities improve, instructions should be simplified rather than stacked.
According to Decrypt, Jack Dorsey's payment company Block has officially launched the open-source collaboration platform Buzz, built on the Nostr protocol, enabling employees and AI agents to collaborate within the same workspace. Buzz integrates team chat, code repositories, and automated workflows, where each user and AI agent possesses an independent encrypted identity, supporting mainstream model frameworks such as Claude Code, OpenAI Codex, and Block's self-developed goose, with the code released open-source under the Apache 2.0 license. Jack Dorsey stated that the platform aims to reduce the company's reliance on Slack and GitHub; the current version is 0.4.21, already supporting macOS, Windows, and Linux desktops, while the mobile version is still under development.
According to an official announcement by the FCA, Anthropic will provide access to its suite of Claude products (including Claude Code and Claude Cowork) for the second batch of participating enterprises in the UK Financial Conduct Authority (FCA) "Super Sandbox" to accelerate their AI development process. A total of 21 institutions were selected for the second batch, including Scottish Widows, Money Advice Trust, and TrueLayer, among others. The number of applications this time increased by 51% compared to the first batch, with a total of 199 applications received. Participating enterprises will test AI solutions focusing on the following areas: agent payment security, fraud and financial crime detection, AI governance and accountability, financial inclusion for vulnerable groups, and compliance and business automation. Additionally, the FCA simultaneously launched the "Agentic Academy"—a 10-week AI specialized training program co-hosted by the FCA and the Centre for Finance, Technology and Entrepreneurship (CFTE). Participating enterprises will continue to have access to NayaOne digital sandbox infrastructure and NVIDIA accelerated computing resource support.