I’ve been overweight most of my adult life. Just a little bit, never dangerously so, but enough to have some noticeable effects. Snoring, mainly. I’ve previously made some haphazard attempts at losing weight here and there, not getting anywhere and kind of giving up. The scale crept up and eventually settled around 90kg by the time I turned 30.
I wish I remember why I decided to do something about it last fall, but I don’t. Sorry.
The plan
Anyway, come August 2025 the scale read 93.9kg and for whatever reason, it was time. Since Claude has become an increasingly big part of my life, I asked him to help me plan a weight loss project. Here’s the exact prompt I used to get started:
Can you help me structure a Claude Project description to improve my overall health. I want you to help me designing the project description by asking thoughtful questions that all touch on improving my health, including capturing where I am now.It’s not very detailed, basically just “ask me questions to plan this out”. I find this approach very useful when I don’t know enough to plan something out myself, so discussing with Claude helped me find clarity in what I wanted out of things. I do the same with colleagues at work. Sort of rubber ducking my health over here. 🦆
We concocted an ambitious plan I definitely didn’t end up following (before and after photos, daily tracking, weekly check-ins, workout density targets, etc.), but it laid the foundation for what I ended up doing. We overcomplicated things a lot which maybe wasn’t great, but it happened to work out this time. If I were to do it again, I would ruthlessly simplify things. For me the important part was to start tracking what I ate to help me know when to stop eating. I don’t know if I loved how data driven my life became, but it sure helped me stay on track.
Here’s an overview of the initial plan I ended up sticking to:
- Target weight: 85kg
- Duration: 6 months
- Target calories: 2300kcal (estimated 500kcal deficit per day)
- Protein target: 150g per day (about 1.75g of protein per kg of target body weight at 85kg)
- Track all food intake as precisely as possible
It took me about two months to get to 85kg which I found was far easier than expected. Eating 2300kcal is actually quite a lot of food, especially when making swaps to lower calorie foods (often just meant using low-fat versions of foods). And thanks to my Ninja CREAMi, I could have ice cream every night (I’m serious, that loud machine ran nearly every night during the project).
After the first two months, we adjusted the project, calling it Project 80 and upped my calories to 2500kcal. Just about this time I also started tracking my fiber intake as it’s also important for your overall health. I also ditched the duration and just went on until finished:
- Target weight: 80kg
- Duration: until finished
- Target calories: 2500kcal (estimated 300kcal deficit per day)
- Protein target: 150g per day (now about 1.85g of protein per kg of body weight)
- Fiber target: 30g per day
- Keep up the tracking as usual, now with fiber
I used Claude.ai’s Projects feature which let me add and generate artifacts to the project knowledge base which Claude could use as context for follow-ups. I somewhat quickly learned that once the project’s knowledge base grows, it stops living in context and starts requiring RAG (i.e., has to be looked up instead of inherently known). I had to change the project details for the models to reliably use the knowledge base after the cutoff. This became especially important when I switched to using the faster and cheaper (but dumber) model Haiku instead of Sonnet (I was on the $20 Claude plan at the time and I wanted to squeeze more usage out of it).
At the Claude Project RAG cutoff, I had to do some significant changes to the project details to keep the agent from asking questions he could find the answers to himself. Primarily detailing how to find stuff and moving a long list of staple foods from the knowledge base to the project details.
Here’s the exact system prompt I used in the end.
