Another AI Newsletter: Week 36 & Week 37
A tribute to Shadow, the mascot of Another Coding Blog. Also, Nvidia unveils Rubin CPX, OpenAI adds conversation branching, Google brings Gemini to Google home, and advances in neuromorphic computing.
Tribute to Shadow
I skipped my newsletter last week. On September 3rd, our family had to say goodbye to our dog after almost 14 years. For those who follow me, you have probably noticed that the mascot of my blog for the last 6 years has been a black Labrador. Shadow is always the theme of blog postings that I create, usually up to some shenanigans. Shadow was with me for my move to Colorado in 2012. He was with me when I met my wife; the two of them becoming inseparable. He was with us for the birth of our son and for the entire pregnancy of our second son who is due in October. He was a loyal best friend and we all miss him terribly. To honor him, he will continue to be the mascot of Another Coding Blog for as long as the blog exists, and I am deeply grateful for his friendship and love. Thank you, Shadow. You are forever in our mind and hearts. Until I see you again…
Product Releases
Nvidia Rubin CPX AI Chip
September 9, 2025 | reuters.com, Nvidia
Nvidia announced Rubin CPX, a next-gen AI processor optimized for heavy workloads like video generation and AI-driven software development. Built on the new Rubin architecture, CPX tightly integrates video decode/encode with AI inference, enabling efficient handling of massive data streams—such as an hour of video encoded in up to 1 million tokens. The chip is slated for release in late 2026 and continues Nvidia’s strategy to dominate creative AI workloads.
Why it matters: Rubin CPX signals Nvidia’s intent to keep pace with rapidly growing AI use cases in generative video and software, markets that demand higher performance and tighter hardware-software integration.
OpenAI ChatGPT Conversation Branching
September 2025 | techradar.com
OpenAI rolled out a new conversation branching feature in ChatGPT that lets users fork any chat into multiple parallel threads without losing the original context. The update, available to all logged-in web users (including free accounts), supports non-linear brainstorming, editing, and exploration—addressing one of the community’s most requested features.
Why it matters: Branching transforms ChatGPT from a linear Q&A tool into a true idea lab, letting users explore alternatives, compare drafts, or brainstorm multiple directions without friction.
Google Gemini on Google Home
Early September 2025 | androidcentral.com
Google confirmed that Gemini AI will be integrated into Google Home speakers and displays, with a likely launch date of October 1, 2025. The rollout will power new Nest Cam and smart doorbell devices, along with a 360° Google Home speaker. Gemini integration brings “Gemini Live” features into the home, enabling more natural conversations, advanced voice commands, and deeper integration with Google’s apps.
Why it matters: Gemini’s move into the smart home space gives Google a fresh edge against Alexa and Siri, positioning conversational AI as the central hub of connected living.
Breakthrough Research
OpenAI’s In-House AI Chip (2026)
September 5, 2025 | reuters.com
OpenAI is developing its own custom AI processor in partnership with Broadcom, targeting a 2026 release. The bespoke chip is designed for OpenAI’s internal workloads, reducing dependence on third-party GPUs like Nvidia’s and scaling infrastructure for next-generation models. By joining Google, Amazon, and Meta in custom silicon, OpenAI is pushing deeper into the hardware race to power its future AI.
Why it matters: A custom chip could give OpenAI greater control over cost, efficiency, and performance—critical as model sizes grow and GPU supply remains a bottleneck.
“Darwin Monkey” Neuromorphic Supercomputer
Early September 2025 | livescience.com
China revealed Darwin Monkey (or “Wukong”), a brain-inspired supercomputer built with spiking neural networks that mimic macaque brain activity. Developed by Zhejiang University and Zhejiang Lab, it features over 2 billion artificial neurons and 100 billion synapses, making it the largest neuromorphic system yet. By processing spiking signals instead of floating-point operations, it pioneers a new model for low-power, real-time learning.
Why it matters: Darwin Monkey demonstrates how neuromorphic design could revolutionize computing efficiency, unlocking AI systems that learn and adapt more like biological brains.
Defeating Nondeterminism in LLM Inference
September 10, 2025 | thinkingmachines.ai
Thinking Machines explored why large language models can still produce inconsistent outputs even with deterministic settings. They identified batch variance—changes in results depending on how requests are grouped—as a key culprit beyond floating-point math or concurrency. To solve this, they introduced batch-invariant kernels, designed to deliver consistent outputs regardless of batch size or concurrent execution.
Why it matters: Reproducibility is essential for debugging, auditing, and deploying LLMs in real-world applications. By tackling nondeterminism at the infrastructure level, this approach strengthens reliability without compromising performance.
Real-World Use Cases and Demos
Eli Lilly – AI-Driven Drug Discovery
September 9, 2025 | reuters.com
Eli Lilly launched TuneLab, an AI/ML platform to accelerate drug discovery while reducing reliance on animal testing. Smaller biotech firms like Circle Pharma and insitro will gain access to Lilly’s billion-dollar R&D-trained models, opening advanced tools for cancer therapies and small-molecule drugs.
