Juicebox Podcast: Type 1 Diabetes

Episode #1767 Bolus 4 – Chicken Pot A.I. Pie
Date: February 10, 2026
Host: Scott Benner
Guest: Jenny Smith

Overview of Episode Theme

This Bolus 4 episode tackles the perennial diabetes challenge of bolusing for a complex, mixed-macro meal—in this case, homemade chicken pot pie. Host Scott Benner presents Real World, “in-the-moment” carb, fat, and protein estimation challenges to CDE Jenny Smith. Together, they demystify the math, strategies, and tech (including AI prompts and diabetes algorithms) involved in making confident insulin decisions for tricky recipes. The episode also explores the human and technological sides of diabetes management, highlighting both Jenny’s deep intuition and Scott’s experiments with AI-powered prompts to estimate bolusing needs for complicated meals.

Key Discussion Points & Insights

1. The Meal Bolt Framework (00:00–02:16)

• Scott introduces the “Meal Bolt” roadmap for food bolusing decisions:

  • Measure the meal
  • Evaluate yourself
  • Add the base units
  • Layer a correction
  • Build the bolus shape
  • Offset the timing
  • Look at the CGM
  • Tweak for next time

• The approach is meant to demystify and make real-world bolusing for specific foods more approachable.

2. The Challenge: Chicken Pot Pie Recipe Breakdown (02:16–09:36)

• Scott texts Jenny a chicken pot pie recipe with minimal info and parameters:

  1. Insulin-to-carb ratio: 1:10
  2. Sensitivity: 50
  3. Target BG: 100
  4. Default (50% for FPU calculation)
  5. Current BG: 130, rising arrow
  6. No insulin on board

• Jenny clicks through a busy, ad-filled recipe site live, demonstrating the mental steps of meal estimation:

  • Identifies 8 servings per pie
  • Highlights importance of reading full ingredient list (including double crust and high-fat content)
  • Estimates carbs/fat/protein per serving just from raw ingredients

• Notable quote:
  “You always want to know how many servings a recipe might make so that you know what a single serving, not what the portion is that you put on your plate.” (Jenny, 05:39)

3. Detailed Bolus Calculations & Justifications (09:36–25:02)

Estimating Macros & Impact

• Jenny walks through rough macro calculations (in her head!) for 1/8 of the pie:

  • Carbs: Estimates ~35g (crust, veggies, flour)
  • Fat: Estimates ~30g (butter, shortening, milk, egg)
  • Protein: Estimates ~20g (mainly chicken)
  • Discusses that this is a high-fat meal; a classic candidate for an extended/double-wave bolus

• Bolus Strategy:

  • Upfront Bolus:

    • 35g carbs at 1:10 = 3.5 units
    • Correction for 130 BG (target 100, sensitivity 50): (130–100)/50 = 0.6 units
    • Total upfront: 4.1 units

  • Extended Bolus (FPUs):

    • Explains Warsaw Method math: calculates combined calories from fat & protein; converts to “carb equivalent”
    • Arrives at another 3.5 units to deliver over 4–5 hours

• Notable quote:
  “I would expect to need an extended bolus for this because the real carbs in here that are the hit carbs are coming from the pie crust, which is high fat.” (Jenny, 08:11)

4. Validation Against AI & Algorithmic Tools (25:35–39:14)

• Scott reveals he has been testing an AI prompt, fed with Juicebox Podcast logic and the Warsaw Method.

  • The AI breaks down the same recipe, returning macro and insulin numbers nearly identical to Jenny’s head-math:
    • AI: 38g carbs, 18.5g protein, 30g fat ⟶ 4.4 upfront, 3.26 extended, 7.65 total units (compared to Jenny’s total of 7.6).

• They discuss potential pitfalls—AI initially calculated for 1/6 of the pie, not 1/8, overshooting insulin.

• Notable moment (31:57):
  Scott’s excitement— “Look how I’m going to curse, look how close you were on all that. That’s amazing.” (Scott, 30:58)

• Takeaway: even with powerful AI, logic and context from real-world wisdom are vital. Macro misestimation (e.g., wrong serving size) quickly leads to errors, even for “smart” tools.

