The cybernetic feedback loop, AI for the science of science, AI in strategy execution, and more

December 22, 2025

“Wherever output is graded by speed, accuracy, or cost, silicon will inherit the earth... Wherever value depends on provenance, uncertainty, and social risk, humans are irreplaceable.” – Scott Belsky

Thank you and wishing you a wonderful year ahead!

Thank you for being with us on the Humans + AI journey! It has been an extraordinary year. Yes, AI capabilities have soared. At the same time many more are understanding that we must frame how we develop and use the tools from a Humans + AI perspective. I have felt increasingly less lonely on this journey over the last three years.

Wishing you a wonderful rest of your festive season. Take the time to unplug and recharge. We need to be ready for a wild next year, where the Humans + AI mission will be even more important.

We’ll be back in mid January.

Be well!

Ross

 

📖In this issue

  • Framework: The Human-AI Feedback Loop

  • Humans + AI update: Humans-AI acceleration of science, faster exponentials on AI capabilities, and plunging prices.

  • Humans + AI Podcast: Lisa Carlin on AI in strategy execution, participative strategy, cultural intelligence, and AI’s impact on consulting

💡Framework: The Human-AI Feedback Loop

To round out the year, a framing of how human intent and AI execution can enable a better future. AI capabilities must always be in service of positive human intent. That will guide our work in the Humans + AI ecosystem in 2026.

🧠🤖Humans + AI

Human-AI collaboration to accelerate the Science of Science

"One key advantage of SciSciGPT is that it empowers human researchers at early, exploratory stages of the research process, rather than replacing their expertise or independently producing publication-ready outputs... SciSciGPT supports rapid prototyping, iterative idea refinement and exploratory analyses—tasks that traditionally involve substantial manual effort and time."

AI work capabilities leap above 7 month doubling rate

METR measures AI performance against the time taken for humans to complete tasks. Over the last 6 years this has doubled every 7 months. Claude Opus 4.5 has leaped above that trendline to successfully complete a 4 hour 49 minute task 50% of the time. The new shape of AI capabilities has powerful implications for Humans + AI work and organizational design.

The evolving state of LLMs and plunging prices

An excellent compilation of the latest data on the LLM market points to plunging LLM prices, exploding open source supply, continued enterprise dominance of commercial models, fast turnover in frontier models, and more.

🎙️This week’s podcast episode

Lisa Carlin on AI in strategy execution, participative strategy, cultural intelligence, and AI’s impact on consulting  

Why you should listen

Lisa Carlin combines deep and extensive experience in strategy and strategy execution with enthusiasm and expertise for applying AI to accelerate value for individuals and organizations. There’s a lot to learn from Lisa.

Thanks for reading!

Ross Dawson and team