Week 1 of 6
Agentic Creativity
Intro to agents · installing Hermes
Live on Zoom · 2.00–4.30pm ET
16 August 2026
Derrick Schultz
By 4.30 today you will have three things.
01
An agent on your laptop
Hermes installed, connected to a model, and answering you.
02
An agent on a server
The same agent on a machine that stays awake after you close the lid.
03
A map of the next five weeks
What an agent is, and the three very different ways artists use one.
Today’s schedule
2.00
What this class is, and what it isn’t
2.15
What an agent actually is
2.45
Three ways artists use agents
3.10
Break
3.20
Install Hermes on your laptop
3.50
Put an agent on a server
4.15
Working time and homework
4.25
Questions
The six weeks
Week 1 · Aug 16
Intro to agentic creativity; installing Hermes Agent
Week 2 · Aug 23
Free your mind — the non-artistic uses of agents
Week 3 · Aug 30
Skills and MCPs
Sep 6
No class
Week 4 · Sep 13
Long-running and evolving agents
Week 5 · Sep 20
An intro to agentic coding
Week 6 · Sep 27
Show and tell
Recordings and notes go out after every session.
What you need to have with you
A laptop
macOS 12+, Windows 10/11, or any Linux distro. You will be typing in a terminal, but you will not be writing code.
One model subscription
Nous Portal, OpenAI, OpenRouter, or an equivalent. Any one of them is enough — we will pick together.
A card on file
Agents spend money as they work. Expect a few dollars a week while you are learning, more if you leave one running.
Optional, from week 4
A cheap server. We will set one up today so you can see the difference.
No prior experience with coding or agents required. Skeptics encouraged.
01
What is an agent?
2.15–2.45 · 30 minutes
a working definition
An agent is a model that can use tools on your behalf.
That is the whole idea. Everything else this class covers is a consequence of it.
A chatbot answers. An agent acts.
A chatbot
Returns text. You are the one who copies it somewhere and does the work.
An agent
Reads and writes files, runs commands, searches the web, calls services, makes images.
A chatbot
Forgets when the window closes.
An agent
Keeps notes, remembers your projects, and can pick a job back up next week.
A chatbot
Waits for you.
An agent
Can run on a schedule, unattended, while you are asleep or making something else.
inside an agent
What an agent is made from
AGENT
Your input
What you asked for, plus the files you handed over
Memory
Notes it kept from earlier work
03 · HARNESS
Model
Reads, reasons and chooses the next action
Runs the loop and holds the working context
Tools
The actions it is allowed to take
Skills
Written instructions for doing a specific job
Only one of these six boxes is the model. Week 3 is about the last two.
how it works
The whole thing is one loop.
Ask. It picks a tool and runs it. It looks at what came back. It picks the next thing.
Nothing more sophisticated is happening. The interesting part is that it can see its own results and change course.
what it can reach
Tools are how the agent touches anything outside the chat.
Some come built in: files, terminal, web search, a browser.
The rest you connect yourself — your email, your notes, an image service, your own site. That connection layer is MCP, and it is where week 3 lives.
Four things change the moment it can act.
It can be wrong loudly
A bad sentence is harmless. A bad command deletes a folder. Sandboxing and permissions stop mattering in theory and start mattering to you.
It costs money per attempt
Every step in the loop is a paid call. A long unattended run is a bill you find out about later.
It has a body
Where it runs decides what it can reach. This is why we put one on a server today.
Your words are the interface
It does what you said, not what you meant. This is the actual skill the class teaches.
02
Agents and creative work
2.45–3.10 · 25 minutes
Three rungs, and they are genuinely different activities.
rung one
It clears the desk
Sorting mail, chasing invoices, renaming exports, gathering references, writing the boring caption.
Mistakes here are cheap, obvious, and easy to undo — which is exactly why you start here.
Week 2
rung two
It works next to you
You are still making the decisions, but it is generating material you react to, finding references you would have missed, and running the tedious middle of your process.
This is where most of the useful work happens.
Weeks 3–4
rung three
It works without you
You give it a direction and it decides what to make, when to stop, and what to show you. You come back to a folder.
