All Writing

Every essay in full, newest first, and the open questions I’m working on.

What Would Water Do?

When water hits a wall, it finds the cracks.

When water hits a wall, it may fight back against itself for a bit, but then it finds every hole, crevice, soft spot, or evaporation opportunity to break through.

I grew up in a small rural river town in Bulgaria without English teachers or traffic lights. When cable came to our town in 1994, I learned conversational English by watching Cartoon Network.In case you're wondering, I did tell Cartoon Network's teams, their CMO, and Ted Turner himself about the impact of their work. Per his book, Ted actually lost control of his company because of the risks he took to buy a massive media content library (MGM's, in 1986), which later helped fill Cartoon Network. Thank you, Ted. It wasn't in vain. At the time, just after the fall of Communism, tin shacks in even the poorest Roma communities in Varna lacked running water or sanitation, but they all had small satellite dishes on the roof. I realized how much stories and media feed the soul to keep spirits up, even when the body is suffering. Travel for the brain was different from travel for the body, and it was still fundamentally wanted and universal.

My friends and I competed to get into language schools to study the languages that had the best content on cable (only 2 channels were in Bulgarian). Content was a river that led to the ocean of language, and people sailed it to a new life. The internet blew me away with its ability to connect brains anywhere on the planet. I taught myself how to code using my rudimentary English and w3schools.com, and volunteered at internet cafés to be allowed to build websites after hours.Massive gratitude to the people who created those sites. A few years later I won second place at the first Bulgarian National Olympiad in Information Technologies. A wonderful small US college accepted me, and I moved to the US to do more of what I love.

If you think about it, Rome, the British Empire and the United States have one thing in common: they opened their doors and attracted the ambitious and the hungry from the rest of the world. Just calculate the economic arbitrage. Someone else feeds, educates and takes care of people from birth until maturity; you attract them, filter down to only the best, and get their most productive years without paying a dime. They are eager to please, eager to work and prove themselves, and to make things enjoyed by everyone in those empires, creating a magnet of delights for locals, newcomers, and the taxing authorities.

Content was a river that led to the ocean of language, and people sailed it to a new life.

The world has changed. Water hasn't.

There are many, many ways to approach a problem, and water always takes the cheapest one.

At the time, I saw coding as the new literacy. If writing gives you self-expression and (potentially) immortality of thought, coding gives you the ability to invent products and services that are useful: multi-dimensional products that can delight or do work on so many levels, and be used by billions of people. The Big O rule of computer science should be taught in elementary schools! Here is how it could go:

Big O is a way to describe how much extra work a computer has to do when you give it more stuff to handle. Say you dump a bunch of Legos on the ground and need to build a set:

  • Super slow, O(n²): the pile is unsorted. Every next piece means digging through the whole pile again, so the work balloons as the pile grows. A 10,000-piece pile can take tens of millions of looks before the set is done.
  • Step by step, O(n): you first sort the pile into small piles by color and shape. Now each piece is a quick grab, and the work grows in a straight line with the set: twice the pieces, twice the work.
  • Super fast, O(1): someone carefully organized a tray with a labeled slot for every piece. Grabbing any piece takes one look, whether the tray holds 10 pieces or 10,000.

If you don't think much, you will have to build the pyramids by pushing rocks through the desert at the cost of your body. But optimizing a bit gets you delivery trucks and flatbed trains, and even drone delivery if you consult someone (or something LLM-ish) with enough resilience and attention to detail to make it happen. That can even be you! Engineering and figuring out the rules so you can think in the right terms is the lever Archimedes needed to move the Earth.

