I spent three years assuming I was bad at feeding my family. Every failed dinner felt like a personal failure. I’d forget to defrost something, or I’d plan a meal that worked for two people but not four, or I’d get to 5 PM and realize I didn’t have the bandwidth to think about what to cook, let alone cook it.

The pattern kept repeating. And the thing is, I couldn’t see it as a pattern. I just saw individual bad nights. It took a long time to realize the failures weren’t about effort. They were structural.

Four people live in this house. Each one has a constraint profile.

I have Celiac disease. Gluten isn’t a preference I manage. It’s a protein my immune system attacks. Cross-contamination from a shared cutting board or a dusted counter means days of symptoms. The kitchen has to treat this as infrastructure, not as something I remember to be careful about.

My son has ARFID. He won’t eat rice, noodles, quinoa, or most grains. He won’t eat cooked vegetables. His acceptable foods are neurologically fixed. The relationship to food is wired differently, and the list doesn’t expand because someone pushed harder or got creative with presentation.

My daughter has ARFID with the opposite profile. She won’t eat meat. Her acceptable protein list: cheese, black beans, hummus, eggs on a good day, chicken under duress. Two kids, same diagnosis, opposite restrictions.

By late afternoon, the executive function I need to plan and sequence a meal has already been spent. Most days. The capacity to hold a recipe in my head, track multiple timers, remember what needs to happen in what order, and adjust when something goes wrong is a finite resource. It runs out before dinner starts.

These don’t cancel each other out. They compound. A meal that works for my son’s grain restriction still has to be gluten-free for me, still has to offer a protein my daughter will eat, and still has to be something I can plan and execute when my working memory is running on fumes. Every dinner is a four-variable problem.

For a long time, I solved this the way most people do. Willpower. Mental checklists. Trying to hold everything in my head and getting frustrated when I couldn’t. Some nights worked. A lot of nights didn’t. And the nights that didn’t work had consequences that cascaded: dinner was late, which meant the kids were hungrier, which meant the ARFID was worse (hunger doesn’t make restrictive eaters more flexible, it makes them more rigid), which meant less food got eaten, which meant someone went to bed without a real meal.

The pattern was invisible because it looked like normal life. Every family has hard nights. Every parent forgets to defrost something. But the frequency and the compounding were signals. There was no system, just a person trying to manage it all, and the failures kept having the same shape.

I’ve spent my career looking at that exact shape.

I’m a design engineer. For twelve years I maintained an enterprise recruiting platform where I had to keep the system running while rebuilding the layers underneath it. I build AI governance tools that hold evaluation criteria across sessions so the work doesn’t drift. I built a protocol that preserves context when working memory (mine, the model’s, the team’s) drops the thread.

In every one of those projects, the work is the same: figure out what the system actually needs, and build structure around that instead of asking it to be something different. An AI model with an 8,000-token context window can’t hold a 40,000-token conversation. A student who can’t process compound instructions can’t complete a multi-step task in one pass. So you break it apart. You scaffold it. You let the environment carry what the person or the system drops.

I was doing this professionally for years before I realized the kitchen needed the same thing. The household has processing profiles the same way a software system does, the same way a classroom does. Four people with documented constraints, each one needing the environment designed around what they actually are.

I stopped treating dinner as a willpower problem and started treating it as a design problem. That shift is what this series covers.

Part of it is the kitchen: how I organize the space, how I structure the cook, how I use tools (sous vide, AI sequencing, modular meal architecture) to reduce the cognitive load of feeding four constraint profiles from one workspace.

Part of it is the garden: growing food, building growing systems, learning hydroponics and soil science from zero. Traditional homesteading practices combined with whatever technology actually reduces the burden.

Part of it is automation: the systems layer. Task sequencing, externalized memory, home infrastructure that carries what working memory drops. Prosthetic architecture for a household where executive function is a constrained resource.

The through-line is accommodation. What each person can and can’t eat. What the climate and the soil can and can’t support. What working memory can and can’t hold.

I do not have this figured out. Tuesday still happens. The systems make the good days more likely and the bad days less destructive, but they don’t dissolve the underlying problem. Four people still have four different profiles. That part is permanent.

What changed is that I stopped blaming capacity for a structural problem. The kitchen was failing because it was designed for a household that doesn’t exist, one where everyone eats the same food, nobody has a medical dietary restriction, and the cook has unlimited executive function at 5 PM. That’s never been my household.

So I started building, and this is where that went.