Interesting, unreliable, and the only rung where the authorship question gets sharp.
Weeks 4–5
rung one · in practice
Mine reads everything I would rather not read.
A scheduled job goes through Slack, my task list and my meeting notes, then posts one message: what needs me today, what is only for information, and a suggested next step on each.
It is not making art. It is buying back the afternoon.
rung two · in practice
On films I use one as an ongoing image source.
I give it a direction, and it makes stills and short clips while I work elsewhere. When I get stuck I go browse what it made.
Some of the stranger results change where the film goes. That is the actual value — not speed, but material I would not have thought to ask for.
Rung three is where the hard questions live.
Who made it
You wrote the brief and chose the model. It chose the images. We will keep arguing about this for six weeks and not settle it.
When does it stop
An open loop will happily polish the interesting version into a competent one. Somebody has to define finished.
What is it optimising for
Whatever you wrote down. Vague instructions produce the average of everything it has seen.
Is any of it yours
The honest answer, most of the time, is: the direction is, and the taste to throw away 90% of it is.
Where I still do not trust it
Taste
It praises plenty of mediocre work, including its own.
Knowing when to stop
Left alone it converges on the safe version.
Spending
Long loops add up, because every pass through the loop costs money.
Intent
It follows the words you gave it, which are never quite the thing you meant.
Confidence
It reports success in the same tone whether or not it succeeded.
the part that does not get automated
You are still the only one in the room with taste.
03
Install Hermes on your laptop
3.20–3.50 · 30 minutes
Why we are using Hermes Agent
It is open source
MIT licensed, from Nous Research. You can read every part of it, including the parts that decide what it is allowed to do.
No model lock-in
Nous Portal, OpenAI, OpenRouter or your own endpoint. You can switch mid-conversation.
It lives on chat apps
Telegram, Discord, Slack, WhatsApp, Signal, email — one agent, one memory, reachable from your phone.
It remembers and schedules
Persistent memory, self-written skills, and natural-language scheduling. That is weeks 3 and 4.
It sandboxes
Local, Docker, SSH, Singularity or Modal. It matters more than it sounds like it does.
Everything you learn here transfers — the ideas are the same in Claude Code, Codex or Cursor.
step 1 · the easy way
Download the app.
hermes-agent.nousresearch.com
macOS 12 or newer takes the .dmg. Windows 10 and 11 take the .exe. The installer brings its own Python and Node, so there is nothing to set up first.
Start the download now while I keep talking — it takes a few minutes.
step 1 · the other way
Or install it from the terminal.
macOS · Linux · WSL2
Windows PowerShell
The installer pulls Python 3.11, Node 22, ripgrep and ffmpeg into ~/.hermes. It does not touch the versions you already have.
step 2
Give it a model.
An agent with no model is a very expensive text editor. Run the wizard and pick one.
If you bought a Nous Portal plan, --portal does the login, sets the provider and turns on the tool gateway in one pass. Otherwise choose OpenAI or OpenRouter and paste your key.
Inside a conversation, /model openai:gpt-5.2 switches without restarting.
step 3 · verify
Make it prove it has hands.
Do not ask it a question — a chatbot could answer that. Ask it to do something small and local.
If hermes doctor is clean and this comes back, you are done.
What’s in my Downloads folder? Just count the files.
terminal · called
There are 412 files in ~/Downloads. 38 of them are screenshots from this month, if you want me to do something about that.
The commands you will actually use
And inside a conversation: /stop to interrupt it, /status to see what it thinks it is doing, /model to swap brains.
When it breaks in the next ten minutes
“command not found: hermes”
Your shell has not picked up the new path. Open a new terminal window, or run source ~/.zshrc.
“It says no model configured”
Run hermes setup again. The wizard is safe to re-run and will not wipe anything.
“My key is rejected”
Almost always a trailing space, or a key from the wrong provider. Paste it again carefully.
“It answers but cannot do anything”
hermes tools — the capability you want is probably switched off.
“Something about Python”
hermes doctor first. If you have a PYTHONPATH set globally, unset it and reinstall.
