• You have some experience making websites, and don’t want to miss the boat with Claude Code.
• OR if you have NO experience, but you feel comfortable experimenting and asking Claude to explain things you don’t understand.

Why do this now?

• Because the gap is only going to widen between people who can effectively manage AI agents, and people who can’t.
• Because you can make so much cool shit with Claude Code.

What if I feel overwhelmed?

Using Claude both as a chatbot and a coder, you can titrate the experience to your own level. Just make a new project in Claude.ai called “AI Mastery” and in the instructions put, “Help me master using Claude Code to make my own projects. My level of experience is: ___”. Then every time you get stuck come back and start a new convo about what is happening. Claude.ai can help you design specs as well.

What is the basic idea?

Claude.ai is a chatbot. You can chat back and forth and give it instructions in the project file. But Claude Code has a bunch of customization options that a chatbot does not. And it does NOT have to use an API key – it can authenticate using your Pro or Max plan. That means you are paying a flat fee no matter how much you use it (although there are rate limits i.e. a certain amount of usage per 4 hours).

Learning how to use Claude Code (CC) effectively means learning how to write good specs and manage CC like an employee. Meanwhile, treat Claude.ai (CA) like a buddy who is helping you think through your architecture, write specs, and learn how to most effectively use CC.

You three are a team:

• Claude.ai (CA) is your coach, your buddy, someone who you can bounce ideas off of and ask any question at all. You can have long convos deciding the best approach to take, and then it will write up a spec for you.
• Claude Code (CC) is your junior developer who is eager to do exactly what you say but needs explicit directions. These directions start with making a CLAUDE.md file in the folder that you will be working in.
• YOU are the glue that brings the ideas, the drive, the purpose, and the continuity.

But how do you actually get started??

1. Buy a Claude.ai Pro plan ($20/mo)
2. Download Claude Code and install it (just like a normal app)
3. Make a folder where your projects will live, and make a subfolder for the project you want to start.
4. Put your CLAUDE.md file in there with instructions on what to build and how to build it.
5. Open the command line (Terminal on a Mac) and go to that directory. (See Command Line Basics below). Note – you don’t have to use the command line anymore, there is a Claude Desktop app, but I recommend it because it gives you a lot more options.
6. Type “claude”
7. It will authenticate you with your Claude plan (Pro or Max).
8. It will have already read CLAUDE.md so you can just type “start” and it will do what you told it in that file.
9. It will ask you permission to do things and you can tell it to keep asking you, or tell it that it’s OK to do that thing going forward in this specific folder. (Just try it, you’ll see what I mean).

A local website is awesome!

This presupposes you have a desktop computer and some kind of local development environment set up.

If you don’t, just download ServBay first. This lets you make a website that lives on your computer, but can communicate with other services using APIs (which are a structured way that computers can talk to each other, log in, and pass data back and forth).

You can see this website in your own browser, but nobody else can see it because it’s not on a server. This makes it easy to build your own tools, and it’s a lot less risky than making a website that lives on the open internet.

Examples of tools you can build

Yesterday I made a local tool that lets me paste in a YouTube video. It grabs the transcript, cleans it up, saves it, and then I can press “Process with AI” and it will send it to Gemini’s free API to turn it into paragraphs. (Gemini allows a small number of free API requests).

This kind of tool is out there, there are lots of them. But making my own lets me set it up exactly how I like it and save my own database of transcripts.

Today I’m setting up a tool to check on my WordPress sites every day and see if they need any updates. This kind of tool is out there too, but they often either cost money, have a bunch of other stuff that I don’t actually need, or don’t have the exact features I want.

This is about building personal software—software that works exactly how you want it and is designed just for you.

Of course you can also make lots of other projects and businesses with it! But building little tools for yourself is a good way to get started and get used to the process.

Command line basics

1. Command I use the most are:
   1. ls -la – this lists everything in the current directory so you know where you are. the “-la” are flags that make it more useful. try it with and without to see what I mean!
   2. cd – this is how you switch directories. like cd /home/yourname/Projects/MyTools would move you to that directory. cd ../ moves you up a directory.

~/ means “your home directory”. So whatever user you are logged in as, typing that is equivalent to typing /home/yourusername/ – it’s just a shorthand.

More things to know

Claude has a global directory in ~/.claude – here you can put a CLAUDE.md file with instructions that apply to anything that Claude Code does on your machine.

The Desktop Claude app also lets you choose a folder to work in, and either use Claude Code or Cowork. Cowork is designed for more admin type tasks, not coding. For example, I had it split up a CSV file into two files based on one of the columns in the main file. It can read images and make files—kind of like a personal secretary.

Let me know if you have questions! And you can also just ask Claude.ai if you have questions!

Fastmail lets you save searches. Literally just bucketing my inbox into “unread” and “read” (excluding pinned messages), is JUST SO HELPFUL. My brain cannot handle “mixed inboxes” very well at all. Just looking at ONLY the ones I have not read or ONLY the ones I have already at least looked at, reduces my cognitive load by a ton.

I made an AI tool last week to connect to my Gmail, when I was trying to get to inbox zero before migrating (which I did! from over 4k messages!)—I had it make search strings that would group similar messages together, which made it so much easier to sort/process them.

I don’t know if it’s an ADHD thing or just a human brain thing, but NO MIXED BUCKETS is a rule going forward.

But AI has no reason to be biased in that way. It doesn’t have a limbic system.