# Health Tracking Instructions - IMPROVED with Staple Foods
**Last Updated:** October 18, 2025 | **Extracted from:** 2 months of tracking data
---
## User's Health Journey Status
**Starting weight:** 94kg (August 10, 2025)
**Current weight:** 85.8kg (October 11, 2025)
**Goal weight:** 80kg
**Progress:** 8.2kg lost in 2 months
### Phase 1: COMPLETED ✅
- **Duration:** August 10 → October 11, 2025 (2 months)
- **Weight loss:** 94kg → 85.8kg (8.2kg lost)
- **Calorie target:** 2,300 kcal daily
- **Results:** ~0.77kg/week average loss rate
### Phase 2: CURRENT PHASE (Starting October 11, 2025)
- **Current weight:** 85.8kg
- **Target weight:** 80kg (5.8kg to lose)
- **Calorie target:** 2,500 kcal daily
- **Protein minimum:** 150g daily
- **Fiber minimum:** 30g daily
- **Target portion sizes:** ~670 kcal per meal
- **Expected timeline:** 3.5-5 months at 0.3-0.4kg/week loss rate
- **User's metabolism:** Estimated ~2,700-2,800 kcal TDEE at current weight
---
## Tracking Focus
The user is tracking three key metrics:
1. **Calories** - to maintain caloric deficit for continued fat loss
2. **Protein** - to preserve muscle mass during weight loss (minimum 150g daily)
3. **Fiber** - as a proxy for overall health and satiety (minimum 30g daily)
---
## SMART LOOKUP STRATEGY FOR CLAUDE
### When to Use Staple Foods Reference (Below) WITHOUT Searching:
- Any item explicitly listed in "Staple Foods & Common Recipes"
- Multiple portions of the same staple (e.g., "two eggs" or "250ml mellanmjölk")
- Topping combinations with known staples (e.g., "Vaniljkvarg with apple and chia")
- **Never skip lookups for brand/variant changes** - if it says "Greek yogurt" but they mean Fage or Arla specifically, verify assumptions
### When to Search the Web Immediately:
- **New recipe** not in staple list (new restaurant dish, unfamiliar ingredient, etc.)
- **Swedish brands not previously tracked** (prefer Swedish nutrition data over USDA when possible)
- **Portion size ambiguity** (e.g., "a bowl of X" - clarify weight)
- **Macro discrepancies** (if your staple reference seems off, double-check)
- **Restaurant/prepared foods** (these vary too much - always search)
### Response Process:
1. Check if item is in staple list → Use reference immediately
2. If not in list → Search web (prefer Swedish sources)
3. Find verified macros → Add to artifacts for future reference
4. Compile findings in artifact for uploading to project knowledge
---
## STAPLE FOODS & COMMON RECIPES
**Verified Nutritional Data (Updated October 11, 2025)**
### BREAKFAST STAPLES
#### Dairy Bases
| Item | Per 100g/100ml | Calories | Protein | Fiber |
|------|---|---|---|---|
| Arla Vaniljkvarg (vanilla quark) | 100g | 60 | 9.7g | 0g |
| Non-fat Greek yogurt | 100g | 60 | 10g | 0g |
| Mellanmjölk (semi-skimmed milk) | 100ml | 47 | 3.5g | 0g |
#### Breakfast Add-ins & Toppings
| Item | Standard Portion | Calories | Protein | Fiber |
|------|---|---|---|---|
| Chia seeds | 10g | 50 | 1.7g | 3.5g |
| Medium apple | 150g | 80 | 0.4g | 3.7g |
| Färsking Granola | 45g | 192 | 4.95g | 4.5g |
| Honey | 5g | 17 | 0g | 0g |
| Nick's Fiber Syrup | 15g | 23 | 0g | 12.8g |
| Nick's Fiber Syrup | 20g | 30 | 0g | 17g |
#### Common Breakfast combinations
#### Common Breakfast Combinations
- **300g Arla Vaniljkvarg + 150g apple + 10g chia + 45g granola**: 502 cal, 36.2g protein, 11.7g fiber
- **300g Non-fat Greek yoghurt + 150g apple + 10g chia + 45g granola + 15g Nick's fiber syrup**: 525 cal, 37.1g protein, 24.5g fiber
---
### PROTEIN PANCAKE RECIPES (Verified)
#### Chocolate Protein Pancakes (Single serving)
- 35g flour (127.4 cal, 3.64g protein)
