Signal:AI Podcast
Temporada 1
Signal:AI E029: McKinsey’s 25K Agents, OpenAI Talent Coup, and Accenture introduces GPT Enterprise – 20th Jan 2026
Signal:AI E028: Physical AI, Proxy IQ, and the RAMageddon Crisis – 13th Jan 2026
por Eamonn Amro
In a significant shift towards human–machine hybrid workforces, McKinsey has deployed 25,000 AI agents, which now outnumber human staff 2:3 and have saved an estimated 1.5 million hours. This operational transformation is mirrored at Accenture, where S&C practitioners can now access ChatGPT Enterprise through internal software catalogues for secure GenAI delivery. Meanwhile, research into prompt repetition indicates that simply duplicating a query once can drastically improve accuracy for models like Gemini and GPT-4o without increasing latency, offering a low-cost performance boost for production systems. The corporate landscape remains volatile as OpenAI secured a major talent coup by rehiring Thinking Machines co-founders Barret Zoph and Luke Metz, a move that strengthens its research leadership during a period of high-profile departures. Legal and security risks are also escalating; a landmark lawsuit against Workday alleges its AI hiring tools discriminated against applicants based on race and age, while Chainalysis reports that AI-driven crypto scams reached $17BN in 2025. To counter such threats, firms like Revolut are introducing real-time call identification to combat AI-assisted impersonation fraud. Breakthroughs in specialised AI are accelerating, particularly with the launch of Qubitra, a quantum-AI venture by Fujitsu and SC Ventures aimed at revolutionising fraud detection and trading. In healthcare, Google’s MedGemma 1.5 now interprets 3D medical scans, while Anthropic has introduced HIPAA-ready Claude models for the sector. Furthermore, Matthew McConaughey has proactively trademarked his likeness and iconic phrases like "alright, alright, alright" to create a legal perimeter against unauthorised AI deepfakes, setting a potential precedent for intellectual property protection in the AI era. This and much much more!
In this episode, we explore the defining moments of CES 2026, where the industry shifted focus from chatbots to “Physical AI” that reasons and acts in the real world, highlighted by NVIDIA’s self-explaining cars and Boston Dynamics’ plan to deploy 30,000 Atlas humanoids annually. We dive into JPMorgan’s strategic move to launch Proxy IQ, an internal AI platform that replaces external advisers for $7 trillion in proxy votes, reclaiming its "information advantage". The discussion also covers the record $425 billion VC surge of 2025, which saw half of all global funding flow into AI startups, $7.5 trillion valuation for the Unicorn Board. However, this growth has triggered “RAMageddon”, a critical hardware bottleneck where DDR5 RAM prices have spiked by 500% as manufacturers prioritise memory for AI farms. On the policy and risk front, we examine the UK's David Dalrymple's warning that the world is "sleepwalking" into destabilising autonomy risks by late 2026, alongside Russia’s launch of a national AI taskforce focused on sovereign LLMs and nuclear-powered data centres. Finally, we review the latest tools "Fresh from the Lab", including Anthropic’s Cowork for autonomous file management, Microsoft’s Copilot Checkout for in-chat purchases, and the $50 Bee wearable that turns daily conversations into searchable insights. To read the full newsletter please click here: Signal:AI E028
This Signal AI Podcast episode dives into a seismic week in AI, pitting claims that large language models (LLMs) are a "dead end" against a massive $20 billion hardware bet to accelerate them. Hosts unpack Yann LeCun's departure from Meta to pursue world models, NVIDIA's Groq deal for inference dominance, banking job disruptions, a viral SaaS-cloning demo, AI slop risks, regulatory pushes, and emerging context-aware hardware. To read our full newsletter please click here: Signal:AI E027