A few weeks ago, the Canadian fires caused a lot of the northern United States to undergo serious air hazards and breathing conditions, and the cycle of people thikng ‘omg something is happening, what is it’ hit the masses again, me included.

Air quality is communicated to the public through something called AQI, or the air quality index. I had heard the term before but never really did the research into what it meant, so here’s a deep dive.

TLDR: turns out our ability to track air quality is pretty accurate!

It’s just that our air quality is bad lol.

What does air quality even mean?

The Air Quality Index (AQI) is genuinely a great piece of information design. It’s used by multiple countries in the world, and it’s able to take a bunch of math and condense it into an accessible metric that tells you if the air outside is ‘good’ or ‘bad’. It’s on a scale of 0-500, with 6 categories of quality (good, moderate, unhealthy, etc).

The first thing to understand about AQI itself is that it’s a man-made index. There are multiples types of pollutants out there, and the ways we can measure them are not always the most comparable. For example, we use micrograms per cubic meter for particles, but we use parts per million for gases. What does that even mean?

It means that we need a way to factor both into the conversation without getting bogged down on how they compare. And that’s what the AQI does.

On any given day, the index reports the single worst out of a group of particular pollutants: PM2.5, ozone, PM10, nitrogen dioxide, sulfur dioxide, carbon monoxide. Some apps allow you see the amounts of all five pollutants, but most just stick to the highest number among them all. AQI runs each pollutant through its own conversion algorithm so that a score of “35” means the same severity, whether it comes from ozone or smoke.

The biggest dangers are PM2.5 particles

Technically any one of the 6 possible pollutants can influence the AQI score, but the pollutant that usually drives it is the PM2.5.

PM2.5 particles are basically everything floating in the air smaller than 2.5 microns. PM2.5 is more like a classification of particles rather than a specific substance, and they’re paid attention to specifically because they’re small enough to lodge into your body and lungs.

It’s the pollutant most tied to premature death worldwide, so when your app shows a low AQI, it’s most likely telling you that the PM2.5 in the air is low. Here’s a table conversion to see how the range of PM2.5 range might compare to AQI scores. If you want to think about it , 12 µg/m³ ≈ one-fifth of a grain of salt dispersed through a cubic meter of air. That’s a lot less dangerous than something like 100 µg/m³, which would be like two grains of salts in one cubic meter of air.

Because PM2.5 is a mass concentration, it can be compared against other size classifications - most commonly PM10.

The naming is kinda counterintuitive. PM10 doesn't mean particles around 10 microns; it means everything 10 microns and under, so it includes all of PM2.5 inside it. A lot of the materials between PM2.5 and PM10 are less harmful; road dust, sea salt, debris, etc.

And then there’s also the issues of ultrafine particles, which are those that are below 0.1 microns. These particles are small enough to pass straight into the bloodstream, but they’re too small to be counted into the PM2.5 mass number. So technically, the AQI scale doesn’t account for those particles even though we may be ingesting the millions of these particles.

How does the US measure air quality?

The US is pretty decent at our air monitor coverage. About one in three US counties has a regulatory PM2.5 monitor. The other counties are filled in with models: satellite data, weather, and interpolation from distant monitors. Which is better than nothing, and often fine for a regional average.

But a 2024 analysis found roughly 2.8 million people live in pollution hotspots the monitoring network simply doesn’t capture, and that translates to about 44% of large metro areas that are under-monitored.

Ok, so we know the state of our monitoring. But what about our AQI scaling?

Recently, the AQI scale has changed in the US, and our classification of ‘good air’ has become a little more strict. Our legal limit is already nearly twice the limit of the WHO’s guideline for safe air, but it did become more strict in 2024.

There’s currently a battle to change it back its standard before 2024, a move which is made to loosen requirements, but that’s currently being discussed in the government.

How is the US doing on air quality?

I’m happy to say that the US is has been getting dramatically at serving healthy air to its citizens over time. US PM2.5 concentration has fallen roughly 40% since 2000, and that’s the main thing we care about due to its links to mortality.

A big part of the success is the Clean Air Act, which encouraged closing coal and gas power plants and incentivizing car fuel efficiency. Plus, violators got sued and fined.

All that’s to say, the current air monitoring network, however incomplete, is still one of the greater environmental data systems working today, and being able to even see gaps in coverage is useful for any next steps.

How is the world doing on air quality?

Compared to the last few hundred years, the world’s air is a lot more polluted than before, which makes sense when you factor the explosion of car use, the rapid industrialization of countless countries, and the heavy reliance on coal and gas up until the turn of the social. As a planet, we’ve been edging downwards slightly, which is good, but we still are far way away from our starting point.

In 2021 the World Health Organization, after reviewing the accumulated science, set its recommended annual PM2.5 limit at 5 µg/m³.

The US EPA’s legal standard, even after it was strengthened in 2024, is still nearly double the health guideline at 9.0 µg/m³.

Once you factor in the standard to the WHO’s 5 µg/m³ line, you read a scary thought: more than 99% of the world’s population breathes air that exceeds the WHO’s annual PM2.5 guideline.

Although the trend is going in the right direction, it’s important to understand that we are still a far way from the goal of what the WHO has decided is clean air. Air quality is something I’ve gained newfound interest in, so I expect to write about it a few more times on this blog, and we’ll see how the situation changes over time.

I recently had to deal with a project that relied heavily on the national weather services’s API. I’m no stranger to the public utilities that governments help provide us, especially the information-heavy ones that are helpful to us citizen journalists, but I always try to do some digging, and find out more about who’s funding it.

So I went down a rabbit hole, and found out that weather prediction is a billion dollar industry that’s gotten good over the years. But I also found out that it’s currently at risk of being underfunded, which seems so interesting to me when we’re actively going under a climate crisis. Is weather forecasting a solved problem, or is it still worth improving upon?

The basis of how weather prediction works

Weather forecasting is fundamentally the science of predicting how the atmosphere will behave over time, and it relies on an enormous amount of real-world data collected continuously from satellites and other devices.

All of this data gets fed into numerical weather prediction (NWP) models. The most well-known of these are the GFS (Global Forecast System, run by NOAA) and the ECMWF (European Centre for Medium-Range Weather Forecasts).

The reason forecasts are broken into 3-day, 5-day, 7-day (and sometimes 10- or 14-day) windows comes down to a hard physical limit known as the chaos problem. The atmosphere is a chaotic system, meaning that tiny errors or uncertainties in the initial measurements compound exponentially over time.

At 1 - 3 days out, forecasts are highly reliable because the initial conditions are well-captured and errors haven’t had time to grow. By days 4 - 7, you’re in a zone of moderate confidence, and past that you're mostly looking at a well-educated vibe.

Weather forecasting has gotten dramatically good

This chart from Our World in Data and pulls its numbers from the National Hurricane Center, and shows hurricane track error - basically, how many miles off a forecast was from where the storm actually went - over the decades. You can see that the track error has been decreasing.

Here’s a different interpretation that is pretty much the same thing.

We can see that this track error, especially for longer-term forecasts, has decreased a lot over time. In the 1970s, a 48-hour forecast had an error between 200 and 400 nautical miles; today this is around 50 nautical miles.

How did it get this good? A few reasons exist, but I see two that have been playing an outsized part.

First, is the fact that we have more and better data. I mentioned the physics-based models we have earlier (GFS and ECMWF). The more atmosphere that we’re watching in real time, the better our conditions are, and that’s ultimately the solution for the chaos problem.

Second reason is the fact that machine learning has recently gone insane with progress.

Traditional statistical techniques like linear regression and logistic regression have long been used in a post-processing step. More recently, deep learning approaches have taken center stage: convolutional neural networks (CNNs) is an example of a model well-suited at pattern recognition across large geographic areas.

There are other examples I can list (Google DeepMind’s GraphCast, or Pangu-Weather), but the main idea is that large AI models trained on decades of reanalysis data can produce forecasts that rival or beat traditional NWP models in a fraction of the compute time.

All of this weather forecasting is actually worth it

People are more satisfied with weather forecasts than before.

There’s a study called the Communicating Forecast Uncertainty study. The first one came out in 2006 that measured Americans sentiment of weather forecasting. And, they replicated the survey in 2022. In short, they asked people: how much would you pay in taxes each year to keep NWS forecasts?

In 2006, the median was about $286 per household per year, but the number grew to $898.50 iin 2022, with a 95% confidence interval of approximately $700-$1,300 per household per year.

Multiply that across the country and you get a public-forecasting value of about $102 billion a year. And people aren’t just willing to pay more; they’re more satisfied than they used to be back in 2006.

So, ultimately, it’s something that people want, and its something that has been ‘worth it’, economically. The NWS estimates its forecasts alone boost economic productivity by around $13B a year.

Weather forecasting is helpful during the climate crisis

We have had more billion dollar disasters because of the effect of climate change. The country averaged about 3 billion-dollar disasters a year in the 1980s. Over the last decade, the average is around 19.