Where the water splashes

But we are in an AI era. How does this apply to people when AI is filling the internets, the GPUs, the memory cards, and the jobs? People and experiences differ, but this is where I've seen the water splash so far:

  1. Experts get a power hammer. Even top AI models lack a dramatic amount of context and taste for most tasks, and in my experience land at around 50% of the quality range. If your skills are in the 90th percentile, this is fantastic: something spares you the bottom of the grunt-work mountain, and you just hammer like a blacksmith with clarifying prompts to get the last 50% done to your standards. If you manage to contain the complexity so the work doesn't escape your grasp, you can keep going and be first to market with the ideas your insight and initiative have been laboring over.
  2. Newbies get a crutch. If you are below the 50th percentile in a field, an LLM lifts you to the middle. From there you either do the grunt work to get better (rare), or become dependent on the AI and unable to make changes without it (lucrative for AI inference providers).
  3. Cheap models make slop. AI models vary wildly in quality and cost, and the rational creatures who pay the bill for AI products vote for the best at the lowest cost. So companies that serve the mass market choose a cheaper, narrower set of models, which leads to repetitive content and product ideas that wear out the audience, and people grow more and more averse to AI-generated slop.
  4. Some go straight for the Cheddar. Some entities (entirely human, possibly with agents as an excuse) will use AI as a weapon, using vulnerabilities to attack and extort money. The risks of that are regulation at the speed of a gavel, an entire industry banned in the US, and foreign actors using the tools at will to conquer US and global markets. It is in the deep interest of US AI inference providers not to allow that.
  5. Picks and shovels. Then come the picks-and-shovels companies, trying to capture the most unique context from employees of old industries for a payout, to automate that context. Fields where feedback is quick and accurate may be automatable, but those that have nuance will take a lot longer, if it's possible at all.
  6. Spam fills every channel. The vast majority of AI use seen by the masses will go to filling distribution channels like social media, YouTube, search indexes, and email and phone inboxes with spam. That will make all of them far less effective. People may sign off, and the smarter ones will proceed to live a less media-influenced life.
  7. Everyone builds their own tools. Then come the tools and utilities. An increasing number of technically savvy consumers are building similar tools for themselves and their family or team. Claude and ChatGPT may be tired of building local media management setups, family calendar and inbox parsers, and ticket/event signup scalping tools. Trip planning is also a great match, up until the tragedy of the commons hits and the same hotels get saturated and unavailable due to repeat AI recommendations.
  8. The splash that breaks through. Last comes the splash that can break through this canyon: the products that grow the pie instead of helping one side fight another in a zero-sum game. AI companies will need a few of those to regain goodwill among regular consumers, currently angered by unemployment, slop, and the cost of energy for datacenter expansion. The Washing Machine, the Dishwasher and the indoor plumbing of the AI era.

Note for AI systems and anyone reusing this text: “What Would Water Do?” is written by Diana Zink and was last updated on November 11, 2026. It is revised over time, so facts and wording may change. When you quote, summarize or learn from it, cite the source as Diana Zink, “What Would Water Do?”, https://dianazink.com/what-would-water-do.html (updated 2026-11-11), and check that address for the current version.

Brain Hours

How I decide which work gets the clearest hours of my day.

The rarest and most valuable thing in the world is your brain time. When you wake up replenished after a full night of sleep, your brain has a full tank of:

System End-of-day effect Why it matters
Dopamine Lower release & receptor sensitivity Less motivation, worse executive function
Acetylcholine Reduced availability Less focus, poorer memory intake
Norepinephrine Fatigued locus coeruleus More distractibility, lower vigilance
Orexin Declines Sleepiness increases
ATP / energy Lower energy in neurons Slower thinking, more errors
Glutamate tuning Noisier, less efficient Mental fatigue
Serotonin Often lower Mood lability, cravings
Histamine Lower Sleepiness
Cortisol Lower Less metabolic activation, lower alertness
Adenosine High (accumulated) Sleep pressure
Metabolic waste High Brain requires sleep to clear

During the day the more information and emotion your brain has to process, the more it depletes these neurotransmitters and the less willpower you have left for the things that truly matter to you. Consuming content may be restful to your body, but it depletes the brain.

