This is an idea which is hard to make money on. Perhaps that's why it's not been made real yet since it's largely based upon Alan Kay's ideas on computing. Though I'm writing it down now because with the advent of AI it will soon become cheap enough to implement it will get done even if it's hard to make money on after.

The essential problem that has prevented computing from really being nice to use is that software tools don't work well together. Each tool does many things because it has to. Any tool that only does part of the job is useless from the perspective of a non programmer since there is no way for them to move the project into another tool to finish it off. The worst thing as a software user is to spend a lot of time learning a tool only to realise that it can't do everything you need and then needing to go learn another tool. Often times the tool that does the whole job is the old thing that awful to use but complete.

The status quo benefits large software firms the need for a complete solution acts like a mote. So does the time users sink into learning their software. Many niche software products charge a fortune for software licenses based on this mote. Even though other products will do 95% of the task for much less. Let's take a look at a couple of examples of what doing things with smaller software will look like.

Let's take ordering food. The first thing you need to do is find restaurants in your area. For this you use a location aware search engine. This produces a list of restaurant objects. You then need to get the menu of each restaurant. This is easy because it's just a property of the restaurant object. You then need to filter out all the stuff that either not available this time of day or that you don't like or that you have had recently. This would be done by menu app that stores your preferences. This app then presents you with a stack of options sorted according to the probability it thinks each item has of being good. You then order the food by opening an order submission app and then either opening your courier app or walking to you car if you feel like getting it yourself. When you hop in you car it checks you recent activity, sees you have a pending order near by and automatically enters the restaurants address into the navigation. Your calorie counting app will notice the order too and add the computed calorie count to your days total. When you open your run tracking app the next day it will adjust the default run length according to your calorie count from the previous day. The menu app will then take the calorie count into consideration when sorting its recommendations. If the food is cold by the time you get home you would just poke it in the microwave until it beeps. The time would be set automatically based on the type of meal and it's starting temperature.

The next example I'd like to take a look at is business accounting and stock management. This space is catered to by a category of software known as enterprise resource planning (ERP). But all large companies roll their own software since evidently none of the available options are good enough. This software problem is made significantly more fragmentary by the differing regulations of each country. The differences are always small but non the less important since failure to comply can put you in jail.

Lets take as an example a simple business that purchases used or damaged cars, restores them and then sells them in Australia. They begin by searching for cars the can be profitably restored. To do this they need to asses the market value of given vehicle in good condition and the cost to restore the given vehicle to that condition. Given the cars VIN it's exact specs can be determined and this can be used to find comparable cars that have sold recently to determine the market value of the car. Calculating the cost of restoration involves identifying the parts that need to be replaced and the labour content required to replace them.