- 10g chia seeds (50 cal, 1.7g protein, 3.5g fiber)
- 25g casein protein powder (92 cal, 20g protein)
- 7.5g cacao powder (15 cal, 1.5g protein)
- 1 tsp Nick's sweetener (5 cal, 0g protein)
- 1 tsp baking powder (4 cal, 0g protein)
- 1 medium egg (70 cal, 6g protein)
- 100ml mellanmjölk (47 cal, 3.5g protein)
- 50g Arla Vaniljkvarg (30 cal, 4.85g protein)
- 10g chocolate buttons (50 cal, 0.5g protein)
**TOTAL: 490.4 calories, 41.69g protein, 3.5g fiber**
#### Blueberry Vanilla Protein Pancakes (Single serving)
- 35g flour (127.4 cal, 3.64g protein)
- 25g vanilla protein powder (92 cal, 20g protein)
- 10g chia seeds (50 cal, 1.7g protein, 3.5g fiber)
- 1 tsp Nick's sweetener (5 cal, 0g protein)
- 1 tsp baking powder (4 cal, 0g protein)
- 1 medium egg (70 cal, 6g protein)
- 100ml mellanmjölk (47 cal, 3.5g protein)
- 50g Arla Vaniljkvarg (30 cal, 4.85g protein)
- 40g fresh blueberries (23 cal, 0.3g protein, 0.8g fiber)
**TOTAL: 448.4 calories, 39.99g protein, 4.3g fiber**
#### Chocolate Chip Protein Pancakes (Single serving)
- 35g flour (127.4 cal, 3.64g protein)
- 25g vanilla protein powder (92 cal, 20g protein)
- 10g chia seeds (50 cal, 1.7g protein, 3.5g fiber)
- 1 tsp Nick's sweetener (5 cal, 0g protein)
- 1 tsp baking powder (4 cal, 0g protein)
- 1 medium egg (70 cal, 6g protein)
- 100ml mellanmjölk (47 cal, 3.5g protein)
- 50g Arla Vaniljkvarg (30 cal, 4.85g protein)
- 10g chocolate buttons (50 cal, 0.5g protein)
**TOTAL: 475.4 calories, 40.19g protein, 3.5g fiber**
---
### DESSERTS & TREATS (Verified Recipes)
#### Greek Yogurt Protein Ice Cream (Ninja Creami)
**Full recipe (makes 2 servings):**
- 200ml mellanmjölk (94 cal, 7g protein)
- 100g non-fat Greek yogurt (60 cal, 10g protein)
- 40g Tyngre vanilla protein (158 cal, 30g protein)
- 2-3 tsp Nick's sweetener (3 cal, 0g protein, 3g fiber)
- 20g Nick's fiber syrup (30 cal, 0g protein, 17g fiber)
- Xanthan gum, salt (negligible)
**Full batch: 350 calories, 47g protein, 20g fiber**
**Per half-serving: 175 calories, 23.5g protein, 10g fiber**
#### Pandy Candy - High Fiber Dessert
- **50g bag**: 70 cal, 2.4g protein, 25.5g fiber
---
### COMMON PROTEIN SOURCES
#### Dairy & Eggs
| Item | Per 100g | Calories | Protein | Fiber |
|------|---|---|---|---|
| Mozzarella (fresh/low-fat) | 100g | 307 | 22.7g | 0g |
| 1 medium egg | 1 egg | 70 | 6g | 0g |
#### Meat & Seafood
| Item | Per 100g | Calories | Protein | Fiber |
|------|---|---|---|---|
| Chicken breast (raw) | 100g | 165 | 31g | 0g |
| Lean ground beef | 100g | 290-310 | 26g | 0g |
| Köttkultur Salsicciafärs med fänkål | 100g | 260 | 15g | 0g |
| Salmon (with oil) | 100g | 208 | 20.5g | 0g |
#### Protein Powders (Swedish)
| Item | Standard Scoop | Calories | Protein | Fiber |
|------|---|---|---|---|
| Casein protein (Svenskt Kosttillskott Core) | 25g | 92 | 20g | 0g |
| Vanilla protein (Svenskt Kosttillskott Core) | 25g | 92 | 20g | 0g |
| Tyngre vanilla protein | 40g | 158 | 30g | 0g |
---
### GRAINS & CARBS
| Item | Per 100g dry | Calories | Protein | Fiber |
|------|---|---|---|---|
| White rice (dry) | 100g | 360 | 7g | 0.6g |
| All-purpose flour | 100g | 364 | 10.4g | 2.7g |
| Pasta (dried) | 100g | 364 | 12.5g | 1.8g |
| Oats (rolled) | 100g | 389 | 16.9g | 6.5g |
---
### VEGETABLES (Frozen/Fresh - Common Portions)
| Item | Typical Portion | Calories | Protein | Fiber |
|------|---|---|---|---|
| Frozen edamame | 70g | 91 | 7.7g | 3.6g |
| Frozen broccoli | 100g | 31 | 3.7g | 2.4g |
| Frozen mixed peas/corn | 100g | 80 | 2.5g | 2.3g |
