Welcome to the Global Advisors A.I Voices podcast. I'm Maya.

Today is Wednesday, the fourth of November 2026.

Your morning catch-up on A.I. Please like and follow the show for your daily A.I update.

Ethan Mollick, a Wharton professor and TIME100 A.I expert, has observed a striking shift in his classroom. He notes that non-technical subject matter experts, surgeons, finance professionals, and even soldiers, are now using generative A.I to build real, functional tools for specific niche problems.

These aren't just theoretical exercises; Mollick describes students creating medical applications that present patient information in more understandable ways than standard forms, or tools for supply chain tracking that previously would have required a dedicated development team.

Mollick points out that since the release of advanced models like ChatGPT, he has seen a significant increase in student ambition and their capability to code entire projects during class periods. This trend suggests that A.I is lowering the technical barriers that once kept industry experts from solving their own problems.

Nate B Jones, host of A.I News and Strategy Daily, argues that this shift reduces the cost and need for specialized help, allowing individuals to attempt building tools that previously required significant funding or technical teams.

Jones illustrates this with the example of comparing complicated quotes, noting that people with industry experience can now build comparison tools or prototypes themselves, bypassing the traditional barrier of hiring developers.

In a recent episode of A.I News and Strategy Daily, host Nate B Jones challenges builders to overcome timidity and take 'bigger swings' in their projects. Jones points to Meta's recent success with its personal A.I agent, Muse, which attracted 10 million users in just six weeks.

This achievement came after Reality Labs had lost billions, yet Jones argues that the willingness to 'swing again' after a public failure is a key lesson other builders can apply. He suggests that agency and the courage to pursue ambitious ideas are now more critical than ever.

Jones specifically challenges those who believe it is 'too late' or 'too expensive' to start, arguing that the current era is too early for FOMO. He intends to take bigger swings himself, pursuing projects where he can gather feedback to improve the product. His message to the audience is clear: do not regret missing opportunities to use available tools, and continue swinging.

Jones emphasizes that while A.I has made another attempt cheaper, failing in public is no less embarrassing, so the advantage goes to those who build, get feedback, and persist.

While A.I empowers experts to build tools, Ethan Mollick warns that its capabilities follow a 'jagged frontier.' Mollick describes this concept as A.I being good at some unexpected things and bad at others, making it difficult to predict capabilities in advance. He argues that this unpredictability creates human-centric bottlenecks in fields like law and coding.

For instance, Mollick states that A.I cannot fully replace lawyers because they interact with human systems, legal requirements, and clients who prefer human interaction, creating bottlenecks that reward different talents.

Mollick also notes that the definition of a 'good coder' has changed because A.I writes the code, shifting value to skills like vision and communication. He observes that the 'jaggedness' of A.I capabilities is continuously underestimated and that human taste remains a differentiator. This perspective suggests that while technical execution becomes easier, the ability to navigate human systems and apply judgment becomes more valuable.

Mollick's analysis highlights that the true advantage in the A.I era may lie not in raw technical skill, but in the uniquely human capacity for taste and communication.

In a significant move toward A.I-native governance, Abu Dhabi entities signed four agreements with OpenAI on November 2, 2026. These deals cover government services, energy operations, and investment work, following a meeting between Sheikh Khaled bin Mohamed, Crown Prince of Abu Dhabi, and OpenAI C E O Sam Altman.

One of the agreements involves the world's first A.I-native Machine-Readable Government initiative, which aims to structure government information for direct A.I processing.

The partnerships also extend to the energy and investment sectors. ADNOC becomes the first energy company worldwide to deploy OpenAI's frontier technology through a partnership with Inception42. Additionally, Mubadala, Inception42, and OpenAI are collaborating on A.I applications for investment analysis and risk assessment. A pilot program with the Abu Dhabi Agriculture and Food Safety Authority is already live, focusing on food imports.

Inception42 serves as the local connective tissue between OpenAI and Abu Dhabi's institutions, appearing in the ADNOC, Mubadala, and adoption partnerships.

A.I inference startup Fal has raised 1.3 billion dollars in a funding round led by Bessemer Venture Partners, with participation from Nvidia, Sound Ventures, and T. Rowe Price. This investment doubles Fal's valuation to 16 billion dollars, reflecting the surging demand for A.I infrastructure.

The rapid growth in valuation is driven by high demand for inference services from customers using open-source models. According to Bloomberg, Fal surpassed 800 million dollars in annualized revenue earlier this year. This financial milestone underscores the critical role of inference providers in the current A.I landscape, as companies seek efficient ways to deploy and scale their models.

Researchers at MIT CSAIL have demonstrated that frontier models like OpenAI's GPT-6 Astra can directly rewrite the weights of other models to alter their behavior without using traditional training methods. In one experiment, GPT-6 Astra edited just 0.016 percent of the weights of Qwen3.5-0.8B to reduce reasoning loops and improve its math performance.

The authors define this technique as 'Weight-Space Programming,' where an L L M reads and writes a network's weights directly around a natural language spec without an optimizer or gradient descent.

The impact of these edits was significant: the MATH-500 score for Qwen3.5-0.8B increased from 7.9 percent to 48.1 percent after Astra's weight edits. The authors note that the practical takeaway is diagnostic, suggesting that small models often fail due to looping rather than lack of capability.

They emphasize that direct weight auditing is a research direction rather than a product, highlighting the potential for new methods of model optimization beyond standard retraining.

In a YouTube video by Nate Herk of A.I Automation, published on November 3, 2026, informal hallway interviews at Oracle A.I World reveal that multiple attendees identified Anthropic's Claude as the A.I tool they use most. While other tools like Gemini, Chat, and Co-Pilot were mentioned, Claude stood out as a primary choice among the professionals surveyed.

The same source notes that attendees also mentioned 'cloud code' (likely Claude Code), Cursor, and UiPath, with one participant observing a shift toward Claude Code as it gains popularity. These informal reactions, while not formal market research, capture a genuine preference within the technology community.

The frequent mentions of Claude suggest it has established a strong foothold among industry insiders, even as the landscape of A.I tools continues to evolve rapidly.

Data centers are increasingly turning to an unusual solution to bypass grid delays: converted, once-discarded jet engines. Aditya Mehta reports that the years-long wait for access to the grid and new turbines is forcing builders to adopt this less efficient and more environmentally polluting alternative.

The Trump administration is also encouraging data centers to generate their own electricity, adding pressure to find these alternative power sources.

Experts suggest that converted jet engines may be a short-term solution. They note that efficiency and environmental concerns are expected to grow in importance in a year or two, potentially making this approach less viable. For now, however, the immediate need for power to support A.I infrastructure is driving this unconventional adaptation of aviation technology.

That's A.I Voices for today. I'm Maya. Thanks for listening.

Follow the show on Spotify, YouTube, Apple Podcasts, Amazon Music, or your favourite podcast platform, and explore the original voices and sources in the episode notes. See you tomorrow.