This is the whole argument for forecasting in one chart: the disasters are getting more frequent and more expensive, and the forecast-plus-alert pipeline is the cheapest thing we have that reduces the body count and the bill. No matter where you move, climate change is going to come after you

When looking at a map of billion dollar disasters and where they occur, you can see that there are a lot of hurricanes in the southeast, and some extreme weather in the middle of the country. Some tornadoes in the midwest, and some wildfires in the southern California.

While short-range forecasting (1–3 days) has become impressively reliable, significant challenges remain. Flash flood prediction at the hyper-local level is still notoriously difficult. Tropical cyclone intensity forecasting (knowing how strong a storm will get, not just where it’s going) lags behind track prediction. as climate change loads the atmosphere with more energy and moisture, extreme weather events are becoming harder to model using historical patterns.

Yet, America tries its best to keep up with alerts, among other things. I won’t spend too much time on our alerting system, but just know that we have a full infrastructure on how and when they get sent out.

Although it’s hard to determine how many weather alerts are sent out each day, a project from weather.com was able to scrape weather alerts over a 10 year period. They retrieved about 300,000 weather alerts during that time. Just think about the sheer volume of information that represents. This chart below shows the breakdown of those alerts by region.

Where does weather forecasting funding stand now?

Remember that NCEI billion-dollar disaster map I just broke down? In May 2025, NOAA retired the database behind it. No updates past 2024. The official reason was “evolving priorities, statutory mandates, and staffing changes”, but former NOAA staff were blunter: it’s part of a broader move to scrub climate material off government platform.

The White House's 2026 budget proposed cutting NOAA's budget by about 27% and eliminating NOAA's Office of Oceanic and Atmospheric Research. They also wanted to eliminate the Office of Oceanic and Atmospheric Research (OAR) entirely, which is the sneaky bad part.

The Trump administration also tried to cut about 600 National Weather Service jobs, but they’ve been allowed to hire again. He’s now looking to increase their budget slightly for 2027, but still wants cuts across the board.

Congress, however, pushed back and, notably, from both parties. The final law (P.L. 119-74) landed NOAA at about $6.1 billion, about the same as 2025, despite the proposed 27% cut, kept OAR alive, and increased NWS funding by 8%.

Personally, I don’t know how much money the NOAA needs to keep running, but I do know that weather forecasting is proving to be a helpful thing we invested in. We have Congress fighting to keep these public utilities well-funded, and nonprofits on the ground trying their best to fill in the gaps. That should not be the system, but at least it’s a guardrail we have for now.

If you’ve been following the tech company conversation in the the United States, you might be under the impression that we’re headed towards a tech oligarchy, something that is bolstered by the ever-increasing presence of tech in our day-to-day lives.

I don’t need to tell anyone that we spend a lot of time on our phones, but here’s another reminder; we spend a lot of time on our phones. I recently wrote about how much energy our phones use doing different activities that cites some sources, but this chart that is a good reminder that it was reaching up to 7 hours back in 2018.

I probably also don’t need to tell anyone how much money tech companies are making from our attention. The estimates vary, but reports out there show that the digital ad industry is reaching between $250 billion and $400+ billion, depending on how specific sub-sectors (like total media buying vs. digital-only services) are categorized. And, in most industries, advertising is increasingly appearing on screens versus traditional formats.

But it took me some time to understand the physics of how the economy worked, and when it comes to social media, a big part of the market is advertising.

My reductive explanation of how the business works goes like this: internet companies track and catalog our behavior, then segment us into populations based off of those behaviors. Customers pay the tech companies money to access these segmented parts of the population…and the business cycle continues.

For many companies, advertising is a big part of revenue

PWC released a report on internet advertising, and its where a lot of my figures will be coming from. Link here.

Every company is different, but for a select few, I wanted to visualize just how important advertising was to each’s business model. After all, Amazon doesn’t need to make as much money from advertising as something like Alphabet (Google, Youtube) or Meta (Facebook, Instagram).

Side note, but to have hundreds of billions of dollars in revenue every year is insane. Yet, the largest companies command hundreds of billions of dollars of revenue every year, and the advertising revenue of some of these companies are large enough to be considered their own companies.

Most remarkable is Meta, which is most-reliant on its ad revenue, and would explain why it takes up so much of the current market share.

Internet revenue is congregating at the top

Out of all the apps that get money from our data, there are a select few who are the most likely candidates of being a part of this tech oligarchy, if not for their market share. This metric is from the PWC report I mentioned earlier, but it shows that the top 10 companies make up like 80% of all digital ad revenue.

What’s even more concerning is that the second largest group is still a concentration select few; companies that are not in the top 25 tech companies only received approximately 7.5% of the money devoted to internet ad revenue.

EMarketer ran a similar study, and while their numbers are slightly different, their conclusion is very similar; large companies like Meta, Alphabet, and Amazon have an outward size advantage on market share for digital advertising. By a large margin, they have more eyeballs.

Not all users are ‘valued’ equally

This might be intuitive to some, but different segments of the population are priced differently. There’s a metric called average revenue per user (ARPU) that defines just how valuable a user might be for a company, and luckily, a lot of tech companies post their numbers publicly.

Now, something that is also interesting to see is how much more ‘valuable’ the attention of Americans are compared to the global average.

It’s no secret that Americans consume more - partially because we earn more, partially because we’re kinda fat and lazy - but to see the trend persisting in each of the companies is something to see. In all the platforms I was able to find the data for, the ARPU for US only users was significantly higher than the global average.

This chart might imply that geography is the only factor that contributes to a person becoming ‘more expensive’ to advertise to, but that’s not necessarily the case; it’s just the easiest to track.

Some doctors might have a higher ARPI than a constructor worker, or a student might have a higher price point to a customer than a parent, just like a Reddit user might be more ‘valuable’ to an advertiser than a Pinterest user.

How do we prevent getting profited off of?

Seeing all of these numbers brings up a simple question; how do we prevent getting profited off of?

I’m not nearly informed enough to write a comprehensive breakdown of how to completely protect yourself against internet trackers, but there are a lot of good pieces already out there.

What I can offer is one piece of advice relating to the security differences between accessing a site via the browser vs via an app.

If you access an app like Tiktok or Facebook on an internet browser, those sites can use things called cookies for cross site tracking…unless blocked by browser. And internet browsers like FireFox and Safari have pretty strong anti-tracking protections.

Apps can collect data inside their own app, but apps cannot access data from other sources unless by asking for specific system permissions. So that means that, while on the Instagram app, Meta will be able to read all your behaviors and maximum profit based on your behaviors, but it can’t do the same to your Hinge app activity, and vice-versa.

All in all, there are probably better options for protecting your behavior on the internet than an app, though remember that not all browsers are created equally. Google Chrome is pretty invasive.

Take that little advice and do with it what you may.

I’ve had the idea of visualizing a harvest calendar for the longest, but finding an accurate data source has been tricky, simply because the equation changes depending on where one lives (I’ve written a post on plant zones already).

I’ve seen many out in the wild, but none that I thought did a great job of telling the story behind the groupings of crops. The ones I’ve encountered have been good to study for an interested super-user, but not fun to look at for a casual glancer. I like to appeal to both audiences. So, I made a few, and along the way, I learned how crops grow, and how climate change is changing those growth patterns.

It’s worth noting also that we live in the days of modern conveniences, so technically there’s a good chance you can get most of these crops at any time during the year.

But, if you ask me, harvest periods are important concerns for the real eaters out there, and I think being a real eater is a lost art.

Citrus can be harvested in the winter

First, we go with the citrus.

A lot of citrus fruit are harvested in winter, after spending months developing on the tree. Their thick rind offers some protection from the cold in their final window before harvest.

Different varieties of citrus have different weather tolerances; lemons and limes are the most sensitive to cold, while mandarins and kumquats sit at the hardier end of the spectrum.

There’s a taste element behind this that I thought was interesting; we might be making our citrus sweeter. When growing, cool temperatures slow the break down of organic acids within the citrus, preserving its tanginess while keeping it edible. Then, warm days allow the tree to keep photosynthesizing, adding sugars/sweetness to the fruit.

Unfortunately, climate change is affecting this balance.

Warmer nights are speeding up the ripening process, and citrus season in some regions is now running a few days to a few weeks ahead of schedule.

That sounds harmless, but it’s fucking up a delicate balance: citrus flavor depends on the right ratio of sugar-to-acid, and the acid develops over time on the tree. If a fruit accumulates enough heat to seem "ripe" before the acid has fully developed, you get something sweet but flat. And a flat citrus is a tragedy; it needs a little sting.

Stone fruit are best harvested in the summer

Similarly, we have climate change issues with stone fruit as well.

A stone fruit is a type of fruit with a fleshy exterior surrounding a single, hard pit (the stone). They evolved in climates with a pretty reliable pattern:

1. A long, cold winter

2. A gradual warm-up in spring

3. A hot growing season

Their biology depends on that rhythm.

During fall, the trees enter dormancy, and their buds basically shut down to avoid getting killed by the cold. The buds require a certain amount of accumulated cold, usually measured in “chill hours”. Only after the tree satisfies that chilling requirement can the buds properly bloom in the spring. Then the tree starts accumulating heat units, which eventually trigger flowering and leaf-out.

