Syntax and Science

Since starting my ketogenic journey on September 18, I've consistently maintained ketosis, confirmed by daily finger-prick testing. Along the way, I've encountered the expected challenges—especially, a heavy, sluggish feeling in my legs during uphill runs and a lack of explosiveness in intense workouts.

Rather than rushing to tweak my nutrition, I gave my body time to fully transition to fat-burning mode. Now, with a solid foundation in place, I'm beginning a deliberate experiment: targeted carb intake during exercise.

Why Experiment with Carbs on Keto?

While fat adaptation offers remarkable benefits—consistent energy, better sleep, weight loss, and improved insulin sensitivity—it has a downside: reduced explosive power during high-intensity exercise. Carbohydrates remain the body's preferred fuel source for these intense efforts.

I recently heard David Roche—a renowned endurance coach and athlete—share his carb intake on the Some Work, All Play podcast: Up to 150 grams per hour—a stark contrast to my 0 grams! Around the same time, KoopCast showed up in my feed. In this podcast, Jason Koop and his guest discussed a study on carb intake during competition. Athletes consumed 60, 90, or 120 grams per hour, with the 120-gram group achieving slightly better performance and significantly better recovery.

This is a big deal. Improved recovery means less muscle damage, faster progress, and higher levels of performance.

First Test: Gu and Ninja

This weekend, I tore open a Gu and thought, "Let's see what happens." My local ninja obstacle-course gym served as a fitting testing ground, as these workouts take both strength and endurance.

Before the workout, I consumed a single gel, delivering 22g of carbs—a significant portion of my daily carb budget (25-30g on keto). For keto purists, this would blow past the 20g limit for the entire day. However, the theory is simple: during intense workouts, your body rapidly burns through those carbs, allowing you to return to ketosis soon afterward.

A line graph showing ketone and glucose levels throughout the day, with ketones represented by a blue line and glucose by an orange line. The graph highlights a Gu supplement ingestion point at around 9:30 am, containing 22 mg of carbs and 40 mg of caffeine. The ketones decrease steadily, while glucose levels rise after Gu ingestion and continue to fluctuate. A shaded region from approximately 12 pm to 1 pm indicates a workout period. The ketone threshold is marked as a dashed line across the graph.

Ninja Training (Saturday)

Protocol: 22g carbs
Duration: 1 hr
Pre: Ketones 0.5, BG 111
Post: Ketones 0.5, BG 121
1hr Post: Ketones 0.5, BG 116
2hr Post: 0.8, BG 93

This experiment wasn't about nailing the perfect carb amount—it was about testing whether one little gel could supercharge my workout without kicking me out of ketosis. The results were encouraging: I stayed in ketosis, and I felt stronger! Next up: testing this approach with different workouts.

A Week of Workouts

Building on my initial success, I tested carb intake across a variety of workouts throughout the week:

Speed Intervals (Monday)

Protocol: 22g carbs at start
Duration: 32 min (1-min on, 1-min off)
Pre: Ketones 0.9, BG 83
Post: Ketones 0.4, BG 144
1hr Post: Ketones 0.7, BG 91

Ninja Training (Wednesday)

Protocol: 2x22g carbs, 40min apart
Duration: 1 hr
Pre: Ketones 0.6, BG 103
Post: Ketones 0.2, BG 146
1hr Post: Ketones 0.4, BG 103
2hr Post: 0.6, BG 105

Tempo Run (Thursday)

Protocol: 2x22g carbs, 30min apart
Duration: 40 min (10-min warmup, 20-min tempo, 10-min cooldown)
Pre: Ketones 0.8, BG 101
Post: Ketones 0.4, BG 179
1hr Post: Ketones 0.6, BG 110

Zone 2 Run (Friday)

Protocol: 22g carbs every 20min (88g total)
Duration: 80 min
Pre: Ketones 0.8, BG 80
Post: Ketones 0.5, BG 183
1hr Post-run: Ketones 0.8, BG 99

Each time, my ketones recovered within an hour or two, and my blood glucose (BG) returned to reasonable levels. And with today's experiment, it's clear I haven't hit the ceiling yet.

What I've Learned (And Where I'm Headed)

These experiments are just the beginning. Two key effects I'm aiming to optimize are:

Performance: The energy boost from carbs is undeniable. For example, on both the first day of keto (9/18) and the first day I took the Gu (11/9), I scaled the 14.5-foot warped wall-—something I couldn't manage during the weeks in between.
Recovery: By fueling my body during exercise, I believe I am reducing muscle damage, potentially lowering the risk of injury and boosting long-term performance.

My goal is to strike the right balance between carb intake and maintaining ketosis. I'm not aiming for surgical precision—just a sweet spot that maximizes my performance within the bounds of ketosis.

A forked path in a field under a dramatic sky. One path leads into a dark, stormy landscape, while the other path is bathed in warm sunlight. A rainbow bridges the two paths, symbolizing contrast and potential for change. This image represents a choice between two diverging futures.

I’ve been using large language models since just a few days after ChatGPT was first released. From the start, it was clear to me that these tools would change how we work. I remember writing a Web Crypto API implementation of Fernet encryption 1 in a couple of days—a task that would likely only take a few minutes with today’s models. At that time, I wasn’t even sure I could fully grasp the concepts without several weeks of study, if at all. Needless to say, I became an evangelist.

But as these tools have become more accessible, I’ve watched an unfortunate, yet inevitable trend emerge. Bad actors are flooding the world with low-quality content and automated noise, sidelining the thoughtful and impactful uses that these models make possible. It’s disheartening to see something so promising being abused, diluting what should be genuine innovation.

A few weeks ago, a friend gave a talk that sparked a new sense of urgency in me. The topic was “100x Engineers,” a play on the “10x engineer” concept, using AI tools to radically speed up development. During the session, we primarily used a tool called Cursor to build an email client connected to Supabase—in an hour. It wasn’t perfect, but the potential was undeniable. After seeing it in action, I took a deeper dive into Cursor and, within a few days, built an iOS app in SwiftUI—without any prior SwiftUI experience.

He wrapped up his talk with an idea that has stayed with me: using AI to amplify our own abilities thoughtfully. AI isn’t inherently good or bad—it’s a tool, and like any tool, it reflects the intentions of the person using it. As he pointed out, these tools can easily “spam the world with garbage.” His challenge to us was clear: “amplify our humanity.” That phrase has stuck with me. It’s a powerful reminder that while AI can accelerate our work, we need to be thoughtful about using it in ways that are impactful and truly human.

With Syntax and Science, I want to create content that’s assisted by AI but never defined by it. This blog is my way of adding not just more AI-generated noise but something meaningful and human to the conversation. Here, I’ll be coding real experiments, diving into science and tech, and hopefully making these topics approachable and engaging. It’s about using technology to expand our understanding of the world and sharing that journey with others.

1. I plan to write a blog about this in the future.