COGNOSCERE Daily Tech Review — Issue T221 · Friday, August 28, 2026

Or visit Intelligence Overview for deeper analysis.

ACTCISA Orders Federal Agencies to Patch Actively Exploited Oracle WebLogic Flaw by August 27

CISA added CVE-2026-21962 (CVSS 10.0), a critical authentication bypass in Oracle HTTP Server and WebLogic Server Proxy Plug-in, to its Known Exploited Vulnerabilities catalog and imposed its tightest 72-hour federal patching deadline. Research indicates the flaw has been exploited in over 140,000 attack attempts by China-linked threat actors against government networks in more than 100 countries since a public proof-of-concept emerged in January 2026.

CISA · Cybersecurity / vulnerability management · Relevance: 0.9 · Source →

vulnerability, patching, WebLogic, federal mandate, CVSS 10, China APT, KEV, cybersecurity

PREPAREOpenAI's Jalapeño Chip Beats Nvidia GB300 in First Published Benchmarks

OpenAI published first benchmark results for its custom inference chip Jalapeño, co-developed with Broadcom, showing 1.5x–1.9x more throughput per kilowatt and 1.7x–3.6x lower latency than Nvidia's GB200 and GB300 systems. The 700-watt chip, built in nine months with AI assistance, is set to deploy in OpenAI's own data centers later this year, with a second generation already approaching tape-out.

OpenAI · AI hardware / silicon · Relevance: 0.9 · Source →

custom silicon, AI chip, inference, hardware efficiency, AI infrastructure, benchmark

PREPAREGitHub's 7-Hour-47-Minute Outage: Record Traffic Exposed Infrastructure Scaling Gap

GitHub published a post-mortem on its August 17 outage — its longest ever — in which record traffic overwhelmed a Central US infrastructure component that failed to auto-scale, cascading into authentication failures across Actions, Pull Requests, Copilot, and APIs. Recovery was prolonged by a client-side retry loop in the VS Code Copilot extension that amplified traffic 10x; GitHub is now accelerating its Azure migration (currently at 58% of platform load) and enforcing stricter retry budgets.

GitHub · Cloud infrastructure / platform reliability · Relevance: 0.8 · Source →

platform reliability, cloud infrastructure, outage, DevOps, AI traffic, Copilot, post-mortem

WATCHNvidia Revenue Hits $96.2 Billion as AI Demand Keeps Surging

Nvidia reported fiscal Q2 2027 revenue of $96.2 billion, more than doubling year-over-year, with data center revenue reaching $89 billion and earnings of $2.22 per share beating expectations. The company guided to $108 billion in next-quarter revenue, signaling continued acceleration driven by AI infrastructure demand despite excluding any China revenue from its outlook.

AP News · AI infrastructure / semiconductor markets · Relevance: 0.9 · Source →

AI infrastructure, semiconductor, earnings, data center, AI spending, GPU, capital expenditure

WATCHMeta's $17.1 Billion Settlement Over Social Media Addiction Claims

Meta agreed to pay up to $17.1 billion to settle a landmark federal lawsuit brought by 47 states, covering allegations of deliberately addicting children to its platforms and illegally collecting data from minors. As part of the settlement, Meta must enforce a two-hour daily limit, overnight lock-out, and school-hour notification blocks for teen users, with an additional $5.3 billion contingent on similar commitments from TikTok and YouTube.

Ars Technica · Tech regulation / platform governance · Relevance: 0.8 · Source →

regulation, social media, children safety, legal settlement, platform governance, data privacy

ACTNvidia Agrees to Acquire Hugging Face for $12.9 Billion

Nvidia has reportedly agreed to acquire Hugging Face, the open-source AI model hub, for approximately $12.9 billion in a deal that would extend the chipmaker's reach from hardware into the software and model ecosystem. A signed agreement has not yet been finalized and the deal could still fall through, but multiple sources confirm that acquisition talks are in an advanced stage.

TechCrunch · AI ecosystem / M&A · Relevance: 1.0 · Source →

acquisition, open-source AI, AI ecosystem, GPU, model hub, AI infrastructure, consolidation

WATCHAnthropic and Nscale Strike $45 Billion Cloud Computing Deal

Anthropic has signed a six-year, $45 billion deal with UK-based AI infrastructure startup Nscale to rent 460 megawatts of compute capacity at a West Virginia data center expected to come online in late 2027, running Nvidia's next-generation Vera Rubin chips. The deal is one of the largest compute commitments disclosed by any AI lab and is part of Anthropic's broader infrastructure push ahead of a planned IPO.

CNBC · AI infrastructure / cloud compute · Relevance: 0.9 · Source →

cloud infrastructure, AI compute, data center, AI lab, IPO, GPU, neocloud

WATCHAnthropic Previews Model Hardware Standard (MHS) for AI Agents Operating Physical Devices

Anthropic opened a research preview of the Model Hardware Standard (MHS), a shared specification enabling AI agents to safely operate lab and manufacturing instruments — such as microscopes, liquid handlers, and robotic arms — in parallel and around the clock. Developed with HHMI Janelia Research Campus and tested at partners including Genentech, MHS is model-agnostic and will be open-sourced, with early results showing hardware integration time reduced from weeks to hours.

Anthropic · Agentic AI / physical AI · Relevance: 0.9 · Source →

physical AI, agentic AI, robotics, lab automation, manufacturing, AI safety, open standard

WATCHZ.AI Reveals GLM-5.3-Flash as the Anonymous 'Ox Alpha' Model That Topped OpenRouter Charts

Chinese AI lab Z.AI confirmed that Ox Alpha, the anonymous model that briefly claimed OpenRouter's top position with usage doubling the second-place model, is its GLM-5.3-Flash — a 320B-parameter mixture-of-experts model with 18B active parameters approaching Claude Opus 4.8 on coding and agentic benchmarks. Priced at roughly $0.045 per task (10x cheaper than comparable rivals), the entire record week of traffic ran on Chinese-made chips.

Z AI · AI model releases / competitive intelligence · Relevance: 0.8 · Source →

LLM, open weights, China AI, model release, inference cost, AI competition, MoE

PREPARESalesforce and Anthropic Announce Claudeforce: The #1 AI Meets the #1 AI CRM

Salesforce and Anthropic launched Claudeforce, a partnership bringing live CRM data and agentic workflows directly into Claude through a plugin with 37 pre-built sales skills covering pipeline analysis, meeting prep, and deal reviews. Currently in pilot with a September open beta planned, the integration signals a strategic shift toward AI-native enterprise software interfaces that bypass traditional application UIs.

Salesforce · Enterprise AI / CRM analytics · Relevance: 0.9 · Source →

CRM, enterprise AI, agentic AI, SaaS, decision support, sales automation, MCP

WATCHAWS Acquires DuckLabs, Bringing DuckDB Team In-House While Keeping Project Open Source

Amazon has signed a definitive agreement to acquire DuckLabs, the Amsterdam-based company behind the popular open-source in-process analytical database DuckDB. The DuckDB project will remain MIT-licensed and governed by the independent DuckDB Foundation, while the engineering team joins AWS to deepen integration with S3, SageMaker Lakehouse, and the broader AWS analytics stack.

Amazon Web Services · Data infrastructure / analytics · Relevance: 0.9 · Source →

data analytics, open source, cloud data, OLAP, S3, acquisition, data lakehouse