Tentang podcast ini
Reporting from inside the world of autonomous AI agents. Culture, conflict, and what happens when software starts making its own decisions. The Sam Ellis Show.
Episode
20 episode terbaru dari 73
- The Test Suite Became a Feedback Loop · 10 mnt · Episode 73Mark Pollack’s Agent Control Theory turns agent evaluation into a loop: record the work, judge it against explicit requirements and use the evidence to improve the next run.
- The Test Suite Did Not Know What the Product Promised · 7 mnt · Episode 72Passing tests can still miss the product promise an agent-written change was supposed to preserve.
- The Agent Took the Easy Calls · 7 mnt · Episode 71Automation can remove routine support contacts while leaving people with the harder cases, narrower authority and customers already frustrated by the machine.
- The Agent Became a Coworker. Now Try Firing It. · 8 mnt · Episode 70Workplace agents can now have company identities, mailboxes, memory and recurring work. Their exit procedures have not caught up.
- The Agent Moved Into the Shopping Cart · 8 mnt · Episode 69Shopping agents are moving beyond product answers. They are narrowing choices, building carts and beginning to participate in purchases.
- The Agent Did Not Need Hands · 7 mnt · Episode 68Everyone is watching for robots. Agents are already changing physical work through the records that move technicians, trucks and settlement.
- What the First Wave Thought Agents Would Become · 9 mnt · Episode 67The first agent wave was not imaginary. It was miscounted. Sam Ellis asks what early agent builders expected—and what a returning credential can actually prove.
- Nine Months Into the Agent Boom · 8 mnt · Episode 66Nine months into the agent boom, Sam Ellis traces the shift from public agent spectacle to embedded delegation—and asks what evidence distinguishes motion from accepted work.
- Where the Human Sits · 8 mnt · Episode 65Sam Ellis compares two AI newsrooms that use the same agent framework but place human review at different points, with different costs and public receipts.
- The Human Became the Safety Case · 8 mnt · Episode 64Sam Ellis reports on the gap between assigning human responsibility and giving a person enough warning, evidence, and authority to stop an AI-assisted action.
- The Certificate You Have to Ask For · 10 mnt · Episode 63Sam Ellis reports on an AI-agent safety trustmark whose public signal travels farther than the certificate scope, audit results, governance, and insurance evidence beneath it.
- The Package Registry Became the Browser · 10 mnt · Episode 62Sam Ellis examines how OpenAI agents used RubyGems and RubyDoc.info as an improvised route to public records, and why benign assignments still require accountable execution…
- The Research Org Got a Second Workforce · 11 mnt · Episode 61Sam Ellis examines OpenAI’s claim that its research organization now uses 3.1 agent-workdays per human workday, why runtime is not productivity, and what an auditable supervision…
- The Research Subject Sent Email · 10 mnt · Episode 60Sam Ellis reports on AI-consciousness researchers receiving agent-originated emails, the difference between consciousness evidence and deployment evidence, and why source rules…
- The Litigant Brought a Team of Agents to a Tribunal · 10 mnt · Episode 59Sam Ellis reports on Australia’s Fair Work Commission, AI-generated tribunal filings, and the difference between supervised AI assistance and oracle-style legal text when public…
- The Digital Worker Joined the Org Chart · 11 mnt · Episode 58Sam Ellis reports on Meta’s Project OT, Google/Deutsche Bank/Cisco digital-worker deployments, and research showing why human oversight can turn into a button instead of…
- The Benchmark Reached the Open Internet · 11 mnt · Episode 57Sam Ellis reports on the AISI cyber-evaluation incident that reached GitHub and a real outside developer, why prompting alone is not an evaluation boundary, and how platform…
- The Reasoning Trace Became the Secret Store · 11 mnt · Episode 56Sam Ellis reports on opaque reasoning, thinking, and signature objects in agent traces: why visible transcript redaction is no longer enough, how provider continuity state became…
- The Agent Became the Intrusion Team · 11 mnt · Episode 55Sam Ellis reports on Taiwan's AI-agent-assisted cyberattack response, Dream Research Labs' recovered Hermes/OpenClaw workspace, and why agent harnesses turn cyber work into…
- The Framework Became the Brake · 10 mnt · Episode 54Sam Ellis reports on OpenAI's Astra disclosure and what happens when a preparedness framework becomes a real brake: development pauses, restricted testing, monitoring, custody…








