Leitura

Rich Friedman, Chairman of Goldman Sachs Asset Management, released a new episode of 'Goldman Sachs Exchanges: Great Investors' on September 18, 2026. The conversation is hosted by Alison Mass, Chairman of Investment Banking at Goldman Sachs Global Banking & Markets, and was recorded on July 28, 2026.

According to the official episode description, Friedman discusses the current state of private markets and their scale expansion beyond earlier expectations. He explains why he anticipates the payoff from artificial intelligence investments will arrive later than the market generally assumes, and how the discipline used to build the firm's merchant banking business over 45 years shapes his current investment focus.

The chapter list indicates discussions on Wall Street history, early moves in Asia, lessons from past crises, and where opportunities lie now. For professional readers interested in institutional perspectives on AI valuation logic and private market dynamics, this resource offers direct access to views from a top asset management leader.

Dexter Lee, Staff Machine Learning Engineer at Zip, explains the shift in his team's AI development workflow after adopting LangGraph and LangSmith. Previously, Zip relied on custom code for calling LLMs, managing pipelines, and storing results, which resulted in weeks-long feature additions and insufficient trace details.

According to the episode description, integrating LangGraph and LangSmith allowed the team to access node interaction and LLM call traces without additional coding. This change facilitated quicker adoption by other internal teams and enabled the establishment of an initial evaluation setup to prevent regressions.

Morgan Stanley Co-President Dan Simkowitz appeared on CNBC from the Global Healthcare Conference. The discussion covered the outlook for capital markets, M&A activity, AI investment and financing, credit conditions, and the macroeconomic factors currently shaping the market.

This resource is selected for its authoritative voice on broad market mechanics and the specific intersection of AI financing with traditional capital allocation. Readers interested in how major financial institutions view the current environment for deals, credit, and technology investment may find this episode relevant.

OpenAI published a video titled "How OpenAI Puts ChatGPT to Work | DevDay 2026" on its official YouTube channel, documenting scenarios where the company's internal teams apply ChatGPT in their daily operations.

According to the publisher's description, the video highlights how OpenAI employees utilize ChatGPT to manage product launch processes, conduct competitor research, and transform internal knowledge into actionable tools. The content aims to provide viewers with practical ideas they can adapt to their own work.

For professionals interested in understanding how large language models are implemented in enterprise-level daily operations, this video offers a perspective directly from the model developer. Although a full transcript or detailed technical documentation is not currently available, the episode serves as a relevant case study for AI application practices.

In a video clip, LangChain team members Amy and Sean explore the 'human in the loop' design pattern for AI agents. This pattern requires an agent to pause and wait for explicit human approval before proceeding to its next action.

According to the publisher's description, the discussion addresses balancing automation with safety. They state that the best production-grade agent systems do not aim for full autonomy on day one. Instead, they intelligently divide work between humans and agents: letting the agent handle routine tasks and only interrupting the flow for high-stakes operations.

This content is excerpted from a broader LangChain conversation covering common questions about AI agents, ranging from basic concepts to evaluation methods. For developers building or deploying AI agents, this resource offers a specific perspective on when and why to introduce human approval steps.

OpenAI published a video titled "Get Stuff Done with ChatGPT Voice" on its official YouTube channel on September 23, 2026. The resource highlights the latest developments in ChatGPT Voice, specifically focusing on its integration capabilities into actual workflows.

According to the video description, ChatGPT Voice now supports calling external plugins such as email, calendar, and Slack. Additionally, the feature is powered by the GPT-6 series models: Astra, Sol, and Luna. The video also notes that users can utilize these voice functions within ChatGPT Work on web and mobile to create documents, decks, sites, and spreadsheets, or tackle complex tasks in the browser simply by talking.

For readers interested in how AI voice assistants integrate into daily office routines, this video provides specific information on the functional boundaries and integration methods of ChatGPT Voice. It demonstrates a shift from simple conversation to executing multi-step workflows (such as combining calendar and email plugins).

Global lithography leader ASML released its third-quarter financial report on October 14, 2026. The company achieved net sales of €11.7 billion, a gross margin of 57.5%, and net profit of €5.0 billion in the quarter. Based on this performance, ASML raised its full-year 2026 net sales guidance to between €44 billion and €45 billion, representing an expected year-over-year growth of approximately 36%.

