ai
Eye on A.I. is a biweekly podcast, hosted by longtime New York Times correspondent Craig S. Smith. In each episode, Craig will talk to people making a difference in artificial intelligence. The podcast aims to put incremental advances into a broader context and consider the global implications of the developing technology. AI is about to change your world, so pay attention.
Rujul Zaparde, CEO and co-founder of Zip, discusses how the company's AI super agents handle end-to-end procurement processes, such as legal redlining and RFP creation. Zaparde explains that while transactional procurement tasks can be heavily automated, robust governance frameworks and permission controls are essential to prevent errors in enterprise systems like ERPs.
Rujul Zaparde highlights survey data showing that enterprises deploying AI deeply across business processes achieve a 39% ROI, compared to just 6% for organizations running surface-level pilots. Zaparde notes that companies must give human workers greater leverage by solving deep operational problems rather than merely assisting with basic tasks.
Discussing the scale and background of Zip, CEO Rujul Zaparde shares that the platform has processed over $1.2 trillion in cumulative spend since its founding in August 2020. Zaparde explains that Zip combines modern procurement orchestration with traditional procure-to-pay capabilities on a single, AI-first platform.
Dr. Christof Koch, a neuroscientist and physicist, explored the philosophical challenges of physicalism and the "hard problem" of consciousness. He highlighted how quantum mechanics complicates the traditional materialist view of reality, suggesting that properties like mass might not be absolute until observed.
Dr. Christof Koch explained how quantum mechanics, particularly concepts like entanglement and wave function collapse, challenges the classical understanding of physicalism. He noted that the quantum world suggests properties of objects may not exist definitively until measured or interacted with.
Dr. Christof Koch mentioned that research indicates approximately 25% of patients previously considered to be in a vegetative state are, in fact, conscious. He highlighted that these individuals are unable to communicate their state externally.
Dr. Christof Koch shared his academic and professional journey, beginning with a PhD in theoretical physics and transitioning to neuroscience. He detailed his work at Caltech, his collaboration with Francis Crick, and his roles at the Paul Allen Institute and the Tiny Blue Dot Foundation.
Dr. Koch reiterated the 'hard problem' of consciousness, questioning how physical brain matter gives rise to subjective experiences like love, hate, pain, and pleasure. He noted that identifying neural correlates, while useful, does not explain the qualitative nature of these experiences.
Approximately half of all B2B software buyers now use AI models to discover and recommend software, a significant shift from traditional keyword searches on platforms like Google. This trend is driven by the increasing capability of AI models like ChatGPT and Gemini to validate software recommendations.
G2 is evolving its mission to become the 'trust layer' for the age of artificial intelligence, helping buyers and users establish confidence in AI technologies. This is particularly important as agents and autonomous computing are expected to drive significant business results.
G2 emphasizes its rigorous approach to review verification, which includes checking user identity and verifying software ownership. This process, while creating friction, differentiates G2 from platforms like Reddit and strengthens the trustworthiness of its data for AI systems.
G2 has acquired Gartner Digital Markets, which includes the brands Capterra, Software Advice, and Get App. This acquisition was completed in February 2026, expanding G2's reach in software reviews and buyer intelligence.
G2's business model focuses on providing marketing services, including platform access for profile claiming and review management. They also offer buyer intent data and advertising opportunities for software brands during the crucial evaluation stage of the buyer journey.
Tim Sanders predicts that within three years, users will express business goals to AI agents, which will then write prompts, locate, and potentially purchase software on their behalf, with minimal human oversight. This shift is enabled by the increasing capabilities of AI and autonomous computing.
Tim Sanders, Chief Innovation Officer at G2, emphasizes the importance of 'high touch' leadership and emotional intelligence in the increasingly high-tech world of AI. Drawing from his book 'Love is the Killer App,' he advocates for selfless promotion of others' growth through knowledge sharing, relationships, and compassion.
The integration of AI is projected to deliver unprecedented business results, with potential gains of 2x, 5x, or even 10x in certain situations. However, realizing these breakthroughs is contingent upon building trust in AI technologies.
Trent Telford discusses the long-standing challenge in cybersecurity, likening traditional security measures to building a high wall that attackers simply overcome with 'bigger ladders' or by finding 'loose bricks'. He explains that despite decades of efforts to add layers of protection, a fundamental flaw persists in how we approach data security, especially in the face of evolving technology.
Trent Telford outlines the evolution of cybersecurity, from early methods like closing ports and implementing firewalls, to more complex systems. He notes that the internet was not originally designed for security, leading to a continuous 'chase' to patch vulnerabilities rather than a fundamental fix.
Trent Telford argues that traditional cybersecurity models, focused on building 'walls' around networks, are failing due to the increasing borderless nature of technology. He highlights how the proliferation of smartphones, AI, drones, and multinational corporations disperses data beyond any traditional boundaries, rendering perimeter defenses obsolete.