This is why stone fruit are intensely seasonal and geographically picky; the timing needs to be the perfect balance between cold and hot. A peach variety bred for Georgia may fail completely in southern Florida because winters are too warm.

Unfortunately, climate change is also affecting this. Weather is also becoming more variable, so trees experience confusing temperature swings on both ends of the spectrum.

Trees may wake inconsistently because they never fully completed dormancy due to warm weather, which may lead to a false spring and can trigger premature blooming in late winter. Then a normal cold snap arrives afterward and kills the blossoms, wiping out the year’s crop. Double oof.

Root vegetables thrive in autumn

Rarely does a brand have as much consistency as the root vegetable. I feel like all of them - carrots, potatoes, beets - all give off a certain vibe of hardiness and sturdiness and being good for a hearty stew.

A big part of that is because of the fact that they grow underground.

As temperatures drop, root vegetables shift from focusing on leaf growth to channeling their energy downward, converting sunlight and nutrients into long-term storage compounds…aka the vegetables. These vegetables are able to thrive in colder environments, so the association to their sturdiness is natural.

The purpose of harvest calendars

I'm not a farmer by any imagination of the word, but there's something grounding about understanding why food tastes the way it does, and how growth patterns ultimately affect the end result.

The seasons are shifting, and some of what I've mapped here is probably already outdated. But that's almost the point. These harvest calendars are meant to be useful tools for me and others, both for having the conversation about how things are working, and as a baseline for noticing when they don't anymore.

As a treat, here’s a harvest calendar for berries. I didn’t have enough time to write a full section on it.

As a web designer, I was taught early on to design with the mobile phone as a priority versus the desktop or laptop, as people are increasingly interacting with the internet on their phones. There are some instances in which I would design something for a power user who would use my tools on a larger screen, but the general understanding is that most people are leaning towards phone usage.

From an energy standpoint, that might be a good thing, especially as the US faces an electricity gap; very soon, we’ll need more power than we can generate.

I looked up the numbers and have to say that phones use a relatively small amount of energy compared to other devices that do the same thing, namely laptops and tablets.

I created a version of the graph with AI figures included because it’s the dominant conversation of our times, but it's worth noting that tits a different figure entirely.

The AI figures measure how much energy is used by the server per query, while the phone/laptop figures show how much energy is being used by the device itself.

And that’s also not to include the energy required to charge a phone, something that takes about 19Wh of energy from 0 - 100%.

The server-side figures are technically available for me to measure the total impact, but the numbers were iffy and I didn’t have time to combine the two this week.

Not all phones use power equally

I covered the topic of electricity and data centers in this blog already, and from it comes this graph that shows the amount of electricity and water certain online tasks do.

When compared to the figures above, some of the trends align, but they did not align 1:1, and when I further researched why that is, I found out that not all online tasks use the same types of servers.

At the risk of being redundant, any time a person wants to do something on the internet, that request (clicking a profile, playing a song, etc) goes to a server in a data center. That server then uses electricity (and, indirectly, water) to fulfill that request and send you the result.

For example, take video streaming. Videos are usually served from CDN edge servers; essentially large caches sitting close to users geographically. These servers are relatively low-power because they’re mostly just reading pre-encoded files from storage and sending them out.

Video calling, however, works through WebRTC servers, which require careful attention to network reliability and infrastructure design And social media servers might require something else entirely.

The energy conversation is a bit tough to compare apples to apples, and while the growing ‘electricity gap’ is a real issue, it’s a bit tough to have these conversations without factoring human usage as well. But, baby steps.

Some housecleaning.

I have an app! It’s a food scanner app called Gourmet Data, and it creates unique charts off a food’s nutrients after you scan it. Check it out today on the iOS app store and leave a review if you feel so inclined (no bad reviews tho)

A few years ago, right when the NFT craze was at its hype, I remember seeing the term ‘web 3’ quite a lot. There’s no ‘official’ way of defining web3, but it was defined loosely as the next generation of how the web was going to work, as there were a few advances happening concurrently at the same time.

That got me thinking about the current technological age we are in, and I thought I’d do my best to link it to the real-world environment on Earth. As more of our lives become digital, what effect does that have on our built environment?

I’ll give you the crude, half-baked answer: it’s basically more data centers and more e-waste.

The increase in software usage

One thing for certain is that there are more software creators. Maybe not full timers, but GitHub (a platform where software developers store their code) added 36 million new developers in a single year, which is 23% year-over-year growth. Excuse me? A total industry grew by 23% in a year? Hmm…..it’s fair to say that is because of AI, which is now writing approximately 46% of all code.

As more of the web is being created by artificial intelligence, it’s interesting seeing how the web is adapting as a result. Sometimes, the web suffers, and hilarious errors appear in everyday applications. But the efficiency with which new software is created is hard to ignore, nor to point out as a recent thing. Since the 2000s, people are just spending more time on their phones and laptops.

More software, unfortunately, means more data center usage, which ends up being one of the biggest changes. The world is increasingly spending more time on the internet, which means that we are using more electricity, and using more land to build data centers. In the US, 60% of the resources we use for electricity are fossil fuels, and our freshwater supply is on a downward slope compared to the rest of the world…both of those resources end up being used data centers keeping up with our digital demand.

Oops.

Everything is digitized

The Internet of Things is where everything is digitized, and different devices can speak to each other. Image automatic lights, autonomous cars, etc.

IoT runs on the principles of general devices and sensor devices, and the fact that everything is connected through a network. The general devices (light bulb, tv) work with the sensor devices (motion camera, app) to do whatever calculations they need to do to trigger whatever actions.

With the growth of IoT we are adding billions of devices to the connected world, with 2025 estimates at 16.7 billion. That means billions of device identities, being used by billions of users, by billions of app instances.

Now, one really unfortunately thing about the internet of things is the increase in devices where manual devices could do the trick. And, often times, that also means there is a huge e-waste issue. The World Health Organization estimates it at 62m tons each year, with only about 22% of e-waste reaching proper recycling facilities. e-waste represents just 2% of landfill trash, it accounts for 70% of toxic waste due to hazardous substances like lead, mercury, and zinc. The UN keeps a really good count on all these numbers.

Not to sound like a broken record or a hypocrite (because I am unfortunately definitely part of the problem), the environmental consequences are intense. improper e-waste disposal releases up to 1,000 different chemical substances into our atmosphere through burning and processing. E-waste toxins infiltrate water supplies through multiple pathways, Heavy metals and flame retardants leach into farmland.

Replacing brain power with AI

One of the more debatable things happening is machine learning and artificial intelligence.

I will not go into the specifics of how machines learn how to do things, but it essentially relies on the idea of training data. This is basically giving a machine a vast amount of information  - either with the ‘right’ answers or not  -  and computers use a lot of math to figure out why the answers are the answers they are.

I already wrote a full piece on the environmental costs of ai, and how it adds extra strain on our data centers. But not only is the electricity usage massive from these data centers - something I wrote about a few weeks ago, but the water used to train these datasets is massive as well, something that gets hidden from the cost. It is estimated we spent 700,000 liters of water training ChatGPT-3 (reminder, we’re on ChatGPT 5 right now).

It’s important to understand that this concept is not new, but we never had enough data for our machines to be smart. Now we do, as well as. But, as mentioned last week, AI uses:

• a decent amount of water

• a lot of electricity

• takes up farmland that is already competitive

So, the major issue from the artificial apps comes from the data centers. As is the case with most of the new software coming out; as we spend more time online doing more complicated things, the electricity and water needed to do these things increases.

Disconnection with the natural world

In the past, 2D tech, we are limited by screens. With virtual reality, glasses can immerse you into a completely different world, and it factors in things like spatial distance and computer processing to present new worlds.

For context, two hours of VR usage on a high end PC would be about 500 watts of energy. The headsets are actually a lot cheaper, energy-wise, ranging from 10 - 30 watts.

Much like AI - and pretty much anything else related to computing - VR technology requires intensive real-time data processing, so it relies on data centers that consume large amounts of energy. And that’s here, again, the sustainability concerns come in.

But one thing that concerns me a little more than the other applications is the idea of disconnection. I don’t need to pull up the countless articles lamenting how much time humans are interactive with the real world vs the digital world, but the issue is covered by countless famous media: The Matrix, Ready Player 1, Wall-E.

In summary, when we’re too busy focusing on the digital realm, maintaining our physical world takes a backseat.

A vision for a more balanced future

Recently, I was in Japan and saw an event thrown by Pokemon Go that collected - no exaggeration - hundreds of people at the park.

And that’s been one of my visions for the future. From my app that I just released which scans food (shameless plug), to Pokemon Go, to QR codes for menus; regardless of the executions, I’m seeing more examples of applications using the physical world as an entry point, and I find that balance a little more reassuring.

Augmented reality has its purpose in its name: it augments what’s already there. Is it enough to say that all issues are null and void? Not even close.

Hey all! Some housecleaning;

I sent this email a few days ago via Ghost, as some of these emails are getting sent to spam or promotions. Gonna align the two sooner or later, but just a reminder to check your spam or promotions folder in case any blogs are gong to them. Thanks for reading!