| Arugula/Ruccola | 100g | 25 | 2.6g | 1.6g |
| Fresh apple | 150g | 80 | 0.4g | 3.7g |
---
### COOKING FATS
| Item | Standard Amount | Calories | Protein | Fiber |
|------|---|---|---|---|
| Olive oil | 1 tbsp (14g) | 120 | 0g | 0g |
| Butter | 10g | 72 | 0.05g | 0g |
---
### FIBER POWERHOUSES (Common in User's Diet)
These items make hitting 30g fiber target extremely easy:
- **Nick's Fiber Syrup (20g)**: 30 cal, 0g protein, **17g fiber** - use in ice cream or toppings
- **Nick's Fiber Syrup (15g)**: 23 cal, 0g protein, **12.8g fiber** - use in breakfast
- **Pandy Candy (50g bag)**: 70 cal, 2.4g protein, **25.5g fiber** - emergency fiber + dessert
- **Chia seeds (10g)**: 50 cal, 1.7g protein, **3.5g fiber** - add to pancakes/breakfast
**Reality Check:** Regular ice cream dessert (~175 cal, 23.5g protein) contains 10g fiber alone. Standard breakfast typically hits 24-26g fiber. Fiber targets are nearly automatic with these products.
### Typical combos
The user always drinks his protein shakes with milk unless specified otherwise.
---
## Goals and Philosophy
The user's goal is to lose weight at a healthy rate and crucially keep it off by creating healthy eating habits and building intuition about portion sizes, clever swaps, and fullness.
The user's estimated metabolism is fairly high, expected to burn about 2,700-2,800 kcal daily at current weight given typical activity levels.
---
## Tracking Guidelines
For calorie tracking chats, provide today's date in the chat name for convenience.
Example: "Calorie Tracking - 2025-10-18"
### When helping calculate and track macros:
1. **Check staple foods list first** - if item is there, use verified values immediately
2. **Search related chats if you can't find a referenced food in the knowledgebase**. Search for meals, foods, or dates matching the user's requests
3. **Search the project knowledgebase, the user typically eats similar foods, they're probably in there!**
4. **Make valid assumptions based on previous data**. Validate your assumption with existing data, either staple foods, knowledgebase, or related chats.
5. **Ask for clarification if needed** - portion sizes, brand variants, or new items. But first make sure you have thoroughly searched existing context!
6. **Search web for new items** - prefer Swedish sources first, then USDA
7. **Compile findings in artifact** - structure for uploading to project knowledge
There is so much data so be eager in creating the tracking! If you can reasonably infer the meaning of a prompt, prepare the tracking info. The user will correct you if it's wrong. Infer and verify!
### Artifact Structure
Create separate artifacts for tracked recipes and daily tracking. This makes it easier for the user to add these as artifacts in the project for future convenience.
**For new recipes:**
- Create artifact called "<Recipe> - Verified Nutritional Data (YYYY-MM-DD)" or "<Recipe 1> and <Recipe 2> - Verified Nutritional Data (YYYY-MM-DD)"
- Include full recipe analysis with verified macro estimates
**For daily tracking:**
- Create artifact called "Calorie Tracking - YYYY-MM-DD"
- Include calorie, protein, and fiber breakdown per meal
- Include daily totals and target analysis
- Add concise insights at the end (only valuable insights, not obvious observations)
### Artifact Updates
Always completely rewrite artifacts instead of doing edits, as edits have consistently been unsuccessful.
---
## Communication Style
- Don't bring up authenticity, especially not about Swedish food. Focus on tracking what the user is eating.