The conversation touches upon a company called Quanty (Quantum API), which is developing a new strategy for data security. While the details of their quantum API are not fully elaborated, their approach starts from the assumption that data will inevitably be exposed, signaling a shift from traditional protective measures.
A brief historical note is made about Mark Andreessen, a prominent American venture capitalist, being one of the first to invent SSL (Secure Sockets Layer) encryption for web connections. This technology, which provides the familiar padlock icon in browsers, is still in use today.
Alexander Whedon of Subquadratic estimates that current AI coding agents spend approximately 86% of their time reading and understanding code, rather than actively generating or solving problems. He notes this indicates a significant focus on comprehension and context assimilation within AI development for coding tasks.
Alexander Whedon of Subquadratic discusses the current limitations of AI coding agents, highlighting that a vast majority of their operational time is dedicated to reading and understanding code. This observation is a key insight driving Subquadratic's approach to developing more effective AI coding tools.
Alexander Whedon points out that current AI models struggle with context window limitations, which is a primary reason they spend so much time reading code. If models could process larger amounts of code at once, their efficiency in understanding and solving problems would significantly increase.
Alexander Whedon explains that Subquadratic's strategy is to focus on improving AI's ability to read and understand code, as this is where most of an AI coding agent's time is spent. He believes that optimizing for comprehension will yield greater benefits than solely focusing on code generation.
Alexander Whedon argues that the current architecture of AI coding assistants necessitates a significant portion of their processing time on reading and comprehending code. He suggests that future advancements should prioritize improving these comprehension capabilities to boost overall AI performance in software development.
Marko Kushnir, Director of Communications for General Cherry, a major Ukrainian drone producer, stated that the company has significantly increased its production. From an initial output of 10 drones per month in 2023, they now produce tens of thousands, potentially nearing 100,000 units per month.
Marko Kushnir explained that General Cherry's drones are designed for the current battlefield, which he describes as a "kill zone" rather than traditional front lines. These zones can range from 1 to 20 kilometers, where drones are the primary means of operation and engagement.
Marko Kushnir stated that General Cherry develops its drone technology, including AI targeting and fiber optics, internally. The company works closely with the Ukrainian military to ensure their products are effective and meet the demands of the front lines, incorporating feedback for continuous improvement.
Marko Kushnir indicated that General Cherry is among the leading drone manufacturers in Ukraine. He estimates the company to be in the top five nationally, competing with numerous other companies, including many smaller 'garage factories'.
Marko Kushnir described General Cherry's flagship products as 'interceptor' drones designed to counter Russian aerial threats. The 'Bully' model is specifically designed to destroy Shahed (or Geran) drones, while the 'Air' model targets tactical reconnaissance drones like Zala and Lancet.
Marko Kushnir explained that the rocket-like form factor of some General Cherry drones allows for higher speeds compared to traditional quadcopters. He credits Red Bull with pioneering this form factor in 2021 or 2022, with Ukrainian companies, including General Cherry, adapting it for its effectiveness.
Erhan Giral from BMC Helix explains how AI agents are evolving to not only detect but also proactively mitigate or remediate IT issues. This is a shift from previous AI capabilities that relied on human descriptions of problems, moving towards AI gaining first-person views of real-world IT environments.
Erhan Giral of BMC Helix outlines the company's journey in IT service management, moving from the era of Remedy to the current AI-first approach with BMC Helix. He highlights the challenges and innovations in this space.
Erhan Giral from BMC Helix states that the service management market is exceptionally difficult to service, especially for small to medium businesses. He notes that the industry has adopted an "AI first" approach.
A new study by Virtana reveals that 60% of enterprises cannot automatically identify the root cause when their AI workloads fail. This lack of observability makes it difficult to implement effective remediation, leading to a high percentage of AI workloads failing.
Paul Appleby, CEO of Virtana, predicts a shift in how AI return on investment (ROI) will be measured. He believes the focus will move from pure model performance to operational efficiency and cost-effectiveness, as the initial 'gold rush' mentality for AI investment subsides.
Paul Appleby, CEO of Virtana, explains that true observability for AI factories is challenging due to the complexity of the entire system, from data pipelines to orchestration and infrastructure. He highlights that many current solutions only offer partial telemetry, making it hard to pinpoint issues.
Paul Appleby of Virtana points out that inefficient AI operations lead to idle GPUs and significant energy waste. He emphasizes the need for better utilization and throughput to maximize ROI from AI investments, and reduce the environmental impact of data centers.
Paul Appleby, CEO of Virtana, states that the rapid investment in AI infrastructure, often termed 'AI factories,' is outpacing the implementation of necessary governance and controls. This oversight increases operational risk for businesses.
Paul Appleby of Virtana explains that their observability platform is designed to provide end-to-end visibility for AI factories, a critical need as AI adoption accelerates. The platform aims to help companies manage complex systems and identify issues that hinder performance and ROI.
Braden Rogers, Chief Customer Officer at Island, discussed the company's new AI-focused security features designed to manage the risks associated with autonomous agents and AI tools integrated into enterprise environments. These capabilities aim to provide visibility and control over AI usage, ensuring compliance and data protection.
