🔥 Welcome to Agent Talk — a brand new series from Factory Fireside where our young engineers in the trenches building enterprise-scale agentic solutions break down what's really happening in AI.

In this debut episode, Kahn Nguyen, Neil Mazunder, and Harry Nguyen dive deep into the AI developments that actually mattered in 2025 — and what turned out to be just noise.

What we cover: → Is Opus 4.5 + Claude Code basically AGI? (Hot take alert 🌶️) → How reinforcement learning became THE word of 2025 → DeepSeek's papers that shook Silicon Valley → The shift from scaling laws to post-training optimisation → Why Chinese labs are outpacing American research → "Code is solved, but software isn't" — what this means for builders → Our predictions for 2026 and the next age of AI research

About Agent Talk: InsightFactory has been heavily invested in developing enterprise-grade agentic solutions. Agent Talk is hosted by three of our engineers who are right in the depths of building these solutions for clients — not just agentic solutions, but enterprise-scale agentic solutions.

🎙️ Hosts:

• Kahn Nguyen — Research Scientist @ insightfactoryAI
• Neil Mazunder — Forward Deployed Engineer @ insightfactoryAI Applied AI Team
• Harry Nguyen — Forward Deployed Engineer @ insightfactoryAI Applied AI Team

📍 About insightfactory.AI: We're an Australian AI company building enterprise-grade agentic solutions through our Forward Deployed Engineer model.

In this episode of Factory Fireside, I sit down with Dr. Sanjay Mazumdar — a pioneer who was building neural networks in the early 1990s before the internet even existed, and who has since shaped Australia's approach to AI, data, and national security.

Sanjay's journey spans developing the world's first mobile wallet at Motorola (in 1999!), leading the Data to Decisions CRC that spawned companies like Fivecast, serving as Chief Data Officer at KPMG, and now running the Defence Trailblazer — a $240 million initiative connecting universities, startups, and defence.

We go deep on:

→ Why most organisations get AI adoption completely backwards — and what to do instead

→ The real role of a Chief Data Officer (hint: it's not data governance)

→ Information warfare, deepfakes, and why guardrails alone won't protect us

→ Australia's quantum computing edge — and the encryption risks that come with it

→ What it actually takes to build a successful AI company (spoiler: it takes years)

→ Practical advice for business leaders preparing for the AI wave

Whether you're a founder, executive, or just trying to make sense of where AI is heading, this conversation delivers hard-won insights from someone who's been at the frontier for over three decades.

🎯 Key takeaways:

[00:00] — Introduction

[02:15] — Doing a PhD in neural nets before the internet existed

[08:30] — The world's first mobile wallet (1999)

[15:45] — Why "AI envy" is derailing business strategy

[24:10] — Where the CDO role should actually sit

[32:00] — Information warfare and the limits of ethical AI

[41:20] — Inside Australia's Defence Trailblazer

[48:30] — Quantum computing: opportunity and threat

[55:00] — How Fivecast went from research project to global success

[62:15] — Three things every business leader should do now

In this episode of Factory Fireside, Peter Inge sits down with Dr. Don Perugini, a pioneer in AI and agentic systems with over 20 years of real-world experience spanning defence, enterprise, healthcare, and startups. Together, they cut through the hype to explore what agents actually are, why most enterprise pilots stall, and what it really takes to move from experimentation to production impact.

This is not a surface-level LLM conversation. Don draws on decades of work building cognitive agents, federated AI systems, and large-scale simulations from predicting bushfire behaviour and modelling entire populations, to improving IVF success rates and enabling privacy-preserving AI in healthcare.

You’ll hear:

• Why agents didn’t start with LLMs, and why that history matters
• The real reasons 95% of AI pilots never reach P&L impact
• Why “narrow problems” outperform grand AI ambitions
• How agentic commerce and the agentic web will change SaaS and B2B
• Why AGI is mostly a distraction and what progress actually looks like
• What founders and technologists should focus on now to build durable AI businesses

This episode is essential viewing for technology leaders, founders, product teams, and anyone trying to understand where AI agents are genuinely delivering value and where the hype still vastly outweighs reality.

🎧 If you care about building AI that works in the real world, this one’s for you.