Build a household dinner-suggestion app
tl;dr. Task: tl;dr A non-programmer built an app with Codex. More interesting than the process is the implication: we are entering an age of personal software, when any of us can use the real power of computers without needing to be an engineer. Background I’ve spent 40 years in tech (in roles including support, community, and product marketing), but I’ve never been a coder. I’ve worked closely with software engineers doing user experience testing and writing documentation, and I’ve done various kinds of tech marketing. I’ve designed simple web apps and done other coding-adjacent tasks like context-sensitive help, and I’ve dabbled a little in HTML and CSS — but that was the sum total of my coding experience. Until now. For family reasons, I’ve become curious about how I could use AI for personal projects. I started small: Years ago I scanned hundreds of personal letters and papers, using a consistent filenaming scheme: [initials of sender] [day] [month] [year] [page number] Now as I finally begin using these files, I realize this scheme does not help get things into chronological order – I should have used: [initials] [year]-[month]-[day] [page number] Manually renaming hundreds of files was a non-starter. Bulk renaming should be easy to do with basic programming skills, but I don’t have those skills. Brendan does, but he has much bigger projects on his mind and we were both interested to see what could be done with Codex.* I set one of the directories of scans as the home for a new project (yes, I did a backup first) and said to Codex: “these files are named according to the following scheme: MB Month Day Year page# can you come up with a plan to convert the filenames to MB Year Month Day page# – show me the plan without doing the conversion”. Out of an abundance of caution, I asked Codex to do this in steps. After I was confident it was working correctly, I set it loose on that directory, and then more. All the files got renamed and are now automatically listed in chron order. It was a silly little job but it got done quickly, easily, and correctly. I felt… powerful. I’m good at using software, but I’d never done anything like this. The real power of a computer was finally available to me. I wanted to do more. I decided to try using Codex to solve a daily problem in our household. Anyone who has cooking responsibilities for a family knows that “What’s for dinner?” can be a vexed question Every Single Day. We all have to deal with dietary preferences, food allergies, what food is on hand (and which urgently needs eating before it goes bad), the cook’s energy level that day vs desire to go out vs expense of ordering in. For our household, this logic is in my head, supported by a list of fresh and fast-expiring items scribbled on a whiteboard on the fridge. What if I could put all that into an app that could store every factor that goes into these daily decisions and make useful suggestions for what to cook? One that I can tune to the very specific needs and preferences of my household? I have tried a few times with apps from the iPhone store, but none could handle our particular cooking and eating habits. For one reason, US recipes tend to assume that an “easy” recipe starts with a packaged main ingredient like a cake mix or frozen dinner rolls. I rarely used those even when I lived in the US, and in Australia they are often not available (cake mix yes, dinner rolls no). What this app is NOT My app is purely for personal use. I have no desire to sell or distribute it in any way, not even as open source. I am simply experimenting with a new tool to build something for myself in a way previously impossible for me. Publishing this blog post is akin to open sourcing – I’m giving you a “recipe” for creating an app. I’ll tell you the ingredients and the process, you can bake your own “cake” – and then you get to eat it, too. Last Friday afternoon (Feb 27), I sketched the bare essentials of an app: Build and maintain a list of all food we have, noting expiration dates for short-lived products Build and maintain a list of what’s in the garden Build and maintain a list of dietary restrictions and preferences in the family/guests Build and maintain a list of favourite restaurants to eat at or order in Build and maintain a list of cooking appliances available Check the local weather – when the outdoor temp is >25C I do not want to run the big oven or stand over a hot stove very long.; tools: Codex. Human role: The original post does not say how a person stayed involved.; limitations: The workflow requires a ChatGPT Pro or higher subscription to access Codex. The author noted that Codex can be overly accommodating, leading to feature creep and unstable code if scope is not strictly managed. Manual intervention was needed to restart the project when bugs became frequent. The app is for personal use only, with no distribution or commercial intent. Integration with external recipe APIs (Edamam) incurs a $9/month cost..