Anything else
Paste the error into the chat. Unmute and read it out. Somebody else has it too.
checkpoint
Nobody moves on until everybody has an agent that answers.
04
Put one on a server
3.50–4.15 · 25 minutes
Same agent, two very different animals.
On your laptopwhat we just did
On a serverwhat we do next
Your files
Reaches your real folders. That is the whole point.
Only what you deliberately copy over.
Staying awake
Dies when you close the lid or lose wifi.
Awake at 3am whether or not you are.
Scheduled work
Does not happen.
Actually runs.
Reaching it
You have to be sitting at the machine.
Text it from your phone over Telegram.
When it goes wrong
A bad command hits your real drive.
Blast radius is a box you rebuild in ten minutes.
What it costs
Nothing. You already own the machine.
A few dollars a month, billed hourly.
Most people end up running both. Today you get one of each.
step 1 · vultr.com
Spin up the box.
Deploy a new Cloud Compute instance. Any provider works — we are using Vultr because the console is plain and it bills by the hour, so a box you destroy on Sunday night costs you almost nothing.
Add your SSH key during setup. Do not use a root password.
Deploy new instance
- Type
- Cloud Compute · Shared CPU
- Plan
- 2 vCPU · 4 GB RAM · 80 GB
- Image
- Ubuntu 24.04 LTS
- Location
- Whichever is closest to you
- Auth
- SSH key · no password
- Extras
- Skip backups and monitoring
About $20/month, billed hourly. 2 GB works if you skip browser automation; 1 GB does not.
steps 2 and 3
Log in and install it.
Skipping the browser? Append -- --skip-browser to the install command and you can drop to a 2 GB box.
step 4
Make it survive you logging out.
This is the one step everybody skips. Skip it and your always-on agent stops the moment you close the SSH window.
step 5 · the payoff
Now it is in your pocket.
The gateway puts the same agent — same memory, same tools — on Telegram, Discord, Slack, WhatsApp, Signal or email.
You stop SSHing into anything. You just text it.
while I'm asleep, make six studies from the dusk brief and put them somewhere I can look at them over coffee
scheduled · 02.00 nightly
Six files in ~/studies/dusk/. Two are worth your time, the other four went muddy. Contact sheet attached.
It is a public machine now. Act like it.
Keys, not passwords
SSH keys only, and disable root login once your own user works.
Do not put your life on it
The point of the server is a small blast radius. Give it the project, not your whole home folder.
Watch the spend
An unattended loop with a scheduling bug is the most expensive thing in this class. Set a budget limit with your model provider today.
Rotate anything you paste
Keys typed into a chat window end up in logs and memory files. Assume they leaked.
Destroy it when idle
Hourly billing means a box you rebuild each week is cheaper than one you forget about.
05
Your turn
4.15–4.30 · 15 minutes
working time
Give it one job you actually do not want to do.
• Something small, boring and real — rename a folder of exports, summarise a long thread, sort your references
• Watch what it does before you approve anything
• Notice the moment it guesses wrong, and what you had left out
Hands up and I will come to you. This is the part where everyone hits the same three problems.
Before next Sunday
Use it every day
Ten minutes, on real work. Frequency teaches you more than one long session will.
Keep a list of failures
Every time it does the wrong thing, write down what you asked and what you meant. Bring the list. That is week 2’s material.
Find your worst chore
The recurring, mildly humiliating task you have been avoiding. We are automating it next week.
If your server is running
Leave it on and check what it cost you by Saturday. Bring the number.
Everything from today
Hermes
hermes-agent.nousresearch.com · curl -fsSL .../install.sh | bash
Install command reference
hermes setup · hermes model · hermes tools · hermes doctor · hermes gateway
The server
vultr.com · Ubuntu 24.04 · sudo loginctl enable-linger
This deck
artificial-images.com/agentic-week1
Stuck between sessions
Email me. Include the exact error, not a description of the error.
Next week
Free your mind
The non-artistic uses of agents
23 August · 2.00pm ET
questions
Ask me anything now, or bring it next Sunday.
Derrick Schultz · derrick@titles.xyz · artificial-images.com