This week’s piece is inspired by this research article I read that examined the amino acids found in cucumbers before and after lactic acid fermentation.

Fermentation is a very popular food technique, and I forget often that more things are fermented than just kombucha (kimchi, sauerkraut, sourdough, etc). We should be embracing it a lot more as a cooking technique.

What actually changes inside the food during fermentation? Not philosophically, but chemically.

The dataset and the lens

For this piece, I’m looking at the data of certain foods in the USDA database, but I’m also leaning on the aforementioned food-science study that analyzed free amino acid (FAA) profiles of cucumbers.

Unfortunately, the database doesn’t have an entry for fermented cucumbers, but I tried to take examples of other foods that have a clear before and after picture of undergoing the fermentation process. That includes pickles vs cucumbers, cabbage vs kimchi, and milk vs yogurt.

Something I learned is that free amino acids are strongly tied to taste (sweet, bitter, umami).

My hypothesis is simple; if proteins are made of amino acids, and taste is partly driven by amino acids, then fermentation should rearrange the flavor of the food itself, and we should be able to track that in the data.

We’ll compare fresh vs fermented versions of the same food, looking specifically at the amino acid profiles. Fermentation is kind of an arbitrary choice, because there are a lot of different ways in which we can alter food, either chemically, physically, or biologically.

Structural look: fresh cucumber vs fermented cucumber

To start, we should look at a common example; pickles.

At the macro level, cucumbers don’t change much when they become pickles. And, actually, pickling is a different thing entirely than fermentation; the latter, you pretty much are using lactic acid bacteria. The former, you store in vinegar.

Maybe it’s because of this that the amino acid breakdowns didn’t really differ in the different forms. Looking at the amino acid structure, the breakdowns are pretty much identical.

Despite the above chart, the study I mentioned earlier found that fermented cucumbers contained roughly twice as many total free amino acids as acidified cucumbers. In other words, fermentation didn’t just preserve existing amino acids, but it also created new ones.

Fresh cucumbers were dominated by amino acids like glutamine and asparagine. Fermented cucumbers, on the other hand, showed high levels of amino acids like GABA, leucine, isoleucine, and lysine. This would correspond to the tastes of bitterness, mild sourness, and a savory depth.

On the micronutrient context, understandably, the raw version of the cucumber could be said to have ‘healthier’ minerals in it; the pickled version had a lot more sodium.

Comparing Cabbages

Raw cabbage is one of the examples of foods that have a few variations across culture; kimchi in Asia, and sauerkraut in Europe.

Now, cabbage and its variations don’t really have that much protein, and finding the amino acid data was difficult. So, for this example, I wanted to focus on the nutrients

By itself, cabbage itself is pretty healthy.

There’s quite a lot of vitamin C in cabbage. Sauerkraut also ends up with a decent amount of vitamin C, and there have been stories of sauerkraut ‘curing’‘ sailors of scurvy.

Kimchi is the same deal, with the added complication that it’s not just fermentation doing the work. We’re also adding garlic, ginger, chili, which all bring their own nutritional contributions to the party. Per 100g, less of the detected minerals are related to healthy vitamins.

The real win for both sauerkraut and kimchi isn’t vitamins, it’s the probiotics. I couldn’t show the data here, but gut bacteria develop through fermentation, as does lactic acid. Both offer benefits.

Comparing Milk vs Yogurt

When you ferment whole milk into yogurt, the bacteria - mostly Lactobacillus - eat the lactose and convert it into lactic acid. Similarly with cabbage, this leads to healthy gut bacteria forming during the process, and it’s what makes yogurt thick and tangy. Lactose-intolerant people can eat yogurt without issue because the sugar that was bothering them basically gets consumed before they ever eat it.

Looking at the macros of milk vs yogurt, the two have very similar profiles, with yogurt having additional fat.

The premise of this blog was the taste profile, and luckily, the data was relatively accessible for these foods. I had to use a nonfat milk for comparison, but hopefully it wouldn’t have too much difference between the amino acids in whole milk vs reduced fat milk.

Milk and yogurt have relatively similar amino acid breakdowns, but the values are different enough for my algorithm to change their placement in a voronoi chart.

Whole milk and plain whole milk yogurt have roughly similar B12 content per 100g, fermentation doesn’t dramatically create more of it. The more interesting angle is that dairy in general has one of the highest B12 absorption rates of any food.

In terms of minerals and nutrients, yogurt seems like it has a little more per 100g, but that may be because of how things like that are calculated. Both are great for calcium.

Why fermentation tastes “deeper”

Different amino acids map to different taste sensations: Glutamic acid → umami, Glycine → sweetness, and Leucine & isoleucine → bitterness.

During fermentation, lactic acid bacteria actively convert amino acids into new forms. And this explains why fermented foods tend to taste more complex than their fresh counterparts.

There are so many other places I can take this blog, so I probably will do a part 2.

Plans to build data centers have been rejected all over the country and over my feed, from New Brunswick, NJ to San Marcos TX, to Coweta, OK.

Being honest, I don’t know how to feel just yet, but its obvious there are two sides of the battle: the techies, who believe that ai is a net good thing we should invest in, and everyone else, who, understandably, don’t welcome the incoming pollution and designation to ‘permanent’ underclass.

So, I’m gonna break down as much of the numbers as I can to help make sense of this thing.

Possibility #1: Electricity

What are data centers?

For most things on the internet to work, including AI, computers need to send requests to a computer server somewhere to run the calculations. Data centers are just massive buildings that hold a bunch of computer servers.

Because data centers are a large continuous load, they run on grid electricity, and that is much of the big debates surrounding them. Currently, the US grid is powered by a few different sources, from solar to nuclear to natural gas. Which comes the first concern; to power these centers, we will probably need to rely on ‘dirty energy’.

Now, compared to agriculture, which is responsible for 1/4 of the world’s greenhouse gas emissions, data centers are responsible for somewhere between 0.5 - 4% of carbon emissions. And, there’s only so much we can improve our renewable energy capabilities in the next few years, especially since the New York Times is forecasting that the demand for electricity for generations willl double in the next few years.

Why does a data center need so much electricity?

There are three main reasons data centers need electricity: cooling, server and IT equipment, and power distribution infrastructure (lighting, UPS systems, etc.). The largest shares go to IT equipment and cooling, but everything depends on something called the PUE, which is basically the total facility power ÷ IT equipment power.

It changes from data center to data center, and the industry average has a PUE of 1.56, which would mean that for every 1 watt going to IT, 0.56 watts go to overhead (cooling, power distribution, etc).

Let’s do some math.

Most typical enterprise data centers consume somewhere in the range of 120-240 megawatt-hours (MWh) per day (equivalent to drawing 5–10 megawatts continuously).

If the average U.S. home uses about 863 kWh/month, a regular data center, at 5 MW, would use the same amount of electricity as over 4,000 homes.

Hyper-scale facilities can sit closer to the 100 MW range, reaching a 100 MW continuous draw; 100 MW an hour for 24 hours a day equals ≈ 72 million kWh/month, or about ~83,000 homes’ worth of monthly electricity. Yikes.

As I said before, each data center has their own PUE, and some companies with resources have gotten better at allocating more power to compute rather than cooling. Rod McLaren created a neat little list on his blog.

How much electricity does AI use, specifically?

The recent increase in data server attention is specifically because of AI, so let’s do some calculations with that in mind.

The data is murky at best, because its often self-reported. There was an old Google claim (from 2009) that each search would cost about 0.3 Wh. And Sam Altman says chatgpt costs 0.34Wh per query. Though, if I were to guess, Google’s energy usage has probably decreased by quite a lot since 2009).

Epoch.AI created this graph that breaks it down smartly, and they actually expand on how they got to those numbers here and in this spreadsheet.

Possibility #2: Land Use

There have been proposals to have data centers in both rural areas, due to cheaper land, and more urban areas, due to proximity to people and shorter bandwidth distances. This map by the Financial Times does a decent job of showing the current landscape for data centers, both current and future.

For land, there are already documented tensions in places like Virginia and the midwest, where data centers are competing with land that could be used for farmland. Landowners may also be impacted by 24/7 lighting and noise from the facility.

If you want to read more, this piece talks more about data center land use.

Possibility #3: Water

Something else we need for data centers is water, for cooling and for other reasons. Computers generate a bunch of heat, and often times, need water to cool them down, as well as contribute to other power generating tasks. One widely cited “average” is around ~1.9 liters per kWh.

A typical enterprise data center drawing 5 MW would use about 120 MWh a day, or 3,600 MWh/month. At 1.9 liters/kWh, that’s roughly 6.5 million liters of water per month, or about 2.5 Olympic swimming pools of water a month. In a world where we already have a need to worry about water scarcity, its understandable why AI isn’t as popular with the environmentalists.

Interestingly enough, the data shows that it’s not reserved only for AI queries. Compartively, AI queries are less of a strain on water than watching Youtube videos.

Moreso, the problem is with the electricity generated, as in America, anything that you do that uses electricity often also uses water. Up to 73% of U.S. power plants (coal, natural gas, and nuclear) generate electricity by boiling water to create steam, which spins turbines. That process requires enormous amounts of water for cooling.

