LLMs are not the only show in town.

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.

📰 This was the week that was...

The start of a new year. Happy new year!

This was also the week AI stopped being just about ChatGPT. Whilst everyone's been focused on the big-name chatbots getting bigger and pricier, researchers have been building alternatives that work differently - and often better for specific jobs. A research team built a model with just 27 million parameters (tiny by AI standards) that beats billion-parameter systems at complex reasoning. That's like a small specialist outperforming a massive generalist - at a fraction of the cost. Models designed for long documents are running 5x faster than the current standard. And AI that can actually understand images is now working as well as text-only tools.

You wouldn't use a Ferrari to deliver parcels. Different tools for different jobs, right?! So, what's in it for you? Real choice about which AI to use for which problem, instead of defaulting to expensive general-purpose models for everything.

This is the foundation of Frugal AI the design practice that massively reduces the cost of AI both environmentally and financially.

Let's get into it.

🔥 Urgent Priorities

✅ No fires to fight this week ✅ The AI world is bigger than just chatbots ✅ Time to ask which tools fit which jobs in your business

This isn't a week for panic. It's a week for asking better questions about your AI choices.

🎯 Strategic Insight

Tension: It's tempting to think "AI" means "ChatGPT-style tools". That leads to using the same expensive general-purpose system for everything - image analysis, document processing, forecasting, reasoning - even when specialised tools would work better and cost less.

Optimistic insight: We're moving into an era of choice. Some AI architectures excel at understanding images. Some handle very long documents efficiently. Some achieve deeper reasoning with far fewer resources. Some simulate physical systems. Each solves different problems brilliantly - and your costs drop when you match the right tool to each job.

What's shifting: The smart question is no longer "Should we use AI?" but "Which type of AI fits this specific problem?" Analysing X-rays? You want AI built for vision, not a text system with image processing bolted on. Predicting equipment failures? You might not need generative AI at all - simpler, specialised models often work better whilst costing pennies per task.

Why this matters now: If you only plan for chatbot-style AI, you'll overspend on the wrong tools for half your problems. If you instead build knowledge of what each type does well, you get three benefits: lower costs, better results on specialised tasks, and real control over your AI choices.

👉 Takeaway: Between now and Q1 2026, list your current and planned AI use cases and ask:

• Which of these actually need text generation? (Probably fewer than you think)

• For image or document work, would vision-focused AI work better?

• For reasoning or calculations, would a specialised model outperform?

• What would we save over three years by matching each problem to the right tool?

🤓 Geek-Out Stories

1️⃣ Tiny reasoning model beats giants on complex puzzles

Researchers built a 27-million parameter model that outperformed billion-parameter systems on visual logic puzzles designed to test abstract reasoning. The difference? It's designed specifically for reasoning, using a brain-inspired approach with one part for high-level planning and another for detailed work. It learns from just 1,000 examples - no massive training runs needed.

Why it matters: Size isn't everything. For leaders, this means specialised AI built for specific reasoning tasks might solve complex problems better than expensive general-purpose systems - and cost far less to run.

👉 Action: Pick one reasoning-heavy task in your business (scheduling, optimisation, pattern-spotting). Ask your team: "Could a compact specialist model work better than our current approach?"

2️⃣ New AI designs handle long documents more efficiently

New AI architectures are processing long sequences of information much more efficiently than current designs. They're particularly good at tasks where context matters - analysing lengthy documents, tracking customer histories over years, processing continuous data streams.

Why it matters: If you're working with long documents or historical data, these newer designs deliver the same quality at a fraction of the computing cost. That's the difference between cloud bills that kill projects and tools you can actually afford to scale.

👉 Action: Identify one task involving long documents or time-series data. Test whether newer architectures would cut costs whilst maintaining quality.

3️⃣ AI that understands images is finally working

AI systems that process both images and text - called Vision Language Models - are now matching the performance of text-only tools. They can read charts, analyse medical scans, understand documents with diagrams, and interpret screenshots, all without needing complex workarounds.

Why it matters: If your work involves visual information - product images, scanned documents, quality control photos, technical drawings - these tools can do things text-only AI fundamentally cannot. Often more cheaply than the elaborate solutions people have been building.

👉 Action: Pick one visual task that currently needs human eyes (document checking, image analysis, quality inspection). Could vision-focused AI speed it up or automate it?

🎨 Weekend Playground

This weekend, try the Hugging Face Vision Arena, where you can test different AI tools side-by-side on the same image.

Upload a business chart, product photo, or scanned document. Ask each tool to describe it, extract data, or answer questions. See which one actually gets it right.

Why this matters: You'll quickly see that different AI systems understand images very differently. Some excel at reading text in images, others at spatial relationships, others at detailed descriptions. Understanding these differences helps you pick the right tool for your needs.

👉 Mission:

• Test 3 different tools on the same business image

• Compare how accurately they extract information

• Note which handles your specific case best

• Calculate what it would cost to process 1,000 similar items

• Ask: "Could this replace something we do manually?"

If The AI Optimist helps you think more clearly, forward it to someone else navigating the shift. If it's not quite landing, hit reply and let me know - I read every message.

Stay strategic, stay generous.

Hugo & Ben