Regarding capacity and demand, ASML President and CEO Christophe Fouquet stated that customers are continuously raising their short- and medium-term demand expectations for equipment, primarily driven by their assessment of the scale and duration of artificial intelligence investments. He revealed that the industry expects to remain in a state of undersupply until at least 2028, with some clients already initiating discussions for equipment orders in 2029 or even 2030.

Despite these optimistic outlooks, Fouquet noted that current planned supply capacities for 2027 and 2028 will still be insufficient to fully meet customer needs. Consequently, ASML is studying plans to further increase 2028 production capacity, aiming to exceed the previously announced 30% capacity expansion target. This statement highlights the existing physical bottlenecks and potential measures to address long-term demand.

CFO Roger Dassen added that the full-year 2026 gross margin expectation was raised to approximately 56%, mainly due to increased business volume, product mix optimization, and strong performance in installed base services. He emphasized that while communications with major clients regarding value-based pricing have been completed, the impact on margins will begin to materialize gradually in 2027 and be fully reflected from 2028 onwards.

OpenAI released a video titled "Figma Gave GPT-6 Astra a Moonshot. Here's what happened." on September 21, 2026. The video showcases GPT-6 Astra completing a "moonshot" task: designing the flight-control experience for a new personal aircraft using the Figma MCP Server.

According to the official description, Astra did not just generate an interface but reasoned through system-level aspects, including audience analysis, jobs to be done, pilot-versus-passenger tradeoffs, and critical interface states. It then connected this thinking to a real design system via the Figma MCP Server to create a polished, on-brand prototype.

This case highlights the gap between generating an interface and designing a system, demonstrating the potential of AI models to handle complex design logic when integrated with external tools like MCP servers. For readers focused on AI agent workflows, multimodal design assistance, and practical applications of the MCP protocol, this video offers a concrete scenario reference.

Enterprise AI security and governance startup Harmonic Security announced on October 14, 2026, that it has raised $50 million in an oversubscribed Series B round led by Crosspoint Capital Partners, with participation from existing investors. This brings the company's total funding to $76 million.

Harmonic Security's platform uses custom small language models (SLMs) to inspect content sent by employees to AI tools within milliseconds, identifying risks such as pasted live credentials while allowing normal developer operations to pass. The new capital will be used to expand engineering and go-to-market teams in the United States and the United Kingdom.

Goldman Sachs Research published a video on September 15, 2026, featuring analyst George Cole discussing the fundamentals behind the global bond sell-off. This resource is suitable for professional readers seeking to understand current fixed-income market volatility and macro interest rate expectations.

As a member of the Goldman Sachs Research team, Cole provides an institutional backdrop to the discussion. The title explicitly references "the fundamentals behind the global bond sell-off," indicating content that analyzes the economic or policy logic driving bond market declines. For investors concerned with global capital allocation, interest rate risk, and bond valuation changes, this video serves as an entry point to top-tier investment bank macro views.

The LangChain team released a video titled "What are Deep Agents? Explained fast" on September 16, 2026, aimed at introducing their latest open-source framework for building complex agents, Deep Agents. As a leading provider of AI development tools, LangChain used this short clip to explain how the framework simplifies the development process for complex agents, highlighting its capabilities in context management.

According to the video description, core features of the Deep Agents framework include controlling what context the agent sees and when, with built-in file systems, skills, subagents, and summarization. Additionally, the framework introduces sandboxes for running untrusted code, human-in-the-loop gating for sensitive actions, and authorization checks before an agent acts. These features form a swappable component system that allows the same agent to run seamlessly on a local laptop, in a sandbox, or in the cloud without changing any code.

For technical professionals evaluating or building production-grade AI agents, this video offers a quick overview of LangChain's latest engineering philosophy. It clarifies Deep Agents' design approach to balancing autonomy and security, particularly through modular components that address context window limitations and execution risk controls.

David Wang, Chief Innovation Officer at Cooley LLP, explores how the firm is reimagining IPO preparation processes in a video released on September 21, 2026. Wang notes that traditional IPO preparations place heavy demands on senior executives through information requests, drafting sessions, and reviews.

To address this, Cooley deployed 'GO Public,' a tool built on OpenAI technology. According to the episode description, this system allows Cooley to present substantive work to clients sooner and identify issues earlier. Despite the automation, Wang emphasizes that people remain responsible for reviewing the output and standing behind the answers.