Why we built BabyFoodTracker

With our first baby, feeding was surprisingly uneventful.

We introduced foods.

She ate them.

Nothing happened.

A few years ago, I built a tiny website to keep track of what she had tried. I’m a developer, so that’s usually what happens when I notice some small inconvenience in my life: instead of downloading an app, I convince myself I can build one in a weekend.

I used it for a while.

Then I stopped.

Because, frankly, there wasn’t much of a reason to keep using it.

When everything is going well, you don’t care that much whether your baby tried avocado on Tuesday or Thursday.

You don’t care exactly how much egg they ate.

You don’t care what time they ate it.

You certainly don’t think:

I should document this carefully in case a doctor asks me about it two months from now.

That would be weird.

Then we had our second baby.

And I realized I had misunderstood the problem entirely.

I – You only know what mattered after something goes wrong

Our second baby was allergic to egg, milk, and tree nuts.

Nobody in my family had food allergies.

Nobody in my wife’s family had food allergies.

So food allergies were one of those things that obviously existed, but existed somewhere else.

Other families dealt with them.

Not us.

Until suddenly we were sitting in allergy appointments being asked questions like:

What did he eat?

How much?

Was this the first exposure?

How was it prepared?

How long after eating did the reaction start?

What exactly did the reaction look like?

Had he eaten this food before?

Did you take a photo?

These are incredibly reasonable questions.

They are also surprisingly difficult questions to answer when you weren’t expecting to ever be asked them.

We had the information.

Sort of.

Some of it was in our memories.

Some of it was in text messages between my wife and me.

Some of it was buried in thousands of photos.

Some of it was probably gone forever.

And this is where I started to understand the actual problem.

The problem wasn’t remembering what foods our baby had tried.

The problem was that you don’t know which ordinary Tuesday will become medically important six weeks later.

That changes how I think about tracking.

Because most tracking products are built around the assumption that the act of tracking itself has value.

I don’t think that’s true.

Nobody wants another thing to log.

Parents definitely don’t.

You are already sleeping less.

Cleaning more.

Trying to remember diapers, naps, bottles, appointments, daycare, laundry, and whatever object your toddler decided to hide inside the couch that morning.

Adding “maintain a beautiful food database” to that list is ridiculous.

The record only becomes valuable later.

Which means the product has to make recording something now almost effortless, because the person doing the recording cannot possibly know whether that information will matter later.

That is the first principle behind BabyFoodTracker.

II – The goal isn’t to track more. It’s to forget less.

There is a strange period when you start feeding a baby solids.

For the first few months of their life, food is simple.

Milk.

Then suddenly the entire world opens up.

Egg.

Peanut.

Salmon.

Yogurt.

Strawberry.

Rice.

Sesame.

Beef.

Blueberry.

One ingredient turns into ten, then fifty, then a hundred.

At the same time, different questions start overlapping.

Can they eat this food?

How should it be prepared at this age?

Is this a choking risk?

Is this a common allergen?

Have they eaten it before?

Did they tolerate it?

When did they last have it?

These sound like food questions.

But they’re actually information problems.

And parents solve information problems the way humans have always solved them:

badly.

One parent remembers one thing.

The other remembers something else.

Grandma feeds the baby something and mentions it three days later.

Someone takes a photo.

Someone writes a note.

Someone says, “I’m pretty sure she already had this.”

Nobody is doing anything wrong.

The system is wrong.

The information is fragmented across people and time.

Once I saw the problem this way, the original version of BabyFoodTracker looked almost silly to me.

It was a food tracker.

What we needed was shared memory.

That sounds like a subtle distinction.

It isn’t.

A tracker asks:

What did your baby eat today?

A memory system asks:

If this becomes important three months from now, what will you wish you remembered?

Those lead to very different products.

III – We learned the important things after we needed them

The allergy experience also taught us something else.

There was a surprising amount we simply did not know.

We learned that, for most babies, common allergens like peanut and egg generally do not need to be intentionally delayed once the baby is developmentally ready for solids.

We learned that if a baby has severe eczema, has already reacted to food, or otherwise seems higher risk, that is a conversation worth having with a pediatrician early.

We learned that introducing an allergen once is not necessarily the end of the story.