What also goes unacknowledged is the systems perspective, and the understanding that when you take water from a place, you have to reduce demand or put water back in that same place. Roughly two-thirds of the data centers built since 2022 have been located in water-stressed regions, according to a Bloomberg News article (that’s been debated), and agriculture, the biggest drainage of water in our country, must reckon with that.

Data centers, not AI, is the real strain on our water, and it will inevitable compete against agriculture for freshwater sources. Technically, salt water can be used for cooling, but the corrosive properties of salt mean that companies prefer to pay for freshwater. And we’re already on the decline for that.

Possibility #4: Pollution

Lastly, there is the pollution that stems from data centers, more specifically, from the diesel generators. Now, I have to admit that this threat seems less of a concern because the diesel generators are mostly used for backup; data centers 99% of the time will be powered the whatever US grid they’re on. Still, when they are used, they are awful.

There have already been reports of health consequences because of being close to data centers. Data center generator pollution can rival power plant emissions in some scenarios. Not to mention, these data centers are construction projects. And, as stated in this NYTimes piece, its possible for these projects to pollute groundwater.

Something that might be a ‘positive’ pollutant (oxymoron, i know) is heat output. These data centers computers run extremely hot, and data centers emit significant waste heat into the surrounding environment. Some farmers have been capturing that energy and repurposing it for their farms.

There’s emerging research on localized temperature increases near large facilities, which could affect microclimates relevant to nearby farming, as well as things like noise that may also disrupt farm-life. Outside of that, the carbon emissions from data centers, and the technology industry as a whole, is quite small compared to other industries. Agriculture is about 11% (we need food so I’ll allot it).

For comparison, agriculture is responsible for about 11% of global carbon emissions.

I want to be one of those people who eat only natural ingredients. Someone who loves honey, and who never eats processed sugar, and who optionally uses deodorant. I channel some of these elements currently, but not nearly enough so it’s not a lifestyle I can say I’m living fully.

I used honey as an example because, if someone were to ask me what the best replacement for refined sugar would be, I would answer honey. Honey is nature’s sweetener, and compared to refined sugar - one of the most consumed sweetener in the us - it seems like it would have more natural ingredients.

But maybe not. I found out that maple syrup could also be a decent contender….

Here’s another data dump for your head top.

A look at the macronutrients

In my experience of doing these things, the more water a food has, generally the healthier it is. And honey is diluted with 17% water.

I compared honey with maple syrup, which has even more water. It’s funny because I’ve always thought the two to have similar consistencies, but maple syrup does end up being a little more thin when poured.

Superficially, honey has a tiny bit of protein, but it is not a significant amount, containing only trace amounts. Maple syrup has a similar thing with protein and fat; by weight, they don’t make up much of 100g. There is a lot more starch in maple syrup though.

And, of course, I wanted to compare the different types of sugar. Sugar is sugar, of course, but brown sugar ends up with just a little more moisture than granulated sugar (eg, the one most of us buy at the store).

When looking at the macro breakdown of all four at once, it’s fun to compare and see the exact differences. Unsurprisingly, the syrups have more water (aka moisture) than the sugars.

The case of high fructose corn syrup

If you live in America, you probably know the curse that is laid upon us in the form of high fructose corn syrup. Propped up by the US industry due to our uncanny ability to farm corn, 40% of caloric sweeteners added to foods is high fructose corn syrup (HFCS), and it goes neck and neck with refined sugar with how much, per capita, Americans eat it.

There’s this crazy analysis published in PNAS found that 100% of the chicken and 93% of the beef served at McDonald’s, Burger King, and Wendy’s had been fed exclusively a corn-based diet. So, unfortunately, a lot of us are eating corn syrup, whether we want to or not.

On the macro level, HCFS looks very similar to a syrup or a honey, with its division between sugars and water.

The case of syrups and liquid sweeteners

I wanted to add a few more sweeteners to the analysis, those that don’t really get as much shine. I created a chart to compare the sugar-to-water ratio in liquid sweeteners; the goal was to see if we can compare a sweetener’s ‘sweetness’ (Even though there are probably a few more factors at play).

I’m a little surprised at fruit syrup being the outlier with its starch content, but the rest of the figures look very similar. And it didn’t hit me till making this post that there are quite a lot of trees and crops that have their own types of syrup.

How healthy are the different sweeteners?

Micronutrients are a decent indicator of nutrition; they don’t tell the whole story, but they tell enough hopefully for us to get an idea.

Now, it’s important to note a few things: 1) nobody is every eating 100g of sugar or 100g of honey at a time; this was just the best way to compare these foods on the same scale. And 2) for whatever health benefits you get, understand that you are still eating large amounts of sugar. Don’t get carried away now; you’re probably still losing more than you’re gaining.

When it comes to micronutrients, potassium and calcium seem to be the minerals offered most from sugars and honeys.

In agave syrup and high fructose corn syrup, sodium and selenium are more present. This might be because they’re industrially processed from plants grown in mineral-containing soils, and trace elements from the raw material, water, and/or the processing steps can carry through into the final syrup.

Not all sweeteners have the same amount of micronutrients, so here is the same chart above, just scaled to represent proportions.

Now, just the raw amounts of the micronutrients isn’t enough to tell the whole store, but it is telling that high fructose corn syrup pales in comparison to the rest of the other sweeteners.

Something more useful than the mass amounts might be something like these radar charts that show the Daily Recommended Intake for a few micronutrients.

Again, this is checking every 100g, so take it with somewhat of a grain of salt, and maybe consider that a tablespoon is about 20 grams. So reduce it by 5.

Still, ever with that caveat, you can see that maple syrup is truly the only sweetener that at least offers something in addition to the large amount of sugar. Honey and Brown sugar both offer modest gains of things like calcium iron, etc. But, among the calculus for which sweetener to pour in your coffee or tea, maybe maple syrup’s lack of sweetness can be made up for its nutritional value.

But, if you ask me, maple syrup in coffee sounds fucking disgusting.

I never really cared for the idea of eating liver, and the texture is a littler too tender for me, but even I’ve heard about the benefits it gives you. What exactly are those benefits? I don’t know, that’s why I’m writing this blog dummy.

I’ve had it once or twice, and from what I can tell, people try to make it smothered in something, or cooked to oblivion. This ends up causing it to go in two directions; chewy or tender. Foie gras is a pate’d example and it tastes ok, but I think the word is just more fun to say than it actually being good.

Also, they force-feed ducks; what the fuck.

Anyway, this post is meant to collect some graphs and talk about liver as a concept.

What a liver represents biologically

All livers do the same jobs, basically; they store fat-soluble vitamins, they process toxins, and they manage energy. They’re very important, and there’s a good reason why health declines rapidly if livers are screwed.

By mass, most parts of the body are mostly water, and the same case goes for liver. There’s some protein in it, of course, but the rest of the distribution is what you expect.

The reason why livers are so healthy is because they store so many vitamins. They also detoxify our bodies, and you would think that eating it would be a base for contamination. Luckily, it only detoxifies us, and all the nasty stuff passes right through it.

In land animals, livers have similar makeups to each other, but you can see the amount of vitamins (niacin, riboflavin etc) relative to minerals (iron, copper) are higher than in other parts of the body.

I couldn’t find the protein breakdown for beef liver, but I did find the one for lamb liver, and I’m taking the liberty that they are more or less the same, with some variation to the relative proportions.

A lot of livers are high in glutamic acid because it sits at the center of nitrogen metabolism, and the liver is the body’s nitrogen-processing factory. When amino acids are broken down, their nitrogen becomes glutamate, and the liver is constantly using glutamate as part of its everyday work, and its proteins reflect that.

Not all livers are created equally, and nutrient differences in a cooked liver reflect on the life of the animal, usually their metabolism and fat storage energy. For example, fish liver tells a completely different story.

The case of Fish liver

Fish liver is a lot different than other types of liver because it has two dual natures; rather than only serving the metabolism of the fish, it also has a parallel goal of storing fat, which reflects its necessity for cold-water survival and need for energy storage.

For land animals, we store our fat in other places, and our liver’s priority becomes our metabolism, and making sure our nutrients power our energy. For lots of fish, a lot of fat is stored in the liver due to their reliance on long-term fat. The difference isn’t that fish use fat more than land animals, but that many fish store much more of it directly in the liver.

When looking at radar charts of the liver of these animals, you can see that land mammals just end up having a lot more B vitamins than the fish animals.

And then, of course, comes the case of foie gras.

The case of foie gras

Foie gras is the artificial fattening of liver. It’s controversial, obviously, and honestly it doesn’t taste all that great for how extra evil we gotta be to eat it. But honestly, because its artificially fattening, its nutrition pales in comparison.

Whereas most other livers have a healthy dose of choline (and niacin, in the case of fish), foie gras’s is a lot less concentrated. In fact, when weighed to 100g, a lot of its weight in minerals come from its sodium content.

It truly doesn’t seem worth it.

Let’s talk about nuts.

The last data-driven comparison piece on milks was a fun one to make, so I wanted to replicate the format for nuts, another food with lots of varieties.

