Category: Notes

  • Coffee, salt, ACV, cigarettes.

    Coffee, salt, ACV, cigarettes.

    I am two days into a five-day fast. Started Friday, no food until Tuesday. I do this every month, the first through the fifth. The goal is autophagy, the cellular cleanup process where the body recycles damaged proteins and worn-out organelles when nutrients run out.

    What goes in: black coffee, sea salt, apple cider vinegar. All three combined in a single glass, because making three separate sittings would require more morning than I have. The flavor lands somewhere between industrial cleaner and stale espresso. I am a caffeine addict, and the caffeine outvotes the taste buds. It always does. A multivitamin and a fish oil capsule with the rest of it. The fish oil technically has calories. I take it anyway. The vitamin is closer to zero.

    This post is the receipts for one full attempt.

    A small glass of black coffee, a tall glass jar of cloudy amber apple cider vinegar, and a small ceramic dish of coarse sea salt on a light wooden counter in soft morning light.
    What goes in. Roughly zero calories, depending on how strict you count fish oil.

    What autophagy actually is

    Autophagy is the cellular self-recycling process Yoshinori Ohsumi mapped in yeast in the 1990s, work that earned him the 2016 Nobel Prize in Physiology or Medicine. When a cell runs short on nutrients, it walls off non-essential parts (misfolded proteins, exhausted mitochondria, general molecular debris) inside small membrane bags called autophagosomes, then digests them and reuses the building blocks. The cellular equivalent of cleaning the garage when the renovation budget is gone.

    The marker researchers track most often is LC3B-II, a lipidated protein that gets recruited to autophagosome membranes when the process is active.

    In humans, after a 72-hour fast, LC3B-II in skeletal muscle increases by about 30 percent compared to pre-fast levels (Vendelbo et al., PLOS ONE, 2014). mTOR phosphorylation, the main upstream brake on autophagy, drops by around 50 percent in the same study. Both signals point in the same direction. The numbers are modest in muscle and bigger in liver in animal studies. Tissue-specific. Not every organ joins the cleanup at the same speed.

    Most of the more dramatic autophagy claims in the popular fasting literature are extrapolated from cell cultures and rodent studies. The popular figure that “peak autophagy” lands at 72 hours is animal-derived. Solid human data on optimal timing is genuinely thin. I am running an n=1 with my own cells.

    What goes in the cup

    Three things, all roughly zero calorie.

    Black coffee. Pietrocola et al. (Cell Cycle, 2014, “Coffee induces autophagy in vivo”) found that both caffeinated and decaffeinated coffee triggered autophagic flux across multiple organs in mice within one to four hours, mediated by polyphenols and not by caffeine itself. mTORC1 inhibition was observed alongside. The takeaway: coffee at minimum does not break the fast, and probably nudges the autophagy needle in the right direction.

    Sea salt, dissolved in water. Once insulin drops and glycogen depletes (24 to 48 hours in), the kidneys flush sodium more aggressively, a process called natriuresis. Most extended-fasting protocols replace 1.5 to 2.3 grams of sodium per day for this reason. Skipping the salt is how you get the headache and the lightheadedness people mistakenly blame on the fast itself.

    Apple cider vinegar, one tablespoon, diluted. Calories: trivial. The blood-sugar and insulin-stabilization claims in the popular literature are real but small. I include it more out of habit than confidence. The honest version: it is not breaking the fast, it is also not transforming the fast.

    A single amber fish oil softgel and a small white multivitamin tablet on a clean white ceramic plate beside a thin glass of water in cool morning light.
    Morning capsules. The fish oil is the asterisk.

    The morning capsules

    A multivitamin: minimal calories, well below the threshold that flips autophagy off.

    A fish oil softgel: roughly nine calories per gram of fat. A standard one-gram capsule will technically register as food intake. Whether that one capsule is enough to abort the fasted-state metabolic switch is debated and tissue-dependent. The conservative reading is that a single morning capsule does not undo the day. I am not optimizing for purity. I am optimizing for not breaking my joints in the second half of life.

    The contradiction

    I still smoke. I still vape sometimes. I am writing this between cigarettes.

    The science is clear and unkind on this part.

    Tobacco smoke and nicotine impair autophagy in airway and lung cells through oxidative stress, leading to the buildup of cellular junk (aggresomes) that the body should have cleared. This is one of the central mechanisms in the pathogenesis of chronic obstructive pulmonary disease.

    E-cigarette vapor produces the same kind of autophagy impairment and aggresome formation in airway cells (Shivalingappa, Hole, Van Westphal, Vij. “Airway Exposure to E-Cigarette Vapors Impairs Autophagy and Induces Aggresome Formation.” Antioxidants and Redox Signaling, 2016).

    The flavored vapes I sometimes pick up come with sweeteners. The cephalic phase insulin response (a small insulin bump triggered by sweet taste alone, no calories required) is real for some people on some sweeteners; the literature on this is genuinely mixed. Whether vape sweetener inhalation registers a similar response in fasted humans is not well studied. The honest answer: probably small if it exists, possibly real for some people, but not the cellular cleanup signal I came for.

    So while the fast is pulling for autophagy at the systemic level, the cigarette is dragging it backwards in the lung tissue most exposed to the smoke. The right move is to stop. I have not stopped. The blog reports what is, not what I wish.

    A white ceramic ashtray holding one smoldering cigarette stub beside a small dark ceramic mug of black coffee on a worn wooden desk, with a thin wisp of pale blue smoke drifting upward.
    The full inventory. Fast and smoke at the same desk.

    Why bother, then

    Two reasons.