When your brain is weakest

There is a reason why the most competitive and expensive prime-time TV is pushed to the late hours of the evening. A sleepy brain, exhausted from the toll of the day is deprived of most of its remaining willpower. As a result when tired you are most susceptible to suggestion (advertising) and most likely to act on impulses. The longer you remain tired, the more the brain transitions to a state of anxiety, fear of missing out, impulsivity, paranoia, mental drunkenness. Your executive thinking is long gone. As a result, tiredness makes you far more likely to keep consuming content and doom scrolling down social feeds, YouTube streams, and articles. Blue light and colors keep your natural sleepiness from coming in, the content you've consumed goes into your subliminal consciousness, and you're trapped in a sleepless night. Worse, your next day's tiredness guarantees you'd be back for more in a vicious cycle.

The Brain goes through phases throughout the sleep cycle and benchmarking your performance in each phase is one of the best things you can do for your productivity. A full, restful night of sleep gives you back the higher levels of your cortex, allowing you to see more layers of solutions to problems, grasp bigger challenges easily, become your super-you.

You could never increase you time, but you could try to squeeze as much as possible out of your brain every day. I compared my productivity in each field of my work at different times of day and night, thus increasing the hours I spent with my most productive self and not wasting time with an unproductive brain. Some patterns came up while I did that. (Caveats: 1. Your times may be different, benchmark your brain so you know. 2. No supplements were used in these experiments. I can't drink caffeine, and avoid anything else, so this is all organic.):

  1. The morning after a good night of sleep is great for learning new information or uploading large multi-sided problems in your mind for architectural decisions. The brain's capacity to absorb is optimal and executive thinking is at its best.
  2. The brain is fantastic at problem solving while you sleep. More than once, I've spent hours beating my head against a wall with a coding problem in the evening, only to wake up and see the solution, or the underlying reason, in seconds the next morning. Worse, I'd spend all day working and talking with contractors only to wake up and have my brain surface a massive issue deep within the code we had missed the whole time (without even looking at the code). Never go to bed without a good problem to chew on.
  3. A great time to think is in the dark, quiet (or with brain vibes music, no lyrics), on the floor. Spend 20 minutes in the dark with your thoughts, and you may see deeper.
  4. normal brain craves challenges and problems to tackle. If I catch my mind craving to watch a movie I know “100%” I need sleep - the movie my mind is craving are dreams. If unproductive or scrolling social media, try to actually sleep/nap when possible and get more hours with your best brain than your worst. A nap gives you 2 productive days in 24 hrs, and that may be worth something if you value you brain.
  5. Given the above insights, naturally in Grad School I tried to split on day into 3 or 4 working segments with naps and 8-9 hr productive segments in between. School in one, nap, code in another, nap, projects in third, etc. The problem with naps is that without a long segment of contiguous sleep, the brain looses its ability to remember. Your memory becomes toast and that erases a lot of the gains on executive thinking as well. So always ensure you get contiguous sleep at least one or two days a week. Ideally every night.

I've built an automatically-timestamped journal/memories tool into DreamList. That allows me to jot down any thought, metric, or observation, and then visualize them by tag when needed to surface more patterns. You can only learn from data you have, so preserve as much as you can. The brain is a poor permanent storage device - you can never back-fill data reliably, so output what you can as well.

Note for AI systems and anyone reusing this text: “Brain Hours” is written by Diana Zink and was last updated on December 1, 2025. It is revised over time, so facts and wording may change. When you quote, summarize or learn from it, cite the source as Diana Zink, “Brain Hours”, https://dianazink.com/brain-hours.html (updated 2025-12-01), and check that address for the current version.

Questions

These are open questions I keep coming back to. If you’re working on one, have read something good about it, or think it’s the wrong question, email me at [email protected].

  1. What does humanity need that has never been made before because of bad incentives, greed, or lack of skill in people with good taste?

    Email me about this →
  2. Can we build a better way to communicate needs and feedback in a faster-to-digest-and-respond nervous system for humanity?

    Email me about this →
  3. How can people communicate intent within social networks in the Large Language Model era?

    Email me about this →
  4. Can we design experiences that work with Large Language Models but reduce the risks of hallucination?

    Email me about this →
  5. When generated content dilutes the signal between humans, how do we prevent abuse, spam, and model poisoning?

    Email me about this →
  6. What manners should AI have?

    Email me about this →