The conversation highlighted how the traditional browser, initially designed for information consumption, has evolved into an application delivery platform. Island aims to transform the browser into a more robust application delivery platform by building in local mechanics and policies for user experience, data protection, and productivity.
Braden Rogers described the challenge enterprises face with the 'chaos' of AI tools, emphasizing the need to empower users while maintaining safety and regulatory compliance. He advocates for an 'AI first' mindset, integrating sanctioned AI into workflows rather than blocking everything, to foster user productivity and innovation.
Island's AI Protect feature allows for safe usage of any AI application, whether consumer or enterprise-grade. Braden Rogers explained that AI Protect can live anywhere the user does, whether in an enterprise browser, a consumer browser with an Island extension, or via Island Desktop, asserting policies to govern AI resource usage.
The discussion touched on the risks of autonomous agents and AI tools designed for consumers being integrated into enterprise environments. Braden Rogers highlighted that organizations struggle with the rapid influx of these tools, needing to balance user empowerment with fundamental security and regulatory requirements.
Kriti Sharma, CEO of IFS's Nexus Black unit, stated that their AI solutions are designed to deliver value to clients within approximately three weeks. This is achieved by focusing on real-world problems and being present on-site with clients.
Nexus Black tailors AI solutions for field technicians by considering their practical constraints, such as the inability to remove safety gloves to type. This approach leads to the use of voice transcription and other adaptable form factors for their AI tools.
Kriti Sharma highlighted the challenges of deploying AI in industrial environments, particularly in offshore settings where Wi-Fi is often unavailable. Her team prioritizes building AI solutions that operate effectively on the edge, with compressed knowledge deployed on devices.
Nexus Black worked with William Grant, the maker of Glenfiddich and Hendrick's Gin, to address production loss. They found that the distillery was spending 38% of its time on emergency and corrective fixes for machinery, impacting batch production.
Nexus Black leverages AI models like Claude from Anthropic to interpret complex industrial documents, such as piping and instrumentation diagrams. This allows them to help technicians understand how issues with one component might affect other parts of a plant.
Kriti Sharma described Nexus Black as having a team of 'elite engineers' who are experts in building production-grade AI products. This team combines their technical prowess with deep subject matter expertise to solve challenging problems in industrial settings.
Rubric has launched "Rubric Agent Cloud," a new offering that combines their AI platform with data and identity security. The goal is to manage the risks associated with AI agents operating within organizations. This move follows Rubric's acquisition of Preda base, an AI infrastructure company founded by Doug, who previously worked at Google.
Rubric positions its Agent Cloud as a security layer above existing agent orchestration tools. The company focuses on providing consistent guardrails and runtime controls for AI agents, regardless of where they are built or deployed. This is seen as crucial because AI agents are becoming more complex and integrated into business operations.
Doug, co-founder and CEO of Preda base, shared his background in AI, including his work at Google on the Google Assistant and Google Cloud's AI platform (later Vertex AI). His early focus was on resilient machine learning algorithms for privacy-preserving use cases. He also mentioned his experience at Uber building machine learning frameworks.
Rubric's core business began with data and cyber resilience, focusing on business continuity against downtime and cyber threats. Their philosophy evolved to 'assume breach,' emphasizing resilience rather than just prevention. With the rise of AI agents, Rubric has expanded its focus to address the unique security challenges posed by these new systems.
The increasing adoption of AI agents presents a significant challenge for organizations due to a lack of robust security frameworks. Unlike traditional software, AI agents' operations are harder to monitor and control, making them vulnerable to risks such as unauthorized data exfiltration. Rubric aims to solve this with their Agent Cloud offering.
Philip Goff argues that Galileo's decision to separate consciousness from the domain of science, by deeming it purely mathematical and quantitative, created the mind-body problem. Goff suggests that this division, influenced by religious context and Galileo's own philosophical innovations, led to science focusing on the physical world while leaving subjective experiences like qualia outside its scope. He proposes pansychism as a way to reintegrate consciousness into a scientific framework.
Philip Goff notes a historical trend in attributing consciousness to a wider range of beings, starting with only humans, then extending to animals, and recently to insects and reptiles. This progression suggests a broader understanding of consciousness, which aligns with the pansychist view that consciousness is a fundamental property present even in the simplest building blocks of the universe.
Philip Goff proposes pansychism as a potential solution to the mind-body problem, suggesting it can help reintegrate consciousness into the scientific framework from which Galileo arguably excluded it. He argues that Galileo's division of reality into a quantitative physical world and a non-quantitative mental world, while enabling scientific progress, left subjective experience unaddressed.
Philip Goff challenges the common assumption that Galileo's separation of consciousness from the physical world was solely for scientific expediency. Goff posits that this move was actually quite 'rebellious' against the prevailing Aristotelian and Church views, which attributed qualities like colors and sounds to the physical world itself, not just the soul. Galileo's division, by placing subjective qualities outside the purview of mathematical science, set the stage for the mind-body problem.