Tools
Published on Agent Work Atlas:
Overview
tl;dr
Workflow
tl;dr A non-programmer built an app with Codex. More interesting than the process is the implication: we are entering an age of personal software, when any of us can use the real power of computers without needing to be an engineer. Background I’ve spent 40 years in tech (in roles including support, community, and product marketing), but I’ve never been a coder. I’ve worked closely with software engineers doing user experience testing and writing documentation, and I’ve done various kinds of tech marketing. I’ve designed simple web apps and done other coding-adjacent tasks like context-sensitive help, and I’ve dabbled a little in HTML and CSS — but that was the sum total of my coding experience. Until now. For family reasons, I’ve become curious about how I could use AI for personal projects. I started small: Years ago I scanned hundreds of personal letters and papers, using a consistent filenaming scheme: [initials of sender] [day] [month] [year] [page number] Now as I finally begin using these files, I realize this scheme does not help get things into chronological order – I should have used: [initials] [year]-[month]-[day] [page number] Manually renaming hundreds of files was a non-starter. Bulk renaming should be easy to do with basic programming skills, but I don’t have those skills. Brendan does, but he has much bigger projects on his mind and we were both interested to see what could be done with Codex.* I set one of the directories of scans as the home for a new project (yes, I did a backup first) and said to Codex: “these files are named according to the following scheme: MB Month Day Year page# can you come up with a plan to convert the filenames to MB Year Month Day page# – show me the plan without doing the conversion”. Out of an abundance of caution, I asked Codex to do this in steps. After I was confident it was working correctly, I set it loose on that directory, and then more. All the files got renamed and are now automatically listed in chron order. It was a silly little job but it got done quickly, easily, and correctly. I felt… powerful. I’m good at using software, but I’d never done anything like this. The real power of a computer was finally available to me. I wanted to do more. I decided to try using Codex to solve a daily problem in our household. Anyone who has cooking responsibilities for a family knows that “What’s for dinner?” can be a vexed question Every Single Day. We all have to deal with dietary preferences, food allergies, what food is on hand (and which urgently needs eating before it goes bad), the cook’s energy level that day vs desire to go out vs expense of ordering in. For our household, this logic is in my head, supported by a list of fresh and fast-expiring items scribbled on a whiteboard on the fridge. What if I could put all that into an app that could store every factor that goes into these daily decisions and make useful suggestions for what to cook? One that I can tune to the very specific needs and preferences of my household? I have tried a few times with apps from the iPhone store, but none could handle our particular cooking and eating habits. For one reason, US recipes tend to assume that an “easy” recipe starts with a packaged main ingredient like a cake mix or frozen dinner rolls. I rarely used those even when I lived in the US, and in Australia they are often not available (cake mix yes, dinner rolls no). What this app is NOT My app is purely for personal use. I have no desire to sell or distribute it in any way, not even as open source. I am simply experimenting with a new tool to build something for myself in a way previously impossible for me. Publishing this blog post is akin to open sourcing – I’m giving you a “recipe” for creating an app. I’ll tell you the ingredients and the process, you can bake your own “cake” – and then you get to eat it, too. Last Friday afternoon (Feb 27), I sketched the bare essentials of an app: Build and maintain a list of all food we have, noting expiration dates for short-lived products Build and maintain a list of what’s in the garden Build and maintain a list of dietary restrictions and preferences in the family/guests Build and maintain a list of favourite restaurants to eat at or order in Build and maintain a list of cooking appliances available Check the local weather – when the outdoor temp is >25C I do not want to run the big oven or stand over a hot stove very long.
Reported result
The file renaming task was completed quickly and correctly, allowing automatic chronological listing. The dinner app reached a usable Minimum Viable Product (MVP) after approximately 20 hours of active work and two major reworks. The author reported feeling empowered by the ability to create personal software, though the app remains in a state of active tinkering with known performance issues and feature gaps.
Human role
The original post does not say how a person stayed involved.
Limitations
The workflow requires a ChatGPT Pro or higher subscription to access Codex. The author noted that Codex can be overly accommodating, leading to feature creep and unstable code if scope is not strictly managed. Manual intervention was needed to restart the project when bugs became frequent. The app is for personal use only, with no distribution or commercial intent. Integration with external recipe APIs (Edamam) incurs a $9/month cost.
Try it
Act as a senior engineer guiding a non-coder. I want to build a local personal app using Codex. First, help me define a strict MVP scope to avoid feature creep. I will provide a list of core features: [replaceable inputs: e.g., pantry tracking, recipe suggestions, weather constraints]. Use ChatGPT to draft a detailed technical prompt for Codex specifying Python, FastAPI, and SQLite. Instruct Codex to implement in small milestones, requiring my approval after each step. Before any file changes, ensure a backup is created. If bugs become frequent, advise me to restart the project with a refined architecture. Do not add features beyond the agreed MVP without explicit confirmation. Output the final prompt for Codex and a step-by-step execution plan for me to follow.
Source reports are not independently verified. Publication date is the catalog publication date, not the original article date.