BRIEFING 10
What Model Distillation Is
The aim is not simply smaller AI. It is more practical AI: a system that does the needed thing well and cheaply.
Short form
Briefings
One idea at a time. Short audio notes from the AI Overview, each published with its video and a full transcript.
Latest briefing
00:00
0:13
BRIEFING 09
What an AI Sandbox Is
A sandbox lets people examine how a system behaves, and what is at stake, without exposing anything real.
Read the transcriptAll briefings
BRIEFING 11
Why Timing Matters in Real-Time AI
Real-time AI is about timing as much as intelligence. An answer that comes too late is a different answer.
BRIEFING 12
What Makes Voice AI Feel Natural
A better voice is not louder. It is more human, and most of that comes from timing rather than from sound quality.
BRIEFING 13
Why AI Interoperability Matters
Better systems should connect without becoming chaotic, and that depends on how the connections are defined.
BRIEFING 08
The Quiet Side of Automation
The most useful automation rarely looks impressive. It removes the small, repeated steps that sit between the work people actually want to do.
BRIEFING 07
Who Is Accountable for AI Decisions
A system that prepares the decision has not made it. Accountability stays with the person who acts on the recommendation.
BRIEFING 06
What Synthetic Data Is For
Synthetic data fills gaps that real data leaves behind. It inherits the standards of whoever designed it, which is both its strength and its limit.
BRIEFING 05
Testing AI on Real Tasks
A benchmark a system has effectively memorised stops measuring anything. Evaluation is shifting towards tasks that resemble the work itself.
BRIEFING 04
Why AI Transparency Matters
Transparency about how content was produced is not a courtesy. It is the information a reader needs in order to weigh what is in front of them.
BRIEFING 03
Matching the Model to the Job
Routing simple work to a smaller, faster model is not a compromise. It is what a considered workflow looks like.
BRIEFING 02
AI That Plans and Checks Its Work
A system that plans, monitors its own progress and reports back is a different kind of tool from one that returns an answer. It is not a different kind of authority.
BRIEFING 01
Delegating Work to AI Agents
Agents change how work gets done, not who answers for it. The process is delegable; the judgement is not.
Looking for the long-form conversations instead? Start with the full episode archive or the written resources.