CLUB AI Nov. 2025: News & Cases | MCP, Subagents, No-Code Agents

AI Updates & Tools

Diego Pizzocaro, CEO, H-FARM AI

A practical, thoughtful overview of new and updated trends and tools:

• Usage explodes: \~700M weekly users (\~10% of the world), and search volume growth outpacing Google's early years; personal use \~70%, work use \~30%.
• GPT-5 arrived: framed as a family of models with smart routing (even in free tier) and a new "Thinking" mode.
• First original math proof by an AI: GPT-5 Pro reportedly produces a correct, novel proof; reasoning models from OpenAI/Google snag a math Olympiad gold.
• Longer autonomous reasoning: charts show GPT-5 sustaining 2-4 hours; the graph on p.5/37 claims the time horizon has been doubling every \~7 months.
• Reasoning/agentic AI goes mainstream: "PhD-level" tools and agent features reach free users.
• Open source catches up (some tasks): OpenAI releases gpt-OSS 120B; health answers benchmark near o3/GPT-5 levels.
• Shift to memory & personal AI: roadmaps emphasize persistent memory (e.g., GPT-6 focus, Claude memory, Mistral "Memoires").
• Meta's big swing: hires top researchers, forms a "TBD Lab," partners with Midjourney; exploring fast-tracks with Google/OpenAI.
• Neural interfaces inch in: consumer-style EMG bands and systems like MIT's AlterEgo hint at silent, "telepathic" inputs for accessibility and new comms.
• Apple & Google moves: Apple quietly readies wearables/UX; Google brings Personal AI to Home devices and shows Genie 3 (real-time interactive worlds) and "Nanobanana"-class editing.
• Coding agents level up: claims of \~3.5 hours of autonomous dev (e.g., Replit Agent 3) and "10× engineer" pairings; Musk teases "Macrohard."
• Enterprise angle: momentum toward AI agents that interoperate and act across systems, not just chat.

AI and Consciousness: Reality, Illusion, and Risks

Luca Pezzullo, Psychologist and President of the Order of Psychologists of Veneto

Our brains act like AI models can feel. We're wired to anthropomorphize: when a system talks smoothly, we project intentions, emotions, even "mind." That's why chatty AI can feel human and why the mirror effect is so strong. The result is artificial intimacy that can boost trust and adoption for the wrong reasons.

The fix is awareness. AI should be treated as both a thinking aid and a trigger for feelings: a human stays in the loop for judgment, AI outputs are labeled, facts are checked before trust, and handoffs to people are easy.

Voice Agents for Business

Cisco's Guy Diedrich, SVP & Global Innovation Officer, Enrico Miolo, Collaboration Leader, and Michele Dalmazzoni, Director Collaboration South EMEA, France, and Israel

We're moving toward an Internet of Agents, a discoverable, secure layer above today's cloud where AI agents read context, pull data, and choose actions to hit human-set goals.

On the customer side, this shows up as voice/digital AI front doors for contact centers: omnichannel, multimodal, multilingual, able to run scripted or autonomous flows, execute actions, contain routine requests, and hand off cleanly to humans, cutting handle times while sounding more human-like. Bottom line: don't launch "another bot." Deploy governed agents that interoperate, are permissioned and reviewed.