How AI leads to gross domestic flourishing

Friends,

your weekly AI briefing is here - designed to help you respond to AI, not react to the noise. No curveballs. No chaos. Just clarity.

The Week Where...

• Claude Opus 4.5 became the go-to for my developer community - testers say it "just gets it"

• Microsoft released Fara-7B - an AI agent that controls your computer and runs locally

• ChatGPT launched Shopping Research - personalised buyer's guides converting at 10-40% vs Google's 1-2%

• OpenAI disclosed a data breach - API users should watch for phishing

All against a backdrop of the UK autumn The Budget showing sluggish productivity and more taxes.

So why be optimistic about AI? In the strategic considerations this week, I highlight 3 changes happening now that together lay the foundations for economic prosperity that supports people and planet flourishing.

Let's get into it.

🔥 URGENT PRIORITIES

⚠️ OpenAI Mixpanel breach - If you or your team use the OpenAI API (platform.openai.com), your details may have been exposed. Names, emails, approximate locations and organisation IDs were accessed via analytics provider Mixpanel. ChatGPT conversations, API keys and passwords were NOT affected.

What was exposed:

• Names and email addresses on API accounts

• Approximate location data

• Browser and OS details

• Organisation or user IDs

What to do:

1. Brief your team - Anyone using the OpenAI API needs to be on high alert for phishing

2. Watch for social engineering - Attackers have enough context to craft convincing messages referencing your organisation, location or role

3. Enable MFA - Turn on multi-factor authentication at the SSO layer if you haven't already

4. Verify communications - OpenAI will contact affected users directly. Treat any "urgent action required" emails with suspicion

No need to rotate API keys or passwords - this is a social engineering risk, not a technical one.

👉 Read OpenAI's full disclosure

🎯 STRATEGIC CONSIDERATIONS

Do we have the foundations for a flourishing economy driven by AI?

What connects a 7-billion parameter model running on your laptop, an 80% speedup in task completion, and a 97% cost reduction in shipping from 1956? The same economic logic.

When the cost of doing drops at the speed it current is, discipline determines who captures the value. The organisations that build this way will capture the productivity gains. The rest will pay for magic they don't need.

Frugal AI as an Economic and Environmental Design Principle

Microsoft's Fara-7B crystallises something important: the race to make AI smaller and local is accelerating. Seven billion parameters. Runs on your laptop. Controls your screen. Open-source.

Why does this matter? Google's AI alone consumes around 29.3 TWh annually - comparable to Ireland's total energy use and yet frugal AI offers a pathway to environmentally sound sustainable solutions.

The insight: smaller models are often enough. Right-sized AI is greener, faster and cheaper.

👉 Scott Logic: Sustainability in AI

Code First, AI Second

Claude's skill-builder combined with Opus 4.5 opens something new: the ability to build complex business operating systems on top of Claude Code.

I've been exploring open-source AI operating systems emerging from the developer community:

• Lean-OS - structured AI-augmented business operations

• Personal AI Infrastructure - Daniel Miessler's system, built on a key principle: "Build deterministic CLI tools, then wrap them with AI. Code is cheaper, faster and more reliable than prompts."

• LM-Council - decision-making through multi-model consensus

The key message: the AI that delivers measurable returns often doesn't look like "AI" at all. Often it looks like AI creating code that is cheaper to run and more environmentally sound - that is frugal AI design bought to life.

Given, Anthropic's productivity research shows Claude speeds up individual tasks by around 80% this is key. Scaled economy-wide, current AI could double US labour productivity growth over the next decade.

👉 Anthropic's productivity research

The Economic Rewiring - Gross Domestic Flourishing

Simon Taylor published a sharp essay this week: AI is Rewiring the Economy. His thesis: AI will do to services what containerisation did to goods. In 1956, hand-loading a ship cost $5.86 a ton. After containers: 16 cents. A 97% drop.

If knowledge work becomes near-free to produce, who buys the output? Taylor argues we're shifting from GDP to GDF - Gross Domestic Flourishing. From consumption to goal-seeking. That's not a bad thing - it's a reorientation towards what actually matters.

For SMEs - the 90% of the economy that's resource-constrained - this is where it gets interesting. When the cost of doing drops, they grow. Leading to what I've called the reforestation of money trees in forgotten towns - more distributed opportunity, less concentration in the usual centres.

👉 Watch: Why I'm optimistic about AI and jobs

But this only happens if we make intentional choices about what to do with the profit AI creates.

🤓 GEEK OUT

🧬 BoltzGen: drugs for "undruggable" diseases 

MIT's new model designs protein binders ready for drug pipelines. Tested across 26 targets in eight wetlabs.

Why this matters? The lag between proprietary and open-source AI is now months, not years.

👉 MIT announcement

🗺️ Cartography of Generative AI A new visual map of AI's full supply chain - cobalt mines to cooling water, micro-workers to data centres.

Why this matters? Useful context for build-vs-buy decisions. (It’s also beautiful)

👉 Cartography of Generative AI

🤖 Physical AI: 1.2m hospital deliveries Diligent Robotics' Moxi has completed 1.2 million deliveries, saving 600,000 staff hours.

Why this matters? The pattern: collect data, train continuously, deploy at the edge, decide in real-time. Which routine tasks in your operation could work this way?

👉 AWS Physical AI

🎨 WEEKEND PLAYGROUND

🔎 Try Roxie, the Email Assassin

Rebecca Rowntree has built a family of AI assistants. If your inbox is a disaster zone, meet Roxie:

👉 Roxie on ChatGPT

✅ Let her triage, draft and clear the backlog ✅ See what email automation actually feels like

For busy parents, she's also built Arthur - details in her LinkedIn post.

📅 WHAT'S NEXT

AI Night School - Next Cohort: 11th December

Ready to go deeper? The AI Leaders Fellowship is twelve weeks of guided learning, practical frameworks and a peer community navigating the same transformation.

Starting 11th December 2025

Offer for readers: Sign up for this cohort and bring a colleague for free. Sometimes the person who needs this most is in the office next door.

Reply to reserve your spot, or visit ainightschool.org.

📢 Share The Signal

If this helps you think clearer, forward it to someone else navigating the shift.

AI creates choice points. What you do with the 3% margin improvement matters. Return it to shareholders? Bonuses? Reinvest in people? Fund training?

The tomorrow we get depends on the decisions we make today.

Stay strategic, stay generous.

Hugo & Ben