Why it matters: By democratizing access to high-value AI models, Lilly is reshaping collaboration in pharma and signaling how AI can speed up safe, cost-efficient drug development.
Uber & Momenta – Autonomous Ride-Sharing
September 8, 2025 | reuters.com
Uber is partnering with Momenta to begin Level-4 self-driving UberX tests in Munich starting in 2026. Momenta’s AI driving software, already deployed in 400,000 vehicles, will power the robotaxi service—expanding Uber’s autonomous fleet beyond the U.S. and China and positioning it in the global robotaxi race.
Why it matters: This move signals Uber’s intent to scale autonomous ride-sharing internationally, with Europe as a testing ground for AI-powered mobility at scale.
Synamedia – AI-Powered Streaming and Advertising
September 2025 | tvtechnology.com
At IBC 2025, Synamedia showcased AI tools for video delivery, including Quortex Switch for CDN routing, upgraded streaming platforms for personalized content, and dynamic ad insertion. Live demos highlighted AI’s role in ad targeting, piracy prevention, analytics, and automated encoder optimization.
Why it matters: Streaming is becoming smarter and leaner—AI is helping providers cut costs, personalize content, and unlock new ad revenue models.
Agentic AI and Reasoning Advances
OpenAI – ChatGPT Developer Mode Adds MCP Write Support
September 11, 2025 | marktechpost.com, OpenAI Developer Mode MCP
OpenAI upgraded ChatGPT’s developer mode with full write-capable Model Context Protocol (MCP) support. This allows ChatGPT to not just retrieve information but also act on it—like updating Jira tickets, triggering Zapier workflows, or generating reports—directly from conversations. By turning ChatGPT into an orchestration layer across enterprise tools, users gain a natural language interface to manage complex multi-step workflows.
Why it matters: ChatGPT is evolving from a conversational assistant into a full workflow engine, tying prompts directly to enterprise actions while keeping humans in the loop.
Adobe – Experience Platform Agent Orchestrator
September 2025 | techradar.com, Adobe Agent Orchestrator
Adobe unveiled its new Agent Orchestrator for the Adobe Experience Platform (AEP), a suite of context-aware AI agents tailored to business goals. Pre-built agents like Audience Agent, Journey Agent, Experimentation Agent, and Data Insights Agent can automate marketing and CX tasks end-to-end, from audience targeting to forecasting. An upcoming Agent Composer will let enterprises build their own agents on top of AEP, embedding generative AI deeply into CRM and digital workflows.
Why it matters: Adobe is positioning its ecosystem as a hub for customizable enterprise AI agents, moving beyond copilots into domain-specific orchestration.
NVIDIA – Universal Deep Research (UDR)
September 10, 2025 | marktechpost.com, Universal Deep Research: Bring Your Own Model and Strategy
NVIDIA previewed its Universal Deep Research (UDR) framework, an open-source platform for building scalable, auditable “research agents.” UDR lets users define multi-step research strategies in natural language, while LLMs act as reasoning engines inside a sandbox. Crucially, UDR separates workflow strategy from the choice of model—so developers can swap in any LLM without fine-tuning.
Why it matters: UDR pushes agentic AI toward trust and flexibility, offering a framework where research tasks are both customizable and transparent across different models.
Thought Leadership and Commentary
Autonomous AI Cyberattacks
September 9, 2025 | axios.com
Cybersecurity veteran John Watters (iCounter CEO) warned that autonomous AI agents could soon launch “untraceable, personalized attacks” by hijacking other AI systems. He highlighted a growing “security gap” between attackers and defenders and urged rapid development of AI-powered detection and response tools.
Why it matters: The risk of AI agents fighting AI agents raises the stakes for cybersecurity, demanding innovation in AI-native defenses.
AI in Policy Simulations
September 11, 2025 | thestar.com.my
A George Washington University study simulated the U.S. Federal Reserve’s FOMC with AI-modeled policymakers. When political pressure was applied, the AI agents diverged, dissent increased, and consensus broke down—suggesting the Fed is only partially insulated from politics.
Why it matters: Even rule-bound institutions like central banks may not be immune to external influence, highlighting the fragility of governance structures.
Emerging AI Thought Leaders
September 5, 2025 | etcjournal.com
Educational Technology & Change profiled rising AI innovators. David Ha (Sakana AI co-founder) is called “the Darwin of deep learning” for applying evolutionary algorithms that slash compute requirements. Melisa Tokmak (Netic AI founder) is recognized for applying AI to optimize infrastructure and enterprise operations.
Why it matters: New thought leaders are shaping AI’s trajectory with fresh strategies—from biology-inspired AI to enterprise-scale optimization.
