#lina notes

Advice for people making DID/OSDD management apps:

Do not use online website builders or online app builders, they will hold you back.

Use an actual development environment installed on your actual computer. If you don't want to do this, something like GitHub codespaces could also work.

AI coding assistants can be helpful, but you must have enough programming knowledge to know whether to accept what the AI wants to do to use them effectively. AI is often confidently wrong.

Store your code on GitHub and use proper git version control. Private repos are available if you would rather keep your code private.

Use one of the big cloud providers for hosting, but be sure that you understand how you are billed and choose resources accordingly. You can end up with a huge bill you can't pay if you set things up wrong.

Most cloud providers have a certain amount of resources that you get for free. It is possible to design things so that you stay within these free limits

Protect your REST APIs with API keys that you don't share, this will help prevent you from getting a big bill.

NoSQL / document type databases are often cheaper than things like MySql or SQL server

consider just storing data locally, no server costs that way!

Javascript or Typescript based app frameworks are probably the easiest for a beginner to get started with.

If building for Android back up your release signing key and store it securely. You will not be able to update your app without it.

If building a web site or a rest API for an app, learn about the OWASP Top Ten security vulnerabilities and make sure that you do not have them in your app.

Allow people to export their data to JSON, data sovereignty is important and is a legal requirement if your app can be downloaded or accessed in EU countries.

Do not check secrets like database passwords or keys into git source control.

If your app grows beyond a few users, at some point you will need to talk to actual lawyer about the legalities of your operations. Terms of service and things like that usually need the help of a lawyer.

The developer develops the app and releases it

The community is excited and embraces it

the community starts demanding more and more features

The developer can't say no because most CDD app developers have a CDD themselves and traumatized people are bad at saying no

More features are implemented and the app grows increasing the amount of work the developer has to do until it completely consumes their life

The community launches harassment campaigns targeting the developer over some perceived failing

The developer issues a statement defending themselves

The community gets angrier.

The developer gets sick of it and shuts down the app

Someone else makes a new app to take the place of the app that shut down starting the cycle again.

I know i should get a lawyer to look at it, but getting a lawyer to draft a TOS costs approximately $2,000 so fuck that.

My app does not collect any personally identifiable information so I should be fairly safe.

VS Code now tries to complete the code you are typing using AI and I can't figure out how to turn it off. You can press tab to accept it's suggestions or just ignore it and keep typing, it is sometimes confusing what is the actual contents of the file and what is an AI suggestion as this is indicated by making the AI suggested text dimmed.

I do accept suggestions occasionally for things like config files, but most of the time it doesn't seem to understand what I am trying to do.

In javascript and typescript it actually works reasonably well.

I think it would choke on C++ or rust.

I am trying to decide if I feel that using this is OK or not.

I am for most part anti AI and my position generally is that AI is dangerous and should be completely banned world wide, except for certain permitted safe uses (eg enemies in video games). There is a book that explains the reasons for this better than I can called If anyone builds it, everyone dies by Eliezer Yudkowsky.

Before we even get into the issues posed by superinteligence, present day AI systems that actually exist learn things like racisim and discrimination from humans and amplify them.

For a practical example, supermarkets in new zealand have started using facial recognition and computer vision systems that try to detect shoplifters. This system had an issue where a number of people who had done nothing wrong were banned from their local supermarket because the computer said they were shop lifting.

The thing these people had in common was that they were brown.

There is no way to prevent an AI system from learning patterns that are present in the training data. You cannot say "learn this, but don't learn that.". These systems are trained on white men and malfunction when presented with faces of other demographics.