Personally, I like pistachios the best, cashews being the second, and I hate pecans. But the fact that I’m eating nuts in somewhat regular fashion is the most important thing, because they’re pretty healthy.

Breaking down the nutrients

First things first, I created the sankey diagrams for some of the more popular nuts.

Compositionally, most nuts are high in fat and/or protein, and generally have a fair-ish distribution between the three macros. The water content sits usually between 2 - 5% for each nut.

In all the nuts I saw, monounsaturated fats seemed to be the predominant type of fat, which doesn’t necessarily mean anything too dramatic, but we can reason that most nuts can be framed as ‘good-fat’ foods, especially in discussions about heart health.

Pecans and macadamia nuts are especially fatty (20 grams more than the average!), which could be part of the reason they are so widely used in desserts. In those cases, the mass seems to distribute away from protein and carbs.

The increase in consumption

Relatively speaking, the nut market is pretty diverse, thought there are like 5 or 6 ‘super nuts’ that are consumed more than others. In 2016, the average American consumed twice as many nuts as the average American in 1970, and most of that increase has been because of almonds.

Personally, I’m most surprised by walnuts. I did not think they were that popular.

The most interesting thing I learned writing this piece was the fact that California is America’s tree nut factory; they are responsible for producing around 80% of the worlds almonds, and 99% of the country’s walnuts. Since the 90s, the benefits of nuts became more widespread and understood, causing what we’re seeing now; a dramatic increase in the consumption of nuts (pause) around the world.

Pistachios still sit at the bottom, most likely due to cultural reasons than anything else. They’re a little more expensive of a nut, and more so dominated in a few countries (Middle East, Italy) than as widespread.

Land Use

America is not the only country on Earth, and the numbers around the world show that other nuts, specifically the cashew, might be grown more often.

These numbers, I think, are more indicative of how widespread cashew production, and not how environmentally draining planting a cashew tree is. Cashews are highly international, and used in West African cuisines, Vietnamese, Indian, a few more.

Nuts are relatively high-value per acre, so farmers and investors and survivalists love them. I wanted to graph more of this story, especially land efficiency and regional trade patterns, but the data was surprisingly thin. But maybe this influences your next trip to the grocery.

Hey y’all. Hope you’ve been staying safe during the winter storm, wherever you’re reading this.

I’m writing this from Atlanta, and it’s not too bad right now. Been missing the past couple of weeks, so this blog is delayed, but that’s not new. Today’s post is about milk.

Whole milk is an indulgence, and I’m tired of pretending it’s not. Not only am I lactose-intolerant, but also I’m an Environmental Ally™ so I try to do my part and drink plant milks on the regular because of their recorded higher carbon footprint than other milks.

Milk is one of the few foods where substitution feels obvious and incomplete at the same time, so here’s a data roundup of all the

Which data did I use?

This should be no surprise; I am getting the information for the USDA database. Specifically, I am getting the data from the Survey (FNDDS), a database that provides the nutrient values for foods and beverages reported in the National Health and Nutrition Examination Survey.

How does milk differ from its alternatives?

The biggest thing for me is the texture of milk; I like the creaminess, and while I know there are some dubious reasons why that texture is so great (hint hint, there’s blood and pus in the milk), I also know that a big part of is it because of the fat content.

When comparing the treemaps for Whole milk and Almond milk, you can see the difference pretty prominently.

With almond milk, most of it is water. Actually, in fact, there are controversies in its cultivation specifically because its high water usage. So, naturally, it makes sense that almond milk would be mostly based in water.

This extends to soy milk and oat milk as well. Both sit at about 92% water, while whole milk’s is 88%, and almond milk is 97%.

Oat milk, arguably the closest in texture to whole milk, does have the considerably more fat and less protein than soy milk, and one could assume more sugars make it thicker as well.

Which milk is the healthiest for me?

Because I consume so much milk, part of my preference for a type of milk is based on health. Ultimately, all of the milks offer modest benefits, and are good in Vitamin B12 and Calcium. Unsurprisingly, Oat milk and Almond milk are two of the less nutritious milks.

Milks are better than each other in specific contexts

At the end of the day, milk isn’t good nor is it necessarily bad. Different milks have different purposes, so it really is a matter of preference. There are a couple of stats on how much resources different milks use listed below, and I actually already wrote a piece about this, with an interactive tool that offers more nuanced comparison. I recommended reading that until I have the time to write the version 2 of this piece.

I love BLTs, but people love hating on me when I get one, so I’m forced to write this fucking blog investigating on if its valid or not.

To be fair, bacon is talked about like a protein-first food, but it doesn’t always act like it. When I cook, I use it for protein, but it competes with eggs as the main thing of the meal, whereas most other proteins easily dominate their plates.

That made me curious: when you actually break bacon down by data, what is it really made of?

How is data going to help me solve this?

I’m kind of a one-trick pony, so I’m breaking down bacon’s nutrient data into its core components (protein, fat, water, carbs). I plan to then compare that breakdown with other foods that sit on the line between meat and something else. I’ll be using the USDA FoodData database.

The ‘other foods’ that I compare it with are animal products that might not necessarily be protein-dominant, but more fat-dominant. I hate to say, but I’m kind of mindless when it comes to what I’m eating, and naturally assume that foods like pork belly or sausage are basically protein.

What does the data say about bacon?

Most of bacon’s weight comes from one of three sources: fat, water, and protein, and there lies my first issue with its categorization as a protein.

Pork bacon has more fat than protein. There are fatty cuts of meat, for sure, but there are also fat products - like foie gras - that we segment into a different category, as seen in the diagram above. And bacon looks like it’s closer to this second category than the first.

This happens not only in the data, but when we cook. Structurally, bacon behaves more like a seasoning, a texture modifier, and/or a flavor amplifier because of the high amount of fat it holds. This is different from most other sandwich meats, which acts as the main star of a meal and have the flavors built around it.

Rarely is a whole meal focused on highlighting just the bacon….except in a BLT.

Other Foods That Sit on This Edge

Bacon isn’t alone here.

Foie gras, for example, is technically liver, and structurally, it’s dominated by fat. It behaves like a spread or a fat-based condiment, which is on the opposite side of the spectrum as a chicken breast. Calling foie gras a “protein serving” feels wrong.

Chicken skin are another edge case. They’re often categorized as poultry or protein, but their behavior is driven by collagen and connective tissue. Skin is about crispness, fat release, and texture contrast, not protein density. When I put them all on a graph, bacon looks a little better in the light; it has significantly more protein than the examples I just listed.

Across all of these foods, the pattern is consistent: protein exists, but it isn’t the dominant thing deciding behavior. When looking at the data, the food’s function diverges from its category.

What This Breakdown Doesn’t Explain

Categories are cultural shortcuts, and datasets inherit those shortcuts, which unfortunately means that we get a situation where bacon is actually not a fat nor a protein, but some weird third thing. If we were to run a clustering algorithm on bacon vs the rest of the proteins, it might sit just at the end of the two.

Unfortunately, this analysis does prove me wrong, as bacon is pretty fat-forward and salt-heavy food; it just happens to include protein.

I’m still ordering BLTs.

When I imagine or hear about something being 95% water, call me dumb, but I imagine it being something wet.

I hear about things like watermelon and cucumbers being 90+% water, and I have no choice but to conclude that water content doesn’t decide texture on its own. The proteins, the starches, the sugars; clearly they have a role as well, so that’s what I want to look at today.

What else are we digesting when we eat cucumber?

The data I used, and how I looked at it

This should be no surprise that I am getting the information from the USDA FoodData database, a government-hosted database of every food item, and their nutrients. As far as I can tell, it’s the most comprehensive open data database on food nutrition.

Specifically, I am getting the data from the Survey (FNDDS), a database that provides the nutrient values for foods reported in the National Health and Nutrition Examination Survey. It lists out all the nutrients, and their mass amounts.

When investigating something, my goals start with decomposition; breaking a complex system into simpler components to understand how it’s assembled. In the instance of food, representing those components as proportions will give me a fast read of the data, so I’m gonna go with a sankey diagram to start.

What is in 100g of a cucumber?

To start, we can look at the sankey chart of how 100 grams of cucumber breaks down, and you can see, according to other USDA FoodData database, it’s 95% water.

The second largest category is Carbs, which makes sense. Most fruits and vegetables are basically water + carbohydrates. Minerals are present too, but they barely register by weight, even though they matter biologically.

How do cucumbers compare to others foods?

So, most foods have water, but different types of foods have different amounts of water, and its mostly aligned by food-type.

Fruit and veggies fruit lies between 70 - 100%. Most proteins are between 50 - 70%, and most nuts are below 10%. Rarely will you see a dry-ass fruit, unless it was intentionally configured that way.

Luckily, the high water content found in cucumbers - and most fruits and veggies - make them healthy to eat. Water is great for us, and cucumbers are a great way for us to satiate our appetites while essentially ‘eating’ water.

Why eat cucumbers if they are mostly water?

So, cucumber has a lot of water. That’s not new. But that’s not why I wrote this piece.

You see, cucumbers, being a fruit, is water + carbs. And carbs are important because a portion of them show up as fiber.