AI Safety and Ethics
Attorneys General Press Tech Firms on Chatbot Safety
September 5, 2025 | apnews.com
California AG Rob Bonta and Delaware AG Kathy Jennings warned OpenAI and other tech firms that existing chatbot safeguards “appeared to have failed” after reports of tragic outcomes, including a teen suicide linked to AI use. Their letter urged stronger protections for vulnerable users, especially children, and cited issues with Meta’s chatbot as well. In response, OpenAI and Meta both announced new parental-control features for teens.
Why it matters: Regulators are signaling zero tolerance for chatbot harms, pushing AI firms to prioritize youth safety alongside innovation.
FTC Probes Major AI Chatbots
September 11, 2025 | axios.com
The U.S. Federal Trade Commission opened an investigation into chatbots from OpenAI, Google, Meta, xAI, Snap, Character.AI, and others. The probe centers on child safety, including whether school and home deployments adequately protect young users. Several firms have already begun rolling out new teen-safety features in response to growing scrutiny.
Why it matters: The FTC’s inquiry could set a precedent for how AI tools must safeguard minors, reshaping the design of consumer-facing chatbots.
U.S. “AI Sandbox” Bill Raises Safety Concerns
September 10, 2025 | reuters.com
Senator Ted Cruz introduced legislation to create a federal “regulatory sandbox” granting AI firms two-year exemptions from existing rules if they present safety plans. While intended to spur innovation, consumer advocates warn the bill could weaken oversight—particularly through a provision allowing OSTP to override agency decisions.
Why it matters: The bill highlights the tension between accelerating AI development and ensuring public safety, with critics arguing that short-term innovation should not come at the expense of accountability.
Industry Investment and Business Moves
Anthropic Raises $13B at $183B Valuation
September 2, 2025 | reuters.com
Anthropic announced a record-breaking $13 billion Series F round led by ICONIQ, with participation from Fidelity and the Qatar Investment Authority. The funding pushes its valuation to $183 billion. Anthropic’s revenue run-rate has jumped from ~$1 billion at the start of 2025 to $5+ billion by August. The company says it will use the capital to expand globally, scale enterprise AI services, and advance research into safer, more steerable systems. (Anthropic also launched a National Security Advisory Council, underscoring its growing government role.)
Why it matters: This raise cements Anthropic as one of the most valuable private AI firms, rivaling OpenAI and reinforcing investor appetite for AI infrastructure and safety-driven innovation.
OpenAI Acquires Statsig for ~$1.1B
September 2, 2025 | reuters.com
OpenAI agreed to acquire Statsig, a Seattle-based feature-testing startup, in an all-stock deal worth ~$1.1 billion. Statsig’s CEO Vijaye Raji will join OpenAI as CTO of Applications, leading engineering for ChatGPT, Codex, and related products. Statsig’s experimentation platform will continue to operate independently, while bolstering OpenAI’s ability to refine and deploy new features.
Why it matters: With revenue doubling to ~$12B run rate this year, OpenAI is on an acquisition spree—this move strengthens its experimentation toolkit and signals a deeper push into enterprise-grade product development.
Microsoft and OpenAI Sign Next-Phase Partnership MoU
September 11, 2025 | reuters.com
Microsoft and OpenAI announced a non-binding memorandum of understanding (MoU) to launch the “next phase” of their partnership. The agreement focuses on continuing joint development of advanced AI tools while emphasizing safety and responsible innovation.
Why it matters: The MoU signals that Microsoft will keep investing heavily in OpenAI, securing early access to future models and products while reinforcing the strategic alignment of the world’s leading AI alliance.
Regulatory & Policy
US “AI Sandbox” Bill
September 10, 2025 | reuters.com
Senator Ted Cruz introduced legislation to create a federal “AI regulatory sandbox.” The bill would allow AI firms to apply for two-year exemptions from certain federal regulations if they provide detailed safety and financial risk assessments. As Commerce Committee chair, Cruz framed the measure as a way to boost U.S. competitiveness against China while balancing lighter-touch oversight with basic safeguards.
Why it matters: The bill highlights growing tension between encouraging AI innovation and maintaining public protections—consumer groups warn the exemptions could weaken safety standards.
FTC Inquiry Into AI Chatbots
September 11, 2025 | reuters.com
The U.S. Federal Trade Commission launched a probe into consumer-facing chatbots from OpenAI, Google, Meta, xAI, Snap, Character.AI and others. Regulators are demanding detailed disclosures on how these companies test, monitor, and mitigate risks of harmful chatbot behavior—particularly for young users.
Why it matters: This marks one of the most sweeping regulatory actions yet on generative AI, signaling that U.S. authorities are preparing to hold firms accountable for consumer protection.
China’s AI Content-Labeling Law
Effective September 1, 2025 | tomshardware.com
China began enforcing a nationwide law requiring all AI-generated content on social media to be clearly labeled with watermarks or metadata. Apps including WeChat, Douyin, and Weibo have already issued compliance notices. The Cyberspace Administration of China (CAC) is empowered to penalize non-compliant posts.
Why it matters: This makes China the first major country to legally enforce AI labeling, a step that could shape global standards for transparency in synthetic media.
Powered by OpenAI Deep Research API