Meta enhances Muse's capabilities: Key memory accuracy hits 90%, and the completion rate of stalled tasks rises by 67%

Beating AI Express: Anthropic Publishes Research on Agent Long-Term Memory, Details Claude’s "Dreaming" Mechanism After agents complete ongoing tasks, they accumulate large volumes of conversations, operational logs, and notes. However, this data can become outdated, contradictory, or hard to retrieve when needed. Anthropic divides memory into four layers: raw records, curated notes, an interconnected Wiki knowledge base, and a concise index that loads automatically each time the model starts. Claude can jot down key information during task execution. Its "dreaming" mechanism reviews historical records between tasks, correcting errors, deleting outdated content, and distilling new insights from multiple task runs. On subsequent launches, Claude only needs to load the concise index and retrieve relevant memories as needed, rather than re-reading all historical data. Anthropic conducted experiments using approximately 350,000 internal documents from Stainless, totaling 420 million tokens. Researchers had Opus 5 complete 128 rounds of sessions, during which it continuously read documents, built the Wiki, merged duplicate content, and verified facts. Across 25 questions about these documents, the model scored 53% when using only a search tool, rising to 61% after integrating the evolving Wiki. It also saved an average of roughly 20,000 tokens per question. The Wiki’s advantage was particularly pronounced for questions requiring synthesis of multiple contracts; for answers contained in a single document, standard search sufficed. Anthropic also tested Claude on *Pokémon Red*. The older Sonnet 3.7 model logged 413 memory files but earned only one badge. The new Fable 5 model retained just 10 files yet secured four badges, demonstrating the new model’s improved ability to distinguish between strategies worth preserving and game states that can be re-accessed as needed.

Beating AI News: Meta launches Spaces feature for its personal AI agent Muse. Users only need to describe their requirements in natural language, and Muse will directly generate an interactive mini-app embedded in its Muse interface. Mark Zuckerberg personally showcased three demos: a 3D barbecue game, an activity tracker, and a smart home control panel. Regular users can also build travel planners, financial dashboards, shopping lists, and even small games with data logging capabilities. The generated mini-apps can be used long-term, and integrate with other software to auto-update information. For example, users can ask Muse to create a travel dashboard that aggregates flights, hotels, and itineraries. Once connected to services like Google Calendar, relevant data syncs continuously. Users can check tasks, add records directly in the dashboard, and request the AI to modify interfaces and functions as needed. These mini-apps can be pinned to Muse’s sidebar or main tab, and can also switch automatically based on schedules: task dashboards on weekdays, travel plans on weekends.

The Hong Kong Securities and Futures Commission (SFC) released a consultation conclusion on November 6, proposing to further restrict unregulated entities from using names that may mislead the public into believing they are regulated, and plans to extend such restrictions to the virtual asset sector. Under the proposed amendments, Hong Kong will introduce similar rules under the Anti-Money Laundering and Counter-Terrorist Financing Ordinance, targeting virtual asset trading platforms as well as virtual assets unrelated to securities or futures contracts that can be traded or settled via such platforms. Unauthorized institutions will face restrictions on using such titles to prevent the public from mistaking them for regulated entities. The proposals have won widespread market support. The SFC said it will further finalize the draft amendments to the Securities and Futures Ordinance and the Anti-Money Laundering Ordinance, and submit them to the Legislative Council for review and approval.

Dongcha Beating AI News Flash: Anthropic has updated the cloud development capabilities of Claude Code. Now, when users instruct Claude to run a GitHub project, the AI automatically checks for missing dependencies, required websites to access, and APIs to call, then configures the runtime environment for future reuse. For example, when tasked with running a weather app, it identifies missing Python dependencies, network permissions for the weather API, and the required API key, then sets up installation scripts and network access rules before requesting the key from the user. These steps previously required manual work from developers. Key management has also been isolated: API keys provided by users are stored outside the sandbox, and when Claude sends requests, a proxy server handles adding the key. This allows the model to call external services without accessing the plaintext key, and eliminates the need to hardcode keys into code or environment variables. The configured environment can be reused in subsequent cloud sessions. If the initialization script completes within approximately 5 minutes, the system will also cache installed tools and dependencies, so users won’t need to re-download and install these components when starting a new cloud session.

Beating AI News Flash: Hermes Agent developer Nous Research has released Arete-27B, a model built directly on Alibaba’s Qwen3.8-27B as its base, specifically trained on “Western values”. Nous acknowledges that Chinese open-source models are currently highly capable, but their value orientation does not align with the firm’s standards. The team then fine-tuned the model with approximately 53 million tokens, focusing on topics including freedom of speech, democracy, and privacy. To verify the effectiveness of the modifications, Nous developed 803 political values test questions and enlisted xAI’s Grok 4.7 as an evaluator. Per its own assessment, the model’s “propaganda rate” fell from 22% to roughly 1%, while its general capabilities showed no noticeable decline.

Despite market panic triggered by stricter data center regulations after U.S. Democrats took control of Congress, Goldman Sachs has revised upward its forecast for U.S. data center construction spending, lifting the 2026 projection from $67 billion to $80 billion, the 2027 figure from $87 billion to $100 billion, and for the first time projecting 2028 spending will hit $125 billion. Goldman Sachs noted that around 86 GW of data center projects are already under construction in the U.S., with an additional roughly $900 billion in high-confidence planned projects, adding the sector’s overall construction cycle remains resilient. Demand and backlogs from hyperscale cloud providers including Microsoft, Amazon and Google have not shown significant cooling; H100 compute rental prices rose 26% year-over-year, while GB200 rental prices have climbed 50% year-to-date. Regarding the data center moratorium proposed by Texas Governor-elect Gina Hinojosa, Goldman Sachs believes its impact will mostly delay project deliveries rather than outright cancel construction plans, though it may lead to higher financing costs, reduced tax incentives and some client contract cancellations. Goldman Sachs also warned that a distinction should be made between the industry’s overall growth outlook and the viability of individual developers. Highly leveraged data center developers may face severe financial pressure due to project delays and rising financing costs, with some firms potentially unable to survive until their projects come online.