Cucumber fiber, like the fiber in many fruits, arranges itself into pressure-resistant cell walls that help the tissue resist being crushed (which is important for fruit to protect their seeds inside.) These cell walls are what give cucumber its structure, and help maintain turgor pressure, aka the outward pressure of water pushing against the cell walls.

What I discovered is that turgor pressure is delicious.

Turgor pressure is responsible for the crispiness and the crunchiness of certain vegetables. So, when we eat cucumbers, we are also ‘eating’ turgor pressure.

Anything else to know about cucumbers?

Outside of a few modest gains of certain vitamins (Vitamin C, Vitamin B1, etc), cucumber’s biggest advantage is probably its Vitamin K offerings. Vitamin K is great for the growth and regulation of cells.

So, safe to say…alongside eating ‘water’, when we eat cucumbers, we are eating vitamin K with a little crunch, thanks to turgor pressure. And honestly, that’s not too bad.

Ok, so the name of this week’s piece is kind of extreme, but as I’ve been diving deep into nutrition and food science, one area ripe for ‘research’ is just the straight up comparison between two foods. So, today, I decided to focus on seafood. Specifically, sushi fillings.

Also, after that, is an interesting chart about the books men and women read and review, according to Goodreads.

Happy Holidays!

What is in 100g of salmon vs eel?

When you’re at a restaurant and you’re indecisive on which sushi role to buy, do you go for the eel or the salmon? Both taste good, both come from the sea, and both offer differ textural and nutritional experiences.

Those differences, of course, come from their difference in composition.

Here’s a sankey chart of salmon and its macronutrients.

Salmon’s great because 100 grams of it has 20 grams of protein. And although salmon is 11% fat, the fat seems to be evenly balanced along saturated and unsaturated fats, making for a stable omega3-omega6 ratio.

Not surprisingly, the composition of eels looks very similar.

Eel has slightly less protein, and all of its fats seem to be focused on the monounsaturated variety; monounsaturated fats are especially good at lowering LDL cholesterol (the bad kind) and raising HDL cholesterol (the good kind).

Which is healthier?

Obviously this analysis is a bit shallow, as the mass of these nutrients do not necessarily account for the complexity of a food’s ‘healthiness’. However, in general, salmon offers more varied nutrients than eel, and the difference lies in their minerals and vitamins. Here are bar charts that show the percentage of our Daily Recommended Value met by 100 grams of each food item.

You can see while both are great sources of Vitamin B12 and Vitamin D, eel edges out in Vitamin A (because of how fatty it is, there’s a lot of place to store fat-soluble Vitamin A). Salmon edges out in the other B vitamins (B3 and B6), which are more related to the body’s metabolism.

Which tastes better? Eel or Salmon?

Taste is truly a subjective metric, but I think it’s ultimately the deciding factor in how people choose between the two.

Obviously, part of the reason the two taste different because of their fat amounts. As we saw above, eel has much higher fat, so its richer and oiler. Salmon, being leaner, has a cleaner taste.

However, something that would be less obvious is protein. Within proteins, the amino acid composition strongly influences perceived taste, because different amino acids map to different taste sensations. For example, things like Glycine is a little sweet, while Leucine is bitter, and Glutamic acid (glutamate) is umami.

I won’t do the comparison between the two, but above is the protein breakdown of pink salmon, just so you can get an idea of what we’re talking about.

Salmon vs Eel: Who wins?

Obviously, this was an exercise done tongue-in-cheek, and truly both options beat each other out in certain things. The true answer with foods, like most things, is variety. But hopefully these charts give you a better idea of what you’re eating when you eat these two. I wonder who will be the next competitors of this deathmatch series…Any suggestions?

Male vs Female Books

As a bonus, here’s a chart that takes all Goodreads reviewers by genre, and separates them by sex. I found it on twitter by the Alex & Books account, and it could imply women read more, and more varied, books than men, as most of the genres have a substantially larger percentage of reviewers who are women.

Well, turns out, women do read more than men. A lot, in fact. And across borders. There is apparently a ‘global book reading divide’ happening right now, adding to the growing divergence between the experience between boys and girls in today’s society.

But, more importantly, I think it’s funny that the genres men read are kind of predictable (comics; who would’ve thunk that?)

Understanding food’s carbon footprint

The most interesting graph I’ve seen this month covers the environmental costs of moving food around.

It comes from Our World in Data, a great site that tries to measure humanity in numbers. There are many reasons why producing food should and does have a certain environmental cost, but moving food around is so easily definable that it’s tempting to analyze and even overanalyze it.

Turns out that very little of global food is transported by air, and for a good reason; it’s expensive environmentally-speaking. So, there is at least some optimization happening when it comes to transport.

Actually, technically speaking, while transport does have a noticeable impact, it’s not nearly as emitting of a stage in the food cycle than other phases. Farming and deforestation are more destructive, and those come from certain foods that we can all try to avoid a little more.

I guess I find it surprising because it makes the ‘buy local’ slogan have a little less bite. For sure, we should try to eat what’s around us, what’s in season, and wha historically our communities have been eating. But, if we’re trying to buy local to save a tree, we’re better off just not eating red meat.

A pictorial map of the world

I personally am mixed on illustration in maps; it helps make finding important info easier (eg Statue of Liberty for NY) but I can’t help but get distracted when I’m looking at them. They’re more art pieces than functional pieces of information, which is fine, but it’s a balance I’m very particular about.

I saw this one, sponsored by the USGS called ‘The Geologic Time Spiral’. It looks cool, don’t get me wrong, so I’m inclined to like it…but I feel like I gotta zoom in on this on a big screen to really understand what I’m looking at.

I found it browsing the David Rumsey Map Collection, and I do think it belongs in the conversation of data viz. As automation is making it easier to simply create line charts of everything, I appreciate these more unique perspectives of what a chart or a map could be, so I’m happy to propagate.

Visualizing Food Nutrition

I spend a lot of time coming up with new ways to tell the story of food, and it’s hard not to default to nutrition-based topics. Data viz works with numbers, and there’s plenty of bar charts and nutrition labels that can explain how many macros and micros a pice of chicken has.

The issue is the fact that it’s not the most exciting-looking, and, because of that, I feel like its more normal for people to not know about their food.

My code sketches leave me back to a circular shape; mainly because it’s pretty, but also because its easy to inscribe visible patterns into circles. Explaining what a chart is showing is tricky, and the wall I’ve been running into. But I’ve made radar charts and radial diagrams and other variations of the same idea, an idea that’s closely related to color. The hope is that making things ‘prettier’ can convey information on its nutritional value.

This section is more of a reflection more than anything, something I hope to look back at after I’ve done a few more experiments. Right now, I want to continue to lean into color and vibrancy as a way to label food as ‘good’ or ‘healthy’, but that requires a more sophisticated coding and understanding of color theory. Real design nerd shit.

But! I’ve gotten back into what may be a daily routine, so here’s something new I’m having a lot of fun with.

Yes, is the short answer, but like everything in nutrition, there's more to the story.

Fertilizer can be organic (manure, compost) or synthetic (made from chemicals), and its the latter that has been called into question, because its one of the things that's killing us (source). More specifically, synthetic fertilizers containing nitrogen can lead to water and environment poisoning via nitrate poisoning when they leach into water bodies. Elevated nitrate levels can lead to serious health issues, and has already been linked to colorectal cancer, thyroid disease, and neural tube defects (source).

The EPA has already set a maximum contaminant level of 10 mg/L for nitrates in drinking water to protect against some of these health effects (source), but with such negative consequences, it begs the question as to why we are still using them in the first place.

Why do we use fertilizer in the first place?

Fertilizer use has grown since the mid-20th century due to the Green Revolution. Some people remember the revolution as a time when agricultural productivity increased dramatically worldwide, particularly in developing countries. I also remember it as when countries had some chemicals left over from World War 2, and decided putting it into our food was a good way to go (source). If you're curious, I've written before about the mid-20th century and its invention of fortified foods, and some of the health effects have come about its cause (source).

Short answer: there was a boom in population, which led to a demand for food, and the world responded with different technologies, fertilizers and pesticides being some of them.

The stereotype is that America is notoriously bad at fertilizer and pesticide usage, and that is kind of true. In reality, the whole world is on this shit, but the US does uses more than twice the amount of fertilizer per capita than the rest of the world. Even Europe tapered off in the 80s and 90s because of the environmental consequences.

While fertilizers make crops grow better and faster, in an average agricultural field, only ~50% of the nitrogen from fertilizers is used by crops [4]. That equates to about 12 million nitrogen elephants (~165 billion pounds) every year. And this is where the problem lies; the lost nitrogen can end up in the atmosphere or in waterways, such as groundwater, streams, lakes, rivers, and oceans (Figure 2), and this lost nitrogen causes a variety of environmental issues that lead to those health outcomes listed above. [2].

Domestically, there was a sharp increase in the 60s and 70s, and our use has eventually stabilized since then, with a slight dip during the 2007 economic crisis.

Is there a way to avoid fertilizer?

