I was invited by @PaulBLaursen to talk about our new paper on his HIIT Science Podcast. It was a very enjoyable and engaging conversation covering elite Ethiopian distance runners and their training practices, AI, LLMs and their role in sports science research, the nature of intelligence, the importance of context, and we even discussed dogs.
Thank you Paul for inviting me and persevering through an hour of my ramblings 😊. Also huge thanks to my co-authors @StephenSeiler , Thomas Haugen, Abay Yisamaw for allowing me to contribute to this amazing study.
YouTube:
youtube.com/watch?v=B73zPUuI…
Paper:
medrxiv.org/content/10.64898…
One of my absolute favorite podcast interviews. Dr. Ranjan Chatterjee was so well prepared. This podcast has the added bonus of being long enough for a green zone training session!
youtu.be/hbfMVRunyGA?si=OENO…
Sufficiently advanced agentic coding is essentially machine learning: the engineer sets up the optimization goal as well as some constraints on the search space (the spec and its tests), then an optimization process (coding agents) iterates until the goal is reached.
The result is a blackbox model (the generated codebase): an artifact that performs the task, that you deploy without ever inspecting its internal logic, just as we ignore individual weights in a neural network.
This implies that all classic issues encountered in ML will soon become problems for agentic coding: overfitting to the spec, Clever Hans shortcuts that don't generalize outside the tests, data leakage, concept drift, etc.
I would also ask: what will be the Keras of agentic coding? What will be the optimal set of high-level abstractions that allow humans to steer codebase 'training' with minimal cognitive overhead?
This paper answers why do living organisms sleep?
And it destroys everything we thought we knew about the brain.
For decades, the consensus was simple. Sleep is a cognitive tool.
We sleep to consolidate memories. To clear out neural toxins. To rest our massive, energy-hungry brains.
But this paper proved that is a fundamental misunderstanding of life itself.
Researchers started studying creatures at the absolute bottom of the evolutionary tree.
Jellyfish. Hydras. Organisms with absolutely no brain. No central nervous system. Just a loose net of cells.
They shouldn't need to sleep.
But they do.
They enter the exact same state of vulnerable, unresponsive suspension that we do.
If a creature with no brain requires sleep, then sleep isn't a cognitive function.
It is a cellular mandate.
The researchers realized sleep didn't evolve to protect the brain. Sleep existed hundreds of millions of years before the brain was even invented.
It is a fundamental metabolic repair cycle. A reset switch built into the physics of living tissue.
Without it, the cellular machinery literally tears itself apart.
We live in a culture obsessed with biohacking. We treat sleep like an inefficiency. A productivity flaw we can optimize away with caffeine, wearables, and willpower.
But this paper is a blunt reality check.
You can't outwork biology.
Sleep isn't an evolutionary luxury. It is the baseline condition for staying alive.
Well, we got there in the end. Thank you @SophieRHerzog for coming along on this thought experiment, exploring the complexities and pitfalls of trying to predict the future.
Palo Galko @pgalko Nov 13 Article I Just Wanna Dance When we think about the future, dancing probably isn’t the first thing that comes to mind. But the more we explored the topic, the more compelling that unlikely connection became. This essay grew out
OK, I finally got around to it and built what I promised. BambooAI 2.0 is now out in the open 🤗 .
I can trace the first seed of the idea back to late 2020, when I first experimented with GPT-3 through the OpenAI API. It returned a snippet of code to calculate the Fibonacci sequence. The code didn't work, but the seed was planted and I was hooked.
What followed a little later on, started as a personal project to help me make sense of an ever growing pile of sports and physiological data I'd been collecting over the years. Over time it evolved into a more general data analysis workbench, although sports science is still very much its home and where most of my ideas and hunches come from.
GenAI models were pretty good at writing code and putting the pieces together, but I often found myself puzzled, looking at the results and pretty charts, and often just a wall of text and wondering, "How on earth did we get here?" I wanted to grow as an analyst too, rather than just blindly trust what came out.
And being naturally curious, I kept wanting to be able to peek under the hood and understand what was actually going on.
That's in essence what I've been working towards with Bamboo. I wanted a tool to help me explore and interrogate data with AI, but also see what it's doing, question its decisions and have the freedom to experiment and steer. Including choosing and tune my own models, whether commercial or open-weight models running locally on my hardware (this bit is important to me).
There's plenty left to figure out, particularly how to assess whether an analysis is genuinely sound, rather than simply plausible and reproducible. But at least now I can see what I'm arguing with. :-)
There have been a few special people that influenced its course over time (knowingly or unknowingly), and I would like to express my gratitude for your thoughts and ideas: @aareskjold , @StephenSeiler , Olav Aleksander Bu, @KittyJohne , @SophieRHerzog , @AartGoossens , @PaulBLaursen , @david_tinker , Paolo Fondin, Thomas haugen, Abay Yisamaw and many others. Thank you!
It's open source. Link below, if you feel like having a peek under its hood 😉.
github.com/pgalko/BambooAI
Humble beginnings 😊. I can't believe that it has been almost 4 years since I started working on Bamboo, back when GPT-3.5 was SOTA, Code Interpreter was yet to be born, and only a few people were talking about AI agents... It is a somewhat different world today.
Just launched BambooAI - a lightweight library for non-programmers to analyze and visualize data using natural language! Total of 126 lines of code. Uses @OpenAI chat models to generate the code for analysis and visualizations. Inspired by @yoheinakajima BabyAGI. Conversational.
Are we living through a new Renaissance period of our human story ? The structural dynamics look very similar to 15th-century Italy. I will draw some parallels below and let you judge for yourselves.
- The Knowledge Shock:
The Printing Press then.
Generative AI now.
A total disruption of how human knowledge is stored and shared
- The Prelude Shock:
The Black Death (1348) then.
The COVID-19 Pandemic (2020) now.
A massive global crisis that shatters trust in legacy institutions, disrupts traditional labor markets, and acts as the ultimate catalyst for a total civilizational reset.
- The Epicenter:
Florence then.
Silicon Valley now.
Hyper-concentrated geographical pockets where immense talent and capital collide.
- The Merchant Princes:
The Medici, Sforza, and Borgia families then.
The ~12 AI tech leaders now.
A microscopic oligarchy controlling the expensive infrastructure of the future.
- The Artisans:
Michelangelo, Leonardo, and Galileo then.
Alec Radford, Andrej Karpathy, and Ilya Sutskever now.
Generational minds quietly obsessed with redefining human capability.
- The Fanatics:
Girolamo Savonarola then.
AI existential doomers now.
The moral backlash demanding we burn the progress down to avoid corruption.
- The Pragmatists:
Niccolò Machiavelli then.
Peter Thiel, Elon Musk, Eric Schmidt now.
The technocrats mapping raw power and shaping state infrastructure.
- The Institutional Fracture:
Ultra-corrupt 15th-century Popes then.
Trump and Putin now.
Hyper-powerful, unpredictable leaders leveraging a world in total flux.
So if you agree that the ingredients are strikingly similar, the next question is “What followed ?” And how does it extrapolate to our modern times ?
Thanks for the seed @SophieRHerzog