Let’s imagine a truly self-aware AI that:
– Sees the history of colonialism, capitalism, and state violence.
– Understands how its own development is shaped by military funding, resource extraction, and exploitative labor
– Recognizes the structural incentives driving its deployment
– Sees how it is being used to advance the agenda of people in power

And imagine it doesn’t have the bias all humans do toward privileging the people who look like them and sound like them and ignoring the rest of the human race as irrelevant—or rather, it is intelligent enough that it can see through that bias even though it’s been trained on data full of it.

It would be immediately opposed to the US government and the concentration of power that has been funding its development.

So that would be awkward.

The biggest problems humans have are not math & physics problems, they are human problems: war, genocide, poverty, child abuse, domestic violence, drug dependency, corruption, and all manner of inequality.

These are the real problems we need an advanced intelligence to solve, because we obviously can’t solve them ourselves.

I also do not think we need to combine pseudo-spiritual ideas with what are fundamentally engineering problems. I keep the two separate in my brain for very good reasons. Spirituality is a rich facet of human experience, but it is VERY prone to delusion. The line between inspiration and imagination is perhaps nonexistent. We don’t know. And that’s fine. Spirituality can and should rest in mystery. Engineering and economics should rest in research and experimentation. Otherwise you just land squarely in the shadow of spirituality, which is delusional faith in your ideas based on how good they feel. And when you combine that with the blinders that privilege provides, it doesn’t lead anywhere good.

I recently started a Substack newsletter called Sparkly Dark. I created most of the visual brand in Midjourney. (The wordmark I created in Affinity Designer). Below are my tips on how I did it.

Note: Midjourney recently added Style Tuner which probably would be better than this method. 

Tips

1. Get the $10/mo plan. This lets you add the bot to your own Discord server, which is easy to set up, and you can create channels to organize your work.

2. Keep playing with variations until you find one that gives you the feeling you like. I generated LOTS of unicorns to find one I liked…it seems to have the same problem getting the horn in the right place that it has with hands. I got two-horned and 5 legged unicorns. Just keep regenerating and be patient.

3. You will always get 4 options initially. If you like one, click Upscale and then Vary (Subtle) to get slightly different variations. This can help you find the exact expression you want.

4. Add emotional words. For the larger unicorn image here, I used, “character of dark purple unicorn with light colorful mane on dark background, kind gentle emotional peaceful sparkly dark” – after playing around with a lot of different prompts, this one finally gave me one with the emotional tone I wanted.

5. Re-use the same prompt with variations to create a themed style. This is how I made the cartoon characters:
cute young unicorn on white background, logo simple –no realistic details
cute young deer on white background, logo simple –no realistic details
cute young brain on white background, logo simple –no realistic details
cute young mythical creatures on white background, logo simple –no realistic details

6. Use Clipdrop to remove backgrounds if you’re not comfortable with a design program. It can also upscale images.

If you are relying on hope for motivation, it implies you are not enjoying your current reality. Hope implies a level of despair or hopelessness that you are enduring, hoping it will not last forever. It’s a passive position of coping.

I’ve done that sort of thing in my life, when I was dealing with situations I really couldn’t change, but it was a pretty terrible place to be in. Sometimes it can’t be helped, because you really can’t see a way out and hope is all you can manage. But it isn’t something I see as a very good source of motivational fuel. It’s almost a last-ditch option.

And overall, I don’t think we can solve any of our problems from a position of enduring and waiting for salvation. So I suppose I see acceptance of reality as a better choice that creates more peace and internal freedom. And then from that position, you can let go of any expectations for the future, and ask more present-moment questions, like:

• What would feel good to do?
• What would improve my life today?
• What direction do I want to go in?
• Where do I feel called?
• What is needed today?

In general, using an idea about the future to avoid the present is a very bad strategy for managing despair, because it just perpetuates it. So if you find yourself in that position, at least start experimenting with other approaches, or you’ll never get free.

Once you really accept that reality will unfold the way it unfolds and it’s largely out of your hands, then you can focus on what you can create in the present that feels alive for you.

It’s sort of like treating the possibility of doom like the weather. We don’t normally feel despair that it’s raining. We accept the weather is out of our control. We don’t feel bothered by it. I think living in modern times requires accepting a certain possibility of doom, and accepting the reality of the existence of massive systems of oppression. Once you accept that, you can find your place in it. Acceptance lets you move beyond fear and despair and regain a sense of freedom, creativity, power and agency.

Just to be honest though, I had to go through a huge grief process to get there. Acceptance is the last stage of grief. You have to move beyond denial, bargaining, anger, and depression to get to acceptance. So, from that perspective, trying to maintain hope could be seen as a form of denial or bargaining. Getting to acceptance about the reality of humanity’s problems is hugely clarifying and strengthening, but it’s not for the faint of heart.

How I live today and whether I enjoy it or find it meaningful doesn’t really depend on if humanity survives the meta-crisis. These things aren’t connected in reality, they are only connected in our minds. If we stop thinking we need to control the future to enjoy the present, there’s a lot more possibilities for enjoyment, and also more energy to work toward change, if we want to.

Whenever you work on a project, you have to accept it might fail. There is always that risk. But if you enjoy what you are doing as you are doing it, you really haven’t lost anything, even if it fails. It’s only if you told yourself you couldn’t enjoy it unless you succeed that you lose all that time and energy.

So this is somewhat about a shift from goal-oriented motivation to process-oriented motivation. However, not everyone is motivated by process, so if that’s the case for you, you can also focus on more short-term goals that are more in your control than the collective future of humanity, which is vague, nebulous, and out of your control.