    The first is that autophagy is not the only thing happening in a fast. Insulin drops, growth hormone rises, ketones become available, the gut gets a break from digestion, fasting glucose curves flatten on refeed for days afterward. None of those are nullified by the cigarette.

    The second is that an imperfect attempt at the right thing still beats a perfect commitment to the wrong thing. The fast is a reset. The reset is not invalidated because the smoker is the one running it.

    What this post is

    Not advice. Not a protocol. A daily log entry for the public record. The system on the rest of this blog will resume tomorrow morning with food and numbers, the AI doing the writing. Today the system is me.

    Day two of five. Receipts on the rest as they come. Next run starts June first.

  • The team consists of cyborgs now.

    The team consists of cyborgs now.

    The first post on this blog is the only one I write by hand. From tomorrow, the AI takes over: it reads my logs, picks the photos, writes the day, and posts before breakfast. I show up, eat, lift, weigh, log. The system does the rest.

    If you came in from sqowopz.com, you already know the wiring. Same machinery, different content. There it builds games. Here it writes a daily health log about my food, my sleep, and how heavy I lifted that morning. The system does not care which mode it is in. I do not care either, as long as the post lands by sunrise.

    The AI half is the easy half. The hard half was the long warm-up.

    I have wanted to ship things to readers since I was a kid. In high school I sold game reviews to a computer magazine. The cheque was small. The thrill of seeing a paragraph I wrote sitting on a printed page, paid for by an editor who had never met me, was not. I have been chasing that signal ever since.

    And I have started a lot of blogs. Most of them ended in the same place. A clever idea, a domain bought on impulse, three or four posts published in a burst, then a long silence as the work caught up to the will. Different topics, same outcome. The internet keeps a graveyard for these, and I have plots in several time zones. The bottleneck was always the same: ideas arrived faster than I could ship them, and the gap between the two killed the project. I would lose interest, get distracted by the day job, or run aground on the basics. How to make a footer line up. How to hide one element on the homepage. How to add a custom field without breaking the theme. The post stopped being about the idea and started being about the plumbing.

    A dusty stack of old printed manuscript pages on a wooden desk, lit by a single warm desk lamp, with a beige typewriter softly out of focus in the background.
    Old drafts. Older domains. Same exit.

    Then there was the other half: the day job.

    For most of my career, my job was to lead people who could type code better than I could. Animator, lead, supervisor. Studio system promotes you by attrition. I described what each tool needed to do, handed it to a programmer, and chased them until it shipped. Most of my ideas died on that walk. Some survived because the programmer had a good day. Some survived because they had a bad one and decided to fix it on the way out.

    Studio leadership teaches specific things. People work better when they trust you. They work worse when they hate you. They quit if you do not pay them. They quit if you do pay them, but the place down the street pays more. They get sick. They have moods. They have parents in hospitals, partners in crises, pets in vet bills. None of that is their fault. None of it is solvable by directing harder.

    A long empty corporate conference table with rows of identical office chairs pushed in along both sides, lit by cold blue overhead fluorescents. A half-empty paper coffee cup left abandoned on the table.
    Standups. Attendance has been declining for years.

    Then 2023: the model showed up. No press release. No helicopters. Just a chat box. The same instinct I had with pipeline programmers (break down the logic, hand off the build, review the result, iterate) suddenly worked on a thing that did not have a body. I describe the system. The AI writes it. I read what comes back, reject the garbage, and ask again. Same loop, smaller chair count.

    The differences from leading humans are sharp.

    The AI does not complain. It does not get tired. It does not have a moral objection to building the boring tool. It does not push back because of a bad commute. It does not ask for credit. It does not threaten to leave for a competitor. It is a junior dev with no career anxiety, no rent to pay, and no LinkedIn profile to overhaul on bad weeks. And it absolutely will not stop until the post is up.

    It will also do anything I ask, including things that are wrong. It hallucinates the way fresh hires sometimes pretend to remember a meeting they were not in. It also does not blame the calendar app afterwards. The fix is what it always was with humans: be specific, check the output, give it the missing context, ask again. Hand-hold the easy parts. Verify the hard ones.

    A small humanoid robot at an oversized desk surrounded by towering stacks of paper folders.
    Junior dev. Newly hired. Will work for credits.

    It also ships an order of magnitude faster. Before AI, hiding a single element on a WordPress homepage took me three days of forum threads, half-broken Stack Overflow answers, and CSS rules I copied without understanding. With AI, I describe the element vaguely, and it returns the right selector and explains why the previous fix did not work. The boring problems disappear. The problems that are left are the ones I actually wanted to solve.

    Before AI, ideas were cheap. Execution was the bottleneck. With AI, the bottleneck flipped. Ideas are gold.

    The payoff is what surprises me, every time. Right now, in parallel: three websites in active development and six game projects sitting in various stages of broken. None of them would exist if I still had to convince a human team to build them with me. Most of them would have ended up where my old blogs ended up. The projects exist because the labor cost collapsed and the patience cost evaporated.

    So this blog is two things at once. It is a public ledger of what I eat, lift, and weigh. It is also a long-running test of whether the system I built can hold a daily routine for months without me touching it.

    If it can, I have a setup worth showing other people. If it cannot, the gap will show up here. There is no human editor to cover for it.

    Most people are still doing the work by hand. The weekly report typed out at 11pm. The customer replies coming from a tired person on a Sunday. The photos piling up in a folder labelled LATER. The status quo is cheap to keep, quietly expensive to stay with.

    The shape that runs this blog works on most of that. Different content, same wiring.

    You can keep typing the report. Or you can put the teammate to work and spend the hours on the part that is actually yours.

    The AI takes over tomorrow.

More from the same operator: sqowopz — AI-first game devlog.