• Jenny’s advice: Always double-check servings and ingredients, even with smart estimators.

5. Usability, Real-Life Application, and “Being Bold” (35:31–41:06)

• Recognizing reality: Few people will break down meals this carefully every time; many will “just guess,” risking swings.

• When to use AI/algorithmic support:

  • Complicated or restaurant meals
  • Recurring, frequently-mis-bolused family meals
  • Storing and reusing strategies for common dishes (write sticky notes in cookbooks!)
  • Better macro calculations than just “guesstimating carbs”

• Notable quote:
  “If you do this for 20 meals that you guys favor over and over, you’re probably going to hit a good 80% of your management, especially after dinnertime meals.” (Jenny, 38:07)

6. The Tech & The Human: Role of Intuition + New Tools (41:06–53:34)

• Scott explains he’s using AI to build embeddable bolus & basal calculators for his website (hidden for now):

  • Basal insulin by weight/insulin sensitivity/carb ratios
  • Insulin sensitivity factor calculators
  • “Strategy engine” for complex meals, taking custom inputs and producing tailored bolus strategies
  • These calculators mirror the podcast’s “mental model,” turning conversational logic into instantly usable outputs

• The greater point:
  Even if you’re not a math nerd or medical pro, AI/tech can help anyone estimate and deliver safer, more effective boluses.

• Notable exchange:
  Scott: “You should all be paying attention to what we just talked about for the last half hour. The future is now.” (43:17)

7. Final Thoughts & Takeaways (53:34–End)

• Jenny and Scott note frustration that big device manufacturers haven’t yet wrapped these kinds of tools—step-by-step, practical meal-insulin calculators—into standard pump interfaces.

• Empowerment message:
  You don’t have to be Jenny, or even understand all the math, to improve diabetes management if you leverage smart strategies and/or next-gen tools.

• Notable quote:
  “My baseline is guy trying to help his daughter…I only know this from talking to Jenny for all these years…Now I woke up in a world two years ago and I was like, hey, I type into a prompt and it’s…almost perfect.” (Scott, 51:16)

Notable Quotes & Memorable Moments

• On estimation skill:
  “You always want to know how many servings a recipe might make so that you know what a single serving, not what the portion is that you put on your plate.” (Jenny, 05:39)

• On high-fat, high-protein bolusing:
  “I would expect to need an extended bolus for this because the real carbs in here that are the hit carbs are coming from the pie crust, which is high fat.” (Jenny, 08:11)

• On AI validation:
  “Look how I’m going to curse, look how close you were on all that. That’s amazing.” (Scott, 30:58)

• On practical use:
  “Put sticky markers or notes in the sides of your recipes that say, this is exactly how to do this meal. Is it going to be 100% every time? No…” (Jenny, 37:50)

• On flattening the learning curve:
  “The future is now. This is not...rocket science.” (Scott, 43:17 / Jenny, 52:08)

Timestamps for Important Segments

• 00:00 – Introduction, Meal Bolt decision map explained
• 02:16 – The pot pie challenge and live recipe breakdown
• 06:21 – Jenny walks step-by-step through macro and bolus estimate
• 14:54 – Blood sugar correction math and “what if you do nothing”
• 18:16 – Extended bolus math (Warsaw Method, FPU)
• 25:35 – Scott reveals AI prompt and its validation vs. Jenny’s math
• 31:57 – Mind-blowing accuracy of AI versus Jenny’s estimation
• 35:31 – When to use such tools, the value of logging favorite meals
• 41:06 – Scott previews AI-based practical calculators in development
• 53:34 – Why this should be part of every pump, and future hope

Final Takeaways

• Bolusing for complex, high-fat/protein meals is hard—but learnable and repeatable with structure and/or tech support.
• AI-powered tools can now estimate meal breakdowns and optimal bolusing nearly as accurately as a world-class educator.
• You don’t have to be “good at math”—you can use tech and frameworks to help (just always double-check serving sizes and ingredient lists!)
• Practical, evidence-based approaches (“bold with insulin”) help remove the fear and guesswork from T1D meal management.
• The future of diabetes is collaborative: your lived experience, your educator’s wisdom, and smart algorithms working together.