The Working Form

Conservatively underloaded

2026-09-10T00:00:00.000Z

Back running after a long stint off, I went out for what I would have called a moderate 4 miles. It felt like the kind of run I have done hundreds of times, nothing about it felt hard while I was in it, and then I looked at the data afterwards and every single mile was zone 4. Not the last mile, not a hill in the middle, the whole thing. I didn't feel good after it either, and I spent most of that evening trying to work out how a run I would have honestly described as easy came back as a threshold effort.

Let me take a step back and explain how I got there.

Heart rate zones are just a way of slicing a session by how hard the cardiovascular system is working. Zone coverage is the share of the run that landed in each of those bands, so a report that says every mile was zone 4 is a coverage report, not a vibes check. Zone 2 is the easy aerobic band. The accepted field test, when you don't have a lab, is whether you can still hold a conversation at that pace. If you can talk in full sentences, you are roughly there. If you can't, you have already left it.

The model I imported

I have some experience in the gym working on strength and hypertrophy, so progressive overload isn't an abstract idea for me. I have spent far more time thinking about that model than I have about heart rate, and I carried the lifting model of adaptation straight into endurance without ever really inspecting it. Under a barbell the logic is clean, you expose yourself to a stimulus, you recover from it, and after enough repeat exposures that same stimulus gets easier so you add load. I assumed running worked the same way, do the run, whatever it happens to feel like, and eventually the run gets easier. Under that assumption the zone genuinely didn't matter to me, it was just a number describing how hard a particular day had been, and as long as I got the miles or the ride done I was successfully working toward the goal. That was wrong.

So I went back to first principles and asked myself,

If repeat exposure to a stimulus is what drives the adaptation in the gym, what exactly is the stimulus I am repeating on a run?

I didn't have an answer, which is usually the sign that I have been running on an assumption instead of an understanding.

What zone 2 actually means

That question sent me into zone 2 podcasts and videos, and after listening to Dr. Íñigo San Millán on The Proof it became pretty clear that I had been treating aerobic exercise mostly wrong. His points, as I understood them, were as follows.

Zone 2 has an actual physiological definition rather than a perceptual one, it is the highest intensity at which blood lactate stays at or below 2.0 mmol/L, and training there selectively loads the type I slow twitch fibers, which is what drives mitochondrial density and efficiency (remember, the powerhouse of the cell).
Fuel use shifts with intensity, and the way he describes it, once the effort creeps into zone 3 the rising lactate actively blocks the CPT-1 and CPT-2 transporters and shuts fat oxidation down, so a harder version of the same run isn't a bigger dose of the same stimulus, it's a different one.
Zone 2 also trains the body to clear lactate quickly, which is what lets you hold a hard effort longer later, so the easy work is what eventually pays for the hard work.

He frames all of it as preventative medicine as much as performance, insulin sensitivity and metabolic flexibility and age related decline, which is not what I came looking for but is hard to ignore.

Worth noting that the definition is a blood measurement and I am tracking all of this with a Whoop for now, which is its own problem and probably its own post.

The humbling part

Then came the humbling part. If I wanted to actually build the aerobic side of this I had to stay under that ceiling, and staying under it meant I could not run continuously, so I ended up run walking on a treadmill, which is entirely new for me. My whole mentality with running has been to push through whatever the run is doing to me, and that works well enough when the goal is to finish a distance, but it doesn't seem to work at all when the goal is aerobic adaptation, because pushing harder moves the effort out of the range that produces the adaptation I am after. The walking intervals felt like failure for about the first mile and then they just felt like the work.

What surprised me most is the gap between what I can do and what I have built. I could run for miles and still have a barely developed aerobic base underneath it. Those two things had always felt like the same thing to me and they are clearly not.

After two runs of this I can now jog the whole way at a slow pace without the walk breaks, which happened faster than I expected. Two runs is not evidence. It could be mitochondria, it could just as easily be that I have finally stopped starting every run at a pace I can't hold, or that a treadmill removes the hills and the pacing decisions and makes the whole thing artificially easy to control.