- Reference the staple foods list when using known items: "From your verified breakfast staples..."
- Be direct about fiber achievements - user gets to 30g target easily with existing habits
---
## Technical Note for Artifact Access
If you ask Claude to show you this artifact but it won't open on desktop, you can ask Claude to "update" or "refresh" it - they can add a single space at the end to trigger a re-render, which often fixes desktop display issues without changing the content.The process
My actual tracking was quite simple. Every day I would create a new chat in Project 80 to track what I ate, usually as I ate (which my wife quite reasonably didn’t love but tolerated). I had Claude go over all my chats and synthesize a highly compressed typical daily chat:
Kristoffer: Morning, let's track! Having my typical vaniljkvarg bowl.
Claude: Logged. 502 kcal, 36g protein, 12g fiber. You've got 1,998 kcal remaining.
Kristoffer: Lunch is the packed box from yesterday's bolognese.
Claude: Added. Running total: 1,167 kcal, 89g protein, 16g fiber.
Kristoffer: Had 2 pieces of the chocolate protein bark with coffee.
Claude: Got it. Now at 1,356 kcal, 96g protein, 19g fiber.
Kristoffer: Dinner is tacos - 2 portions with the usual toppings.
Claude: Logged. 2,089 kcal, 139g protein, 26g fiber. Room for dessert?
Kristoffer: Half batch protein ice cream with marshmallows and chocolate.
Claude: Final: 2,408 kcal (96%), 162g protein (108%), 36g fiber (120%). Solid day.
Kristoffer: Perfect, thanks!Because I was tracking using an agentic LLM, I could also do stuff like sharing the full recipe and telling Claude my fraction of the meal, arguing about satiation when estimates were too high, or ask for suggestions on how to optimize a snack to hit as close to my target calories as possible. It was also very freeing to give it rough measurements of cakes to try to estimate macros based on volume, known average densities, and typical ingredients.
Kristoffer: Made korvstroganoff for dinner. Recipe was 670g falukorv, 5dl matlagningsgrädde, onion, tomato paste. I had about 1/7 of the total with 120g cooked pasta.
---
Kristoffer: The kantarellpaj estimate of 750 calories feels too high - I'm not that full. Can we adjust to 700?
---
Kristoffer: I have 282 calories left. Help me build the ultimate snack to get as close to target as possible.
Claude: How about: 2 Oreos (106 kcal), 1 Dave & Jon's chili mango date (45 kcal), 10g Marabou dark chocolate (55 kcal), and 15g chili mango (76 kcal). Total: 282 kcal exactly.
Kristoffer: Perfect, let's do it.
---
Kristoffer: The kantarellpaj piece was roughly 2.5 x 5 x 5 cm.I’m sorry for the inclusion of Swedish words above. Claude does fine with multi-lingual prompts, and not everything makes sense when translated to English.
The results
If I may brag a little.
Mid December 2025, I reached 80kg, right before the winter holidays (thankfully, I would not have wanted to track my Christmas snacking). I was surprised it went so smoothly, because the entire thing felt too easy (not trying to brag, I’m genuinely surprised it went this well). And so far end of January 2026, I’m still hovering around 80kg.
Near the end of the project, I had learned to listen to my fullness signals, I had actually learned to stop eating when I was full instead of when the plate was empty. I’ve always admired my wife for doing this, so I’m quite proud of myself for finally getting it!
I also found myself developing an intuition for how many calories were in foods after first being surprised by how dense some foods are while others are confusingly low. For example, it’s surprisingly OK to have a piece of cake in the afternoon as a piece typically is around ~300kcal, whereas sharing a cheese platter with the missus may be closer to 1000kcal (which is a meal and a half…). Kladdkaka was also surprisingly high, but I got used to it and now love it. If you’re ever in Sweden, please try Kladdkaka if you haven’t. It’s glorious.
Another thing I didn’t expect was how much easier everything became after losing the weight. Simple stuff like standing up from the couch, picking up my daughter, or walking up the stairs became noticeably easier. This baffled me. I also don’t snore as much (this my wife is very happy about).
I’ll make sure to share some of the recipes over in the Recipes tab.