2026 Ai Coding Workstation Setup

We've been using Generative Ai for the last few years for non-coding tasks like [Image Gen](https://github.com/nelsonic/nelsonic.github.io/issues/912),
But have so far resisted using it for Code Gen @dwyl because we still
**_Prefer_ `Humans`** to _think_ about the problem we are solving for other **`Humans`**. 💭
Until recently the output of Code Gen was not production-quality: https://arxiv.org/abs/2502.11844

We're half-way through 2026 and a _lot_ has changed in the Ai-Coding Landscape! 👩‍💻 🤖
We are aware that they have improved sufficiently to be _useful_ without too many errors. 📈
And provided the `Human` Software Engineer is using the Code Gen and still auditing for security/privacy issues,
we feel this is now at the point where we can dive in.

This setup/guide will assume that we've gone full Ai-[Kool-aid](https://en.wikipedia.org/wiki/Drinking_the_Kool-Aid) e.g: [youtu.be/DmU9uovmT2A](https://youtu.be/DmU9uovmT2A)
and are building an ["AI-native startup"](https://claude.com/blog/the-founders-playbook).

# Required

1. Create new file named `ai-setup.md` and start documenting. (Please don't use any `Ai` to do the writing) 🤖 🚫
2. Document the setup from scratch for a "**_Dedicated_ Coding Workstation**" that will use the latest Code Gen tools.
For security reasons, the machine will not contain any (personal) sensitive info and not be used for _any_ sort of personal use; just work. This separation of concerns allows using tools that might otherwise be considered risky. (within reason; not going to `sudo npm install` anything ...)
2. Describe exactly which tools need to be installed on a fresh Mac/Linux machine, e.g. [Claude Code](https://claude.com/product/claude-code), [Cursor](https://cursor.com/), etc.
3. Document how to setup/secure any necessary **API Keys**.
4. Any other tools that are helpful for building Apps using a "**Ai First**" Approach.
5. Once the environment is setup, create some sample prompts to show it working.

# Assumptions

1. Assume there is a **Budget** of **`$200/month`** per Engineer/Machine (covers `Cursor` [Ultra](https://cursor.com/pricing) or `Claude` [Max](https://claude.com/pricing) at current pricing)
2. Assume we will be writing code in **`Elixir`/`Phoenix`** and `Dart`/`Flutter`. Ideally also `Swift` (`iOS`/`macOS`)
3. Attempt to follow security best-practices wherever possible, document these.

Create the required ai-setup.md file and use the issue requirements as the outline. Cover a fresh Mac/Linux workstation for Elixir/Phoenix, Dart/Flutter, and Swift development, including Claude Code, Cursor, API-key security, the $200/month budget, security practices, and sample prompts. Done means the setup can be followed from scratch and the sample workflow is demonstrated.

Written by the indexing model from the issue text.