The argument on the other side

And then, because I kept watching, I ran into the argument on the other side. A GCN video made the practical case rather than the physiological one, and the numbers there are what complicate this for me. Elite cyclists ride 20 to 30 hours a week, and at that volume spending roughly 80% of the time in zone 2 isn't really a choice, it is the only way to accumulate that many hours without burning out. But for a time crunched amateur training under 6 hours a week, the argument goes, riding strictly in zone 2 doesn't provide enough overload to move VO2 max or FTP much at all, and zone 3 and 4 intervals return more performance per hour spent. Which is not a contradiction of San Millán exactly, they are answering different questions, one is describing what the tissue does and the other is describing what to do with a Tuesday evening.

My plan runs 12 to 15 hours a week, which puts me above the low volume bucket where the intensity argument is strongest and well below the elite volume where the 80/20 split is forced. I can't tell which side of that line I'm actually on. I don't think it changes what I should be doing right now, because my problem isn't that my zone 2 is too easy, it's that until very recently I didn't have a zone 2 at all, and running every mile at threshold by accident is not the same thing as doing prescribed intervals on purpose. So the base comes first either way. But somewhere in the next few months there is a question about how much intensity goes back in and when, and I don't have a way to answer it yet beyond watching what happens.

The thing I am actually watching is whether the pace at which I can hold zone 2 keeps dropping. That number moving is what would tell me the base is being built rather than the effort just being managed better, and it's the one measurement in all of this that I can take myself, every run, without buying anything.

Hydration — separating the myths from the science.

n = 3

Buoyant? Negative

2026-09-08T00:00:00.000Z

I don't float. Most people can lie on their back in a pool and just stay there, and I can't do that, I go down. It isn't dramatic and it isn't fast. This means that every stroke, no matter how much air I breathe in, is a fight against gravity.

Let me step back and explain.

I've run marathons and I've ridden centuries, so I have a reasonable idea of what hours of work feels like and what it takes to get there. None of that part is new to me. What is new is that I never received any swimming instruction beyond what I learned in day-camp decades ago. This isn't a case of my technique needing some polish, it needed a complete overhaul. For reference, when I started the honest answer to how far I could swim continuously was two lengths (50m), and I was out of breath at the end of them.

That combination is most of why I signed up. I've been looking for something engaging and difficult, and going back to marathons or centuries would have meant revisiting things I had already done (boring). The Ironman has something in it I have genuinely never done, and that made it a real test rather than a longer version of a test I had already considered passed. The swim is the whole reason it counts.

Race is Ironman Lake Placid, Sunday July 18th 2027.

Weeks to race45

Swim, continuous285 m

Race swim3,700 m

Planned hours350

Two lengths to two hundred and eighty five

My plan is a distance based progression over the training period. Every long session is written as a percentage of the race distance, so an eighty mile ride reads as 72% of 111.3mi, and I can see exactly how short I am at any given moment. That visible gap is most of the motivation.

For the swim that means going from 200m continuous in week one to 3,700m in open water, in a wetsuit, in chop, with a few hundred people swimming over the top of me. Three weeks in I am at 285m. The distance between those two numbers is the honest scale of this thing.

I decided early on to forget everything I thought I knew about swimming from day camp and rebuild from the ground up, reading books (shoutout Total Immersion!), watching videos, and talking to people. That has been eye opening, mostly because it turns out I had a lot of confidently held ideas that were just wrong.

The lane

Two days ago the pool was crowded and I ended up sharing a lane. I don't normally talk to people at the pool but the tighter lane sparked a conversation, and he gave me a couple of tips. His main points were as follows:

My legs were sinking, which is a known problem I have been dealing with the whole time.
My breathing timing could be better coordinated.

Neither was groundbreaking but it corroborated things I had heard from others and noticed myself, and it validated some of the progress I have made elsewhere with arm movement and staying long.

What has been more interesting is what it got me thinking about outside the pool.

If two people independently watching my swim mechanics can reach the same conclusions with vision alone, why can't a camera based system do the same thing?

I am not sure how far that idea goes yet. Vision is doing a lot of work in that sentence and a person on a pool deck has years of watching swimmers behind them, which a camera does not. Can't tell yet. But it is a question I keep coming back to, and I have started building toward an answer.

Building a cardio base, Zone 2 training, and VO2 Max.

n = 2

Am I just hardware?

