Six weeks ago I sat in a chair at the Royal United Hospital in Bath, had a needle stuck in my port, and started Cycle 1 of my latest round of chemotherapy. FOLFIRI plus cetuximab, if you want the technical name. Poison plus more poison, if you want the honest one.

What made this round different from the ones before it? This time, I tracked everything. Every single day. With AI.

Not in some vague "I noted how I felt in a journal" way. I mean proper tracking — symptoms, energy, mood, nausea, pain, rash, sleep, recovery, strain, heart rate variability, blood oxygen, skin temperature. Every day, logged into Claude, cross-referenced with my WHOOP data, analysed against previous entries, and spat back at me as daily insights I could actually use.

Three cycles. 42 days. Here's what happened.

The Setup

I built a daily tracking protocol in Claude — you can read more about the tools on the Chemo Tracker page. The short version: every morning, I tell it how I'm feeling. Every evening, I pull my WHOOP data. Claude combines the two and gives me a daily report — what's happening now, what happened at the same point in previous cycles, and what I might expect tomorrow.

It backs up into Google NotebookLM for long-term memory. So the daily tracker handles the here-and-now, and NotebookLM holds the bigger picture.

Before Cycle 1, I was tracking from a standing start. No historical data. No patterns. Just a blank document and a vague hope that this would be useful.

Forty-two days later, it's the most useful thing I've built.

What the Data Actually Shows

Here's the bit that surprised me. I've done chemo before — this isn't my first round. I thought I knew the pattern. Infusion day, steroid high, crash on day 3, gradual recovery, repeat. And broadly, that's right.

But the data told me things I'd been too brain-fogged to notice on my own.

The steroid crash is real, predictable, and manageable. Day 3 of every cycle, like clockwork. Dexamethasone wears off and your body just... stops. Energy goes from a 7 to a 3. But here's the thing — by Cycle 3, I was planning for it. Rest day scheduled. No meetings. Food prepped the day before. The AI flagged it before each cycle: "Based on your previous data, expect a significant energy drop tomorrow." And it was right. Every time.

My WHOOP recovery score is a better predictor than how I feel. There were days I woke up thinking "yeah, I'm alright" and my recovery was 18%. And days I felt rough but was actually in the green at 68%. The AI learned to weight the WHOOP data alongside my subjective scores, and the combination was more accurate than either on its own.

The rash follows its own cycle within the cycle. Cetuximab gives you a facial rash — it's actually a sign the drug is working, which is the most cancer thing ever. "Good news, your face is falling off." The tracker showed me that the rash peaks around days 4-5, improves by day 8, then has a secondary flare around day 10. I'd never spotted that second peak before. Now I adjust my antihistamines for it.

Sleep is everything. The correlation between sleep performance and next-day wellbeing was the strongest pattern in the data. Not sleep duration — sleep quality. WHOOP measures both, and the AI picked up that anything below 70% sleep performance basically guaranteed a rough following day. Above 80% and I'd rate my wellbeing at 7 or higher. Every single time.

The Things AI Got Wrong

It's not perfect. Obviously. This section matters because if I only told you the good bits, I'd be doing exactly the kind of wellness bullshit I hate.

Early on, the AI tried to be encouraging in a way that felt hollow. "You're doing great!" No I'm not — I'm on chemo, I feel like shit, and my recovery score is 18%. I had to retrain the tone. Now it's honest. "Recovery is low. Rest today. Don't push it." That's what I need.

It also struggles with one-off events. A bad night's sleep because the kids were up doesn't mean my chemo is hitting harder. A great mood score because I had a good day with friends doesn't mean the drugs are working better. Context matters, and AI isn't always brilliant at separating signal from noise when the sample size is one human.

And the skin temperature data from WHOOP has been consistently odd — running below baseline in a pattern the AI keeps flagging but can't fully explain. I've mentioned it to my nursing team. The AI was right to flag it, even if neither of us knows what it means yet.

What Changed in My Behaviour

This is the part that matters most. Data is useless if it doesn't change what you do.

I fast before every infusion now. 24 hours minimum. The data showed my Day 1 nausea scores were consistently lower when I'd fasted beforehand. Sample size of three, so take it with a pinch of salt (which is about all you can have while fasting), but it's enough for me to keep doing it.

I take a nap every single day in the first week post-infusion. Not because I'm lazy — because the data shows my afternoon energy crashes between 1pm and 3pm on days 2-7, and a 90-minute nap correlates with better evening wellbeing scores. I used to fight the tiredness. Now I schedule around it.

I know which foods trigger GI issues. Fried food and rich cheese on days 3-5 of a cycle? Bad idea. The tracker caught it. I adjusted. Less time in the bathroom, more time being a functioning human.

And I go into each infusion calmer. Not because I'm less scared — I'm always scared — but because I know what's coming. The AI gives me an outlook the day before each infusion. "Based on your Cycle 1 and 2 data, here's what Days 1-3 will probably look like." Having data to anchor to when your brain is screaming is genuinely useful.

The Quantified Cycle

I've put together a presentation — The Quantified Cycle — that walks through what this looks like in practice. The actual data, the patterns, the daily reports. If you want to see what AI-tracked chemo actually looks like beyond this blog post, that's where to look.

It's also the first step toward something bigger. The tracker I built for myself is becoming a tool anyone can use. You don't need to be technical. You don't need a WHOOP (though it helps). You just need to be going through treatment and want a better way to understand what's happening to your body.

Six Weeks In

I'm writing this on the eve of Cycle 4. Recovery score yesterday was 65%. Mood is 9 out of 10. Rash is a 3 — I can feel the spots coming on my nose, which means tomorrow they'll be properly visible. Sleep was 85% last night. All things considered, I'm going into this one in decent shape.

I know all of this because I tracked it. Because an AI tool I built reminded me to log it, pulled my wearable data automatically, compared it to 42 previous days of entries, and told me where I stand.

That's not a gimmick. That's not a tech demo. That's a person with incurable cancer using the tools available to him to understand his own body a bit better and make the shit days slightly less shit.

Which is the whole point of FC:AI, really.

Fuck cancer. Track everything. Use the data.