The Big AI Questions to sit with this Christmas

Friends,

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

As 2025 draws to a close, OpenAI reminded us that the pace isn't slowing for the holidays. This week they launched GPT Image 1.5 - up to 4x faster image generation with dramatically better instruction-following and editing precision. It's a clear shot across the bow at Google's Nano Banana Pro, and a signal that the image generation wars are just getting started.

But this week's newsletter isn't about the race. It's about the pause. The questions worth sitting with over mince pies and mulled wine. The kind that need space to percolate.

Let's get into it.

🔥 URGENT PRIORITIES

✅ No major model changes requiring immediate action

✅ No critical security updates this week

✅ No compliance deadlines before the new year

Use the calm. This is your window for reflection, not reaction.

🎯 STRATEGIC CONSIDERATIONS

The Christmas Question: What If We Got More Ambitious?

As we wind down for the festive break, this is the perfect time to let some big questions percolate. Not the tactical questions - those can wait for January. The deeper ones that need time, space, and perhaps a long walk after too much turkey.

Here's the one keeping us up at night: How do organisations become more ambitious with AI, not less?

We talk about this in our AI Optimist videos - AI genuinely enables us to solve bigger problems; Superhuman problems. The kind of challenges that were previously impossible because they required processing scale, pattern recognition, or coordination that human teams simply couldn't achieve.

And yet. The dominant narrative is rationalisation. Cost reduction. Efficiency. "Doing more with less" - which too often translates to "doing the same with fewer people."

This is understandable. Economic headwinds are real. Boards want certainty. Shareholders want returns. But here's the tension: if we only use AI to optimise for what we already measure, we miss the transformation entirely.

Why this matters: What we measure gets done. Our obsession with productivity risks becoming a trap. Productivity measures output per input - it says nothing about whether the output matters. You can be phenomenally productive at the wrong things.

What if we measured something different? Something like Gross Domestic Flourishing - not just what we produce, but whether it makes life better. Whether work becomes more meaningful. Whether businesses solve problems that actually need solving.

This isn't utopian dreaming. It's strategic positioning. The organisations that use this technology to expand what's possible - rather than simply shrink what's necessary - will define the next decade.

Your move: Over the break, between mince pies and family debates, let one question sit with you: What problem could your organisation now tackle that was previously impossible? Not "how can we cut costs?" but "what becomes achievable when cost and complexity constraints shift?"

The Upskilling Paradox

There's a second question that follows from the first: How do we find time to upskill when the work keeps coming?

This is the catch-22 facing every leader we speak to. You know your teams need new capabilities. You know AI fluency matters. But the day job doesn't pause while people learn.

Here's what we've observed: the organisations making real progress aren't choosing between training and working - they're using structured learning to accelerate how quickly people can learn through doing. Without a foundation in AI literacy and fluency, people fumble. They don't know what questions to ask, what's possible, or where the risks lie. With that foundation, every interaction with AI becomes a learning opportunity that compounds.

This is exactly why we built AI Night School - to give teams the frameworks and fluency that turn daily work into continuous development. Not theory for theory's sake, but the practical foundation that makes "learning by doing" actually work.

Your move: If up-skilling has been stuck on your "important but not urgent" list, the new year is your moment. Our January cohorts are filling now - explore AI Night School and give your team the head start that makes the rest of 2025 count.

🤓 GEEK OUT

🎄 "Slop" is officially the word of the year

Merriam-Webster has crowned "slop" - defined as "digital content of low quality produced usually in quantity by means of artificial intelligence" - as their 2025 word of the year. From absurd videos to talking cats to "workslop" reports that waste colleagues' time, the flood of AI-generated mediocrity has earned its own four-letter dismissal.

👉 Read the announcement

Why this matters: The word sends a message to AI that's "less fearful, more mocking" - when it comes to replacing human creativity, sometimes it doesn't seem too superintelligent. A festive reminder that quality still trumps quantity.

🧠 How Claude's memory actually works

Developer Manthan Gupta reverse-engineered Claude's memory system and found something fascinating: while ChatGPT injects pre-computed summaries into every prompt, Claude uses on-demand tools that selectively retrieve past conversations only when relevant. User memories are stored in XML format, and Claude decides in real-time whether to search conversation history or pull recent chats.

👉 Read the full breakdown

Why this matters: Understanding how these systems work under the hood helps you use them more effectively. Claude's approach trades automatic continuity for on-demand depth - it can access more detail when needed, but you may need to prompt it to "remember" relevant context.

🏠 Local AI: The trend to watch in 2026

Small Language Models (SLMs) running entirely on your own devices are set to reshape enterprise AI next year. The promise: AI that never sends your data to the cloud, works offline, costs a fraction of API calls, and can be fine-tuned for your specific domain.

👉 Explore local AI search engines

Why this matters: As data privacy regulations tighten and cloud costs compound, running AI locally shifts from "nice to have" to competitive advantage. Start experimenting with tools like Ollama now - by this time next year, local-first AI will be table stakes.

🎨 WEEKEND PLAYGROUND

🎄 Try Comet - Perplexity's agentic browser that doesn't just search, it does

This week's mission: Plan your Christmas viewing schedule for the whole family

Ask Comet to research what's on BBC, ITV, Channel 4, and your streaming services on Christmas Day. Have it compare options, note the timings, and compile a viewing schedule that keeps everyone happy - from the kids who want animation, to Gran who wants the King's Speech, to the teenagers who'd rather be anywhere else.

Your prompt: "Research what's on TV and streaming this Christmas Day. We've got kids aged 6-10 who love animation, grandparents who won't miss the King's Speech, teenagers who like comedy, and adults who want something everyone can watch together. I'll be busy cooking until about 2pm, then eating until 4pm. Find the highlights across BBC, ITV, Channel 4 and Netflix, and create a viewing schedule that gives everyone something they'll enjoy."

Why this matters: This is what "agentic AI" actually looks like in practice. You're not searching, clicking, and compiling - you're delegating. Comet navigates multiple sites, extracts the relevant information, and synthesises it into something useful. No passwords, no payments, no risk - just a glimpse of how browsing is about to change.

Notice:

• How many tabs would this have taken you manually?

• Where does Comet stumble? (It will - this tech is early)

• What other research tasks could you hand off this way?

👉 Download Comet (free, Chromium-based - works like Chrome)

📢 Share The Signal

If The AI Optimist helps you think clearer, forward it to someone else navigating the shift.

Not quite landing? Hit reply. We read every message.

Have a wonderful Christmas break. Let the big questions percolate. We'll see you in the new year.

Hugo & Ben