Avoiding nitrate contamination depends more on where you live and the quality of your water supply, as nitrates can still leach into groundwater and surface water from nearby conventional farms. Because of this, rural and agricultural areas are often at higher risk. There's a map on the USGS website that allows you to see current water systems in the US that have nitrate violations (and organizes by state, as well), and currently, there does seem to be a slow increase in nitrate violations over time in the US (source).

Eating organic food may reduce your exposure to synthetic fertilizers because organic farming prohibits the use of synthetic fertilizer, so that's +1 point to the organic folks out there.

Something else that should also help is filtering out the water, specifically with a reverse osmosis system and water bottles (source). According to a large scientific review of 2,743 studies, reverse osmosis systems water filters can reach nitrate removal rates of 85.03% (source). It won't guarantee our water won't still exceed the EPA's safety limit of 10 mg/L, but its a start.

Lastly, there's small scale farming. With greater transparency on their sources of food and water, and embrace of organic fertilizers versus synthetic ones, your best bet is to buy food from clean places. I used to be an avid composter (still plan on doing so), and that's one of the first things to come to mind. I leave you with the same challenge I'm giving myself: learn more about growing food.

The U.S. Food Dollar is a metric developed by the USDA to better understand the costs of food spending. Essentially, it takes the total amount Americans spend on food, and segments how each dollar is distributed among the various industries along the supply chain. For instance, food service establishments received a share of about 31.5 cents for every 'food dollar' spent in 2023.

Here's a full screenshot of the USDA's version. I color-coded it to make it more clear, but then I got carried away and decided to make a new design, one that's a little more interesting to look at. Pick your fav.

The food dollar is exactly the type of mnemonic device I love, so it kept my attention these last few weeks as the Thanksgiving holidays came around and gaslit Americans (me included) to spend even more money on food. Looking at the 12 or so subcategories of the 'food industry', I could see how they could be re-arranged into specific contexts, and there were three subcategories that came to mind in particular as being the most interesting:

• the people associated with the actual production and processing of the food (farmers, processors, etc)

• the intermediaries who essentially connect producers to consumers (wholesale distributors, supermarkets, etc)

• the food service industry; basically the people that serve food directly to the consumers

How much do the Food Producers get?

Arguably, the most important part of the food supply chain starts at the beginning, at the farm and processing levels.

From the start, we can consider the farmers. In the Food Dollar series, farm production is defined as the cost of the raw materials used for all types of food production. The cost includes seeds, fertilizers, pesticides, animal feed, labor, and farm equipment, and most of it goes directly to the farmers. Its important to distinguish it from food processing, the part of the supply chain where we convert these raw materials into consumable items eg, turning raw milk into cheese.

I'm also adding in transportation and packaging into this section of food production because they make up the ability for food to move throughout the supply chain. This felt better than grouping them with the people businesses managing the distribution, and better than grouping them with the people serving the food. Technically, though, they are present in all parts of the supply chain. The same can be said about Energy.

According to my grouping, the actual production part of food takes about a third of the total expenditure, and has become slightly cheaper since 1993.

The most noticeable insight is the decrease in cost of food processing; technological advancements probably account for most of it, though I wouldn't be surprised if the changes in consumer preferences and the stink of processed foods had a noticeable effect. That's a story for another blog, though.

The total allocated to the farmers has increased slightly, though much of that increase has happened in recent years, no doubt due to COVID. There has been substantial federal financial assistance to help boost the agriculture sector, but the longer term horizon suggests a dip in allocation until recently.

How much goes to Business Operations?

This next grouping is focused on costs associated with running the businesses that connect producers to consumers; the intermediaries, essentially.

Wholesale trade involves the costs of distribution between processors and retailers or food services (distributors, warehousing, logistics, etc). It's had a steady increase since 1993, as have retailers and insurers.

Retail trade represents the operational costs of grocery stores, supermarkets, and other specialty outlets. This is specifically separate from the food service industry, but does include employee wages of workers in those stores. All the costs associated with these stores are the reason why a loaf of bread at a grocery store and its cost at wholesale.

Insurance should be self-explanatory: crop insurance for farmers, vehicle insurance for trucking companies, and liability coverage for restaurants.

Unsurprisingly, when compared to 1993's numbers, we can see a steady increase in costs for the three categories in this grouping. One can assume the trend that more of the profit seems to be flowing to the distributors rather than the producers.

How much goes to the Food Service Industry?

The biggest increase of share of the food dollar goes to the food service industry.

Food services include all costs associated with preparing and serving food outside the home. This ranges from labor (chefs, servers) to utilities and food ingredients in restaurants or catering services.

Since the 90s - and even longer before then - Americans have thoroughly embraced convenience culture, and many more people have eaten out many more times than they've been doing. The growth of fast-food chains and casual dining spots in suburbs and cities undoubtedly helped cause this increase. And that isn't even doing a full investigation into takeout and delivery; but again, a topic for a different blog.

Overhead and Support Costs

These are indirect but essential costs that ensure the smooth functioning of the food supply chain. It's a bit of a catch-all category that includes anything from property taxes to regulatory compliance to advertising. It's generally decreased in its share of the food dollar.

Ultimately, the food dollar is an observational tool, and what I'm observing is a clear winner in the food industry; those serving food to people away from home. I hate moralizing observations, so I'll leave simply on the note of this graph, which shows how, since the 90s, food away from the home has been overtaking food served at home for a while now.

The 21st century is off to a fascinating start.

Microplastics in Common Food

Recent tests by Consumer Reports on 85 foods show that phthalates and bisphenols are decently prevalent in the American food supply. Here's a methodology of their process.

What are the Reference Doses (RfD?)

Reading the report, the numbers are scary because many of the tested food items have recorded tens of thousands of nanograms of microplastics in the test. But I’d be lying if I said I really know what the actual numbers meant, so I looked it up, and came upon the EPA’s system of setting reference doses (RfD) for certain phthalates. In short, reference doses are estimated daily exposures of a chemical that is likely to be without risk of harm.

We can use that as a reference point.

To better understand reference doses, it’s helpful to walk through an example. There's a phthalate called DEHP, and, according to the EPA, it has a RfD of 0.02 mg/kg/day). That means, everyday, we should try to limit our intake of DEHP should be limited to .02mg of the substance per kilogram of body weight.

The data from Consumer Reports is reported in nanograms, so to further contextualize the report, we should try to convert the mg metric to micrograms (µg), and then convert that to nanograms.

1. Convert the RfD (milligrams) to micrograms:

   God bless the metric system.

• 1 mg = 1,000 µg

• 0.02 mg/kg/day = 0.02 * 1,000 µg/kg/day = 20 µg/kg/day

2. Convert phthalates per serving (nanograms) to micrograms:

   Technically, we need to do one more conversion.

• 1 µg = 1,000 ng

• 20,000 ng = 20 µg (as an example)

3. Calculate safe consumption per day:

• Convert weight: 70 kg = 154.32 lbs

• Safe intake: 70 kg * 20 µg/kg/day = 1,400 µg/day

Thus, if I weight about 155 lbs, realistically, I should be able to eat 1,400 µg/day of DEHP a day without risk, which is about 1.4m nanograms/day. In reality, that’s equivalent to something like a grain of salt. One grain of salt weighs about 60 mg, so converting 1.4m nanograms to 1.4 mg, we get a mass of about 1/40th the weight of a grain of salt. A very, very small amount; no wonder they call it microplastics.

When factoring into how this puts the Consumer Reports in context, luckily, many of the foods don't reach that level on their own. Here’s a brief look:

How many phthalates are in our food?

I think most doctors would agree with me that the less plastics, the better, so while the recorded levels of some phthalates are in the tens of thousands of nanograms, the RfDs for those phthalates are in the millions.

The study tested for 10 phthalates (BBP, DBP, DiBP, DCHP, DEP, DEHP, DnHP, DMP, DiNP, and DNOP), and three chemicals used as phthalate substitutes (DEHA, DEHT, and DINCH). I tried my best to find as many of the RfDs as possible. Thus, for a 155lb person, here are your limits:

• DEHP: 0.02 mg/kg/day = 1.4m ng/day (link)

• DEP: 0.8 mg/kg/day = 56m ng/day (link)

• DBP: 0.1 mg/kg/day = 7m ng/day (link)

• BPA: 0.05 mg/kg/day = 3.5m ng/day (link)

• DIBP: 0.1 mg/kg/day = 7m ng/day (link)

• DINP: 0.059 mg/kg/day = 4m ng/day (link)

When lookin at the numbers, few make it past 35k of any one type of phthalate in one serving. Now, it’s technically possible for someone to eat ~50 servings of a certain food in a day, which would put the number of phthalates over some of the limits, but it would be very unlikely. However, the popularity of some of these food items recorded in the test are very popular, from Coca-Cola to Wendy’s chicken nuggets. It wouldn’t be the biggest surprise to understand how or why microplastics are present in ultra-processed foods.

Consumer Reports also found variations among similar products. The 33,980 phthalates in nanograms per serving in Wendy's Crispy Chicken Nuggets, for example, was more than four times the level in McDonald's Chicken McNuggets. So...I guess that's good news.

Thanks for reading!