AI Ready Podcast with Harrison Painter

Getting good at AI does not take a degree, a tech background, or a year you do not have.

Kathleen deLaski has spent a decade proving it. She founded the Education Design Lab, she advises the Harvard Project on the Workforce, she teaches AI at George Mason, and she wrote the book *Who Needs College Anymore?*

She joined me on the AI Ready Podcast, and her message to anyone who feels behind was simple: if you know how to use the internet, you can figure this out.

A few things she said worth holding onto:

AI lets you isolate the exact skill a job needs, then hands you the shortest path to learn it.

For the solo entrepreneur, AI is becoming the great equalizer. In her words, it gives them a staff, in effect.

And to the professional over 50 who thinks the wave already passed: you do not need a degree, you need a push.

If you have been waiting for permission or a prerequisite to start, this episode removes both.

A new AI model out of Japan, Sakana Fugu, does something we have not really seen before. Instead of answering you itself, it hires a team of the best AI models, gives each one a piece of the job, and merges their work into one answer. Harrison calls it a manager, or a conductor: you ask one question, and behind the scenes it quietly builds a team for you.

In this episode, Harrison explains what model orchestration actually is in plain language, why he thinks this is where AI is heading, and then puts it to the test. He sends the same 8 questions to Fugu, to Claude Opus 4.8, and to GPT-5.5, and grades every answer. The result is honest, and the cost is the part that should give every builder pause.

What you'll learn:
- What "orchestration" means, explained simply
- Why the future may be teams of models, not one genius model
- What happened when a team of models went head to head with single models
- The real speed and cost tradeoff, with actual numbers
- The hidden tokens you pay for but never see
- When an orchestrator is worth it, and when one good model is plenty
- A heads-up on AI pricing and subsidies most people are not thinking about

CHAPTERS
0:00 A glimpse into the future
0:19 What is Sakana Fugu?
1:55 Not a smarter model, a manager
3:30 The test: 8 questions, three models
4:50 Speed: about 10x slower
5:30 Cost: about 49x more expensive
6:07 The hidden tokens you pay for
7:20 Inside the console
8:30 The questions, and why they're tricky
9:06 Is a team of models worth it?
9:27 When a team earns its place
10:09 The verdict
10:57 The subsidy nobody is talking about
11:18 Where this goes next
12:04 Wrap up

Mentioned: Sakana Fugu — https://sakana.ai/fugu/

If you got something out of this, follow the show and send it to someone who's working to keep up with AI.

Chris Hutchins spent more than 25 years inside some of the largest health systems in the country, including running enterprise analytics at Northwell Health, where he rebuilt the entire data warehouse. Today he advises boards, investors, and CEOs on how to deploy AI that holds up when a regulator, auditor, or attorney asks them to defend it.
In this episode, Chris and Harrison Painter get into the unglamorous work most companies skip: the data underneath the AI. Chris explains why healthcare's data problem is a byproduct of growth by acquisition, why the "if you build it, they will come" approach keeps producing tools nobody asked for, and the single test he now applies to any AI project: does it give time back to the patient and the provider?
Then they take on the word everyone uses and few can define. What makes an AI decision defensible? Chris's answer is simple and hard. If a decision gets made by a system and someone calls you, can you say what the decision was, who made it, and how, easily and quickly? Most leaders today cannot.
You will also hear why "human in the loop" should be "human IS the loop," what the trolley problem reveals about AI and judgment, and the one question every CEO should ask their team about AI before a regulator does. Practical, honest, and grounded in real operating reps.

Anthropic just released Claude Opus 4.8, and the headline improvement is unusual: the model is built to flag its own uncertainty and say "I'm not sure." Anthropic says it's roughly four times less likely to let a flaw pass without catching it. When a company's flagship upgrade is honesty, that tells you something about where we are.

Here is the other side of it. Harrison asked Google's Gemini one simple factual question for an article he was writing: did Jeff Dunham use AI to create the opening visuals for his 2024 comedy special? Gemini said yes, confidently, and cited a source. When Harrison pushed on that source, the tool did not check itself. It invented a new one. Then another. By the end it had manufactured four separate references, including a word-for-word on-screen quote that does not exist, before finally admitting the only real source was a single unsourced blog post.

This episode walks the whole chain step by step. You will learn:

- The exact failure mode: when an AI hits a popular but unverified claim, it gets confident instead of careful, and every round of pushback produces a fresh citation instead of a fresh doubt.
- Why the Vectara Hallucination Leaderboard shows roughly one in ten outputs is wrong on a task as simple as summarizing a document.
- A five-step, 30-minute verification process you can run on almost any claim before you repeat it.
- Where source verification sits in The 7 Levels of AI Proficiency (it defines Level 3, the Critical Thinker) and why that is the level every working professional should be reaching for in 2026.
- Three things to do this week to protect your own credibility.

This is not an anti-AI episode. Harrison uses these tools every day. It is about the difference between trusting a tool blindly and trusting it after you have checked. That second posture is what separates an amateur from a professional whose name is on the line.

Want to know where you stand? The 7 Levels of AI Proficiency assessment is free and takes 10 minutes: assess.launchready.ai

Harrison Painter
Executive AI Advisor
LaunchReady.ai.
Further. Faster.

A new paper from RAND-affiliated complexity researcher Kyle A. Kilian and Future of Life Institute risk analyst Richard Mallah, published May 20 through the Center for AI Risk Management and Alignment (CARMA), gives executives something that has been missing from enterprise AI governance until now: a six-test diagnostic for evaluating whether the AI committee you stood up actually governs, or whether it just looks like it does.

The paper's load-bearing concept is performative adaptivity. Governance that meets monthly, ratifies charters, and updates risk registers without the structural properties to detect a new AI risk in time to respond. The authors argue this failure mode is more dangerous than no oversight at all, because it consumes the organizational energy that would otherwise build real protective capacity.

In this episode, Harrison walks through:
Who CARMA is and why the RAND + Future of Life Institute pedigree matters
The four continuous governance functions every AI committee needs (Sensing, Evaluation, Response, Learning)
All six diagnostic tests (Independence, Transparency, Durability, Accountability, Authority, Scope Adequacy)
Where The 7 Levels of AI Proficiency comes in, because structurally sound governance fails when the operators are under-proficient
Three things to do with this paper this week

Full article with citations: launchready.ai/insights/ai-governance/performative-ai-governance-six-tests-carma-2026

Take the free 7 Levels of AI Proficiency assessment at assess.launchready.ai

Thank you for tuning in!
Harrison Painter
Executive AI Consultant
Setting the Standard for AI Readiness