If a food is tolerated, keeping it in the diet matters too.

And we learned how useful the boring details can become:

The food.

The preparation.

Roughly how much.

The time.

What happened afterward.

A photo if there was something visible.

None of these ideas are particularly complicated.

That’s almost what bothers me.

Because when I look back, I don’t think we needed some giant medical textbook.

We needed the right information at the right moment.

That is a very different problem.

There is more information available to parents than at any point in history.

And yet parents are still constantly anxious.

That sounds contradictory until you understand that access to information and access to the right information are not the same thing.

You can Google almost anything.

But when you are holding a six-month-old and a spoonful of peanut butter, you do not want 400 search results.

You want to know:

Is this appropriate now?

How should I serve it?

What should I pay attention to?

What do I do next if everything goes fine?

And when guidance changes based on the food or the age, that context should live next to the decision.

Not inside seventeen browser tabs.

That became the second principle behind what we are building:

Good information should appear at the moment it becomes useful.

Not before.

Not after.

IV – Our third baby made me rebuild the whole thing

We recently started this process again with our third baby.

This time, I opened the old project.

And instead of thinking about which features I could add, I started thinking about everything I wished we had known the second time.

So we began going through foods one by one.

This became much more work than I expected.

Because “can a baby eat broccoli?” is actually a terrible question.

A better set of questions is:

Can a 6-month-old eat broccoli?

In what form?

What about at 9 months?

What about at 12 months?

What preparation creates a choking risk?

Does the risk come from the ingredient or the shape?

Is this an allergy issue?

Is this an official recommendation?

Or are we simply choosing to be more cautious?

When you look at enough foods, you start to see the same pattern.

Parents don’t need more rules.

They need distinctions.

That matters because baby feeding advice has a tendency to become strangely absolute.

Never do this.

Always do that.

This food is good.

That food is bad.

Reality is usually more nuanced.

Age matters.

Preparation matters.

Development matters.

Medical history matters.

And sometimes the honest answer is:

This is our more cautious recommendation, not a universal medical rule.

I would rather tell parents that than pretend certainty exists where it doesn’t.

Because BabyFoodTracker is not supposed to make decisions for you.

It should help you make better decisions.

There is a difference.

V – The product becomes obvious once the problem is clear

When you understand the problem this way, a lot of the product decisions stop being product decisions.

They become consequences.

Logging should be free.

Obviously.

The history is the foundation of the entire thing. Putting the basic act of remembering what your baby ate behind a paywall would defeat the purpose.

The family should share one history.

Obviously.

Because a baby does not care which parent’s phone contains the information.

Foods should explain how preparation changes by age.

Obviously.

Because “baby can eat carrot” is useless advice if the way the carrot is served is what matters.

Allergens should have their own flow.

Not because I want parents to obsess over them, but because I want the process to feel boring.

One food.

A small amount.

You know what was introduced.

You know when.

You know whether anything happened.

If everything is fine, great.

Keep going.

Boring is the goal.

And if something unusual ever does happen, you have a history.

Not a diagnosis.

Not an AI pretending to be a doctor.

A history.

That distinction is important.

BabyFoodTracker cannot tell you a food is safe for your child.

It cannot diagnose an allergy.

It cannot replace a pediatrician or allergist.

I don’t want it to.

Doctors are very good at being doctors.

Software is very good at remembering things.

We should probably let each do what it is good at.

VI – I built the same product three times

The first time I built BabyFoodTracker, I thought the problem was:

Parents need a list of foods their baby has tried.

After our second baby, I thought the problem was:

Parents need better allergy tracking.

Now, with our third baby, I think both were incomplete.

The real problem is that parenting creates thousands of tiny moments whose importance is impossible to know in advance.

Most of them disappear.

That is fine.

Some of them matter later.

That is when you wish they hadn’t disappeared.

So the job of BabyFoodTracker is not to make parents track everything.

That would be miserable.

The job is to make the important things easy enough to record that you actually do it.

Then organize those tiny pieces of information into something useful when you need them.

That’s it.

And weirdly, it took three babies for me to understand that.

Baby number one gave me the idea.

Baby number two gave me the reason.

Baby number three gave me the perspective.

I’m building BabyFoodTracker again.

But this time, I finally know what I’m building.

— Nil