EPISODE 2026-06-01
The launch of the daily show — a live morning broadcast with Andy Fernandez of HYCU on the AI-agent observability gap, David Villalón of Maisa on auditable digital workers, and Snehal Antani of Horizon3.ai on autonomous offensive security.
Episode one of AI in the AM as a daily live weekday morning show — a true live broadcast, growing pains included. Launch day stacked three builders working where AI agents meet the enterprise: watching what agents do, making their work auditable, and turning them loose as attackers so defenders can keep up.
Launch day aired on X only — no YouTube VOD or transcript — so the timeline below shows the planned show flow without deep links.
Launch day was broadcast on X only, with no YouTube VOD and no transcript — so this timeline reflects the planned show flow rather than caption-derived segment boundaries. From June 2 on, every episode carries real timestamps, deep-linked watch buttons, and a full auto-transcript.
Andy Fernandez, GM of AI & Cyber Resilience at HYCU, on the enterprise observability gap: AI agents now take real actions inside Salesforce, GitHub, and Okta at machine speed, while most security tooling assumes a human-speed threat model. HYCU's answer, launched in May 2026 as aiR, turns the backups companies already keep into a natural-language-queryable layer — backups as memory, not just insurance. The hosts probed the honest limits of a snapshot-based view (point-in-time backups are closer to forensics than prevention) and how a data-protection company positions against the data-security-posture tools enterprises already fund.
David VillalónThe bet that enterprise AI fails on reliability, not the model — and whether an auditable chain of work is enough to trust an agent with real tasks.David Villalón, co-founder and CEO of Maisa (and a former professional magician, building from Valencia and San Francisco), on his core argument: enterprise AI doesn't fail on the model, it fails on reliability. Maisa's digital workers run on a model-agnostic reasoning layer it calls the KPU, which plans in code and logs every step as an auditable chain of work; the company raised a $25M seed in August 2025 and describes production deployments in banking and energy. The hosts pushed on his 2025 head-to-head demo (a striking result, but a founder-run demo on his own product), whether he'd welcome an independent blind test, where auditability actually becomes correctness, and what stops the frontier labs from absorbing the reliability layer into the models themselves.
Snehal AntaniAn autonomous pentester that chains exploits to domain compromise — does building the world's best AI hacker help defenders or attackers?Snehal Antani, co-founder and CEO of Horizon3.ai, former CTO of Splunk, and previously a technologist with US Special Operations, on NodeZero — an autonomous pentester that runs continuously in production, chaining exploits from reconnaissance through domain compromise. In August 2025 the company reported NodeZero was the first AI to fully solve the Game of Active Directory benchmark, and in May 2026 it published a paper claiming formal Lean 4 safety proofs that the agent cannot degrade security posture (company-reported, not independently audited). The recurring question: if you build an AI that's genuinely good at attacking, does that help defenders or attackers? The hosts pressed on proliferation (why "defenders run it first" is enough when the same capabilities spread fast), what formal proofs do and don't guarantee, and the offense-defense balance twelve months out.