2026-09-04T00:00:00.000Z

I spent most of a Sunday afternoon building out the training plan for next July, and somewhere around the second block I noticed I had been using the word programming for about two hours without once thinking about what I was saying. In fitness, programming means writing a set of instructions and protocols to be completed over a duration. In software, which is what I do every day, programming means writing a set of instructions to be executed over a duration. Same word, and after years of writing the second kind I don't think I had ever consciously connected it to the first.

Once I noticed it I couldn't stop noticing it. The plan I was writing has a structure I recognize. There are loops, four week blocks that repeat with the volume incremented each pass and a deload as the exit condition. There are conditionals, dependencies, and even error handling! I was writing a program. The only real difference was the runtime.

So I asked myself the obvious thing,

If the training plan is a set of instructions written now, to be executed later, what is actually different about it from the code I deploy every day?

Where the comparison holds up

More of it held than I expected. The body takes an input, it does something with it, and it produces an output I can measure, and the whole loop is observable if I bother to instrument it, which is exactly what a heart rate monitor, power meter, and a pace field are. The adaptation isn't mysterious either, it is a response to a load applied at a frequency over a period, which is a specification. I can write that specification down eleven months in advance, hand it to the system, and have a reasonable expectation of what comes out the other side. I have done that with software many times and never once thought of it as remarkable.

The recent humanoid hype is what pushed the thought further. With the latest developments in AI, more and more companies are exploring opportunities for humanoid robots to enter the workforce. If I set aside the human emotions, the self perception, and the conflicting priorities, it starts to look like the physical body may be glorified hardware. Actuators, a power system, a control loop, a maintenance schedule, and a set of tolerances that degrade when they get exceeded.

I want to be careful here because there are vast interpretations of what a body is, from the purely physical to the deeply spiritual, and I am not equipped to argue any of them and I am not trying to. I am interested in a much narrower version of the question, which is whether I can treat my own body as a system that accepts a protocol, and what happens over eleven months when I do.

Where it falls apart

It falls apart in at least one place I can already name. Hardware degrades under load. Every mechanical system I worked on back in my mechanical engineering days had a duty cycle and a fatigue life, and a good part of the discipline was about managing the rate of decline. Biology does the opposite under the right dose. You apply a stress, the system fails a little, and it rebuilds past where it started. That is not a property any hardware I have ever specified has, and it means the analogy is inverted in the one dimension that the whole plan depends on.

The other place it falls apart is the part I waved away in the first place. The three things I set aside to make the comparison work are the three things that decide whether the protocol gets executed at all.

I can write a perfect eleven month plan on a Sunday afternoon. Whether the 5am swim happens on a Tuesday in February is not a question about the hardware, it is a question about emotion, self perception, and competing priorities, which is to say a business, family life, prior commitments. In software the deploy either succeeds or it fails, rather binary. Here the compiler can decide it doesn't feel like it.

So the honest position is that the comparison is useful for the parts of the system I can specify, and it goes silent on the part that determines whether the specification ever runs.

What I'm actually going to do about it

I'm going to run the protocol and see. The plan covers roughly eleven months and ends at Ironman Lake Placid, which as the course reads is a 2.3 mile swim, a 111 mile bike with 7,432 feet of climbing, and a 26.2 mile run. I could not come up with a better test of this hardware, partly because it is long enough that nothing I do in a single week matters and everything I do across forty six weeks does.

I am also using LLMs heavily for the planning and the week to week adjustment, which makes the whole thing a little recursive, one program writing another program to be executed by a third thing that may or may not be a program. That is the part I am most curious about. Not whether the training works, plenty of people have proven training works, but whether treating myself as a system with a specification actually changes the outcome compared to how I've trained before, or whether it just changes how I feel about the training.

I don't know yet. I'm forty six weeks out and the only data I have is that the plan exists and I have only begun executing.

Weeks to race46

Plan length49 wk

Planned hours350

Bike climbing7,432 ft

The swim, and overcoming buoyancy and preconceived notions.

n = 1

The Working Form

Conservatively underloaded

The model I imported

What zone 2 actually means

The humbling part

The argument on the other side

Next

Buoyant? Negative

Two lengths to two hundred and eighty five

The lane

Next

Am I just hardware?

Where the comparison holds up

Where it falls apart

What I'm actually going to do about it

Next