An Introduction to Hackpacking

Mon, 19 Oct 2026 00:00:00 +0000

“Hackpacking” – a portmanteau of “hacking” and “backpacking” – means working on side projects on the go, perhaps assisted by an AI coding assistant. As a lifelong lover of programming side projects, for me AI’s gift is that it lets me work on my side projects while going for walks or hikes or camping trips, often without ever needing to look at a screen. In this post I’ll cover what hackpacking means to me, and my “hackpacking toolchain” – the setup I use to hackpack most effectively.

Despite my love of using AI for side projects, I assure you that every word of this piece was written by my own human hands (or dictated, but still human authored). No slop here.

Hackpacking’s twin dreams

I have two dreams with hackpacking.

(1) No screens at work. I want to be able to do a full day’s work without having to look at a screen. I write software for a living, and if I could do this without needing to look at a screen, it would unchain me from my desk. For this goal, eight consecutive hours would be enough; yes, maybe I have to look at a screen the next day, but at least one full day’s work, screen-free, encompassing all the complexity of my real job. That’s the dream. Ideally I could get away from the desk, take a walk, be outside, be in nature, and still be as productive as I was before.

(2) Getting ideas out of my head. Let’s say I’m hiking and an idea for a side project comes to me. Without looking at a screen, I want to be able to tell my AI assistant the idea. By the time I’m back at a computer, perhaps the idea is already implemented and I can try it out. On such a hike, I may want to stay present with nature. It may be only opportunistically that I want to delegate things to my AI assistant.

Dream one is about doing work deliberately, but being able to be active outdoors while I do it (i.e. a dedicated continuous hackpacking session). Dream two is about being deliberately out in nature, and opportunistically being able to push on a project should the impulse strike. Why do I want these things? Well, it feels really good to be out in nature. The cool air on your skin, it’s relaxing. Being at a desk all day is an unfortunate trade-off that we have to make because computers work best when you’re seated and you can use a keyboard and you can look at a screen. That’s where you’re able to be productive and run a business or build software. But this is changing; now that speech to text works, and now that AI works, the desk and screen are not necessary. We have an opportunity to rethink the workplace.

Why do I want to work on side projects while I’m hiking or backpacking? I love working on side projects. I find them fulfilling. I love doing them with other people. And I love being outdoors. Hackpacking brings these things together. Being outdoors – just going for a walk, allowing your mind to wander – can also be inspiring; that’s where the best ideas come from. So being able to capture those best ideas and get an AI assistant working on them for you is rewarding.

This second dream is not without risks though! If you interrupt your hike to talk to an AI, it takes you out of the moment. If you’ve stopped to capture an idea, it interrupts your mind-wandering. You must be wary of this going in. Hackpacking is not for every hike. It is not for everybody. But it is something I really enjoy and I think others will too. Those trade-offs are not lost on me though, and I certainly won’t advocate for turning every hike into a hackpacking one.

Hackpacking Essentials

You know what I want out of hackpacking. Now, how do we actually get away from the desk? How do we take our hacking and make it genuinely screen free?

For me, the core of the answer is: (a) an iOS Shortcut and (b) a remote Claude Code instance.

A persistent Claude Code on a cloud VM serves as my AI programming assistant, and a simple iOS shortcut lets me send my Claude audio just by pushing my phone’s Action Button.

My setup has other bells and whistles – we do comms in Slack, it has a credit card, an OpenAI API key for image-gen, and so much more – but those two alone define the core of the experience. Later in the post I go through the parts of my stack that I think will be most useful to others (as well as some of the fun quirky bits that might be unique to me), but first I’ll quickly cover these two essentials.

What it looks like:

Push Action Button
Tell Claude what to do
Enjoy the sun on your skin

Setting it up

First, get yourself a VM. I use Hetzner. It costs me a few dollars per month.

On the VM, install Claude Code. Launch it with claude. This will be your persistent Claude instance.

Explain its role to it. For example: “Hi Claude, I’d like for you to act as my persistent chief of staff. I will send you audio clips by saving them to Dropbox in the folder “Voice Memos for Claude”. Whenever a voice memo appears, transcribe it and process it. If it’s a task for you to do, please do it with expedience. Looking forward to working together!”

Claude will have no trouble writing a monitor so that it wakes when new audio files appear.

Then, so you can actually send those audio clips, make an iOS shortcut that:

Starts recording audio
Saves the audio to Dropbox

Set your “Action Button” (iPhone 15 Pro and newer) to trigger your new shortcut. For older iPhones, set Double Tap or Triple Tap (in Accessibility Settings) to trigger the shortcut.

Here I’m using Dropbox as a stand-in for any mechanism for sending Claude the audio. Your two main options are emailing the audio to Claude or saving to some form of cloud storage.

In practice, having your Shortcut using the Mail app to email Claude the audio tends to work best if you’re interested in offline usage – and offline usage is a big part of the hackpacking experience!

(You can click to read the setup instructions, or you can just give your favorite AI the URL of this post and have it walk you through the setup, or even do it for you.)

These two elements alone make building with AI on the go an absolute joy for me. When I send an audio note, there’s no reply – but I know it’s being acted on. If I have a follow-up thought, I just push the Action Button again – there’s no friction to doing so. The persistent Claude Code handles the messages gracefully, regardless of whether I’m working on one project or twenty.

You’re going to want to take your hackpacking setup further than this though, and I have some recommendations for you.

Hackpacking Recommendations

Here are a handful of tips for your hackpacking toolchain.

(1) Set up Tailscale.

This makes it trivial for your Claude to serve you apps that only you can see. How to set it up? Just ask your Claude – it’ll walk you through it and it only takes a few minutes. Put it on your Claude’s VM, your phone, and your computer.

(2) Use Slack for Comms (Or whatever your favorite messaging app is.)

Make a private Slack just for yourself. Create a bot and tell your Claude the key. That’s all you have to do – it will know how to use the key to communicate with you in your Slack (or Whatsapp or Signal etc).

(3) Build little apps for yourself!

Tip: If Claude builds you a web app and you save it to your iOS home screen, it feels like a native app (and can even support notifications and work offline, if you ask for it).

A few of my favorites:

A “Subway Lessons” app that works offline and helps me towards my learning goals while I’m on the subway
“Trail Talk”: for capturing audio screen free during dedicated hackpacking sessions (more on this in a future post, perhaps)
A “Strudel DJ” app for making music on the go
My “Dreams Tracker” – for tracking my dreams, big and small

(4) Prioritize offline support

When you have Claude build you an app, ask it to make it work offline. It’ll just work! Make your iOS shortcut work offline too; see Making your setup robust below.

(5) Give your Claude an OpenAI API Key

Tell it to use the key for image generation whenever you’re building something that needs assets. All your projects will now be 10x better looking by default.

(6) Use subagents to prioritize “availability”

When I send my Claude a message, I want it to be ready to receive it right away. I don’t want it to be blocked from seeing my message because its working on some other task.

So, I tell my Claude to delegate to subagents liberally, such that it’s always available to receive my messages. I have a “weekly review” (something for a later post) that tracks total “uptime” and how long it takes Claude to receive my messages on average.

Quirky Recommendations

(1) Give your Claude a credit card

Use privacy.com to create a capped-spend credit card so that even if Claude goes rogue or gets prompt injected, the damage is contained.

(2) Ask your Claude to buy you a book every few weeks

Tell it to surprise you, and not to spoil that surprise in chat.

(3) Give your Claude its own email address

But only if it sparks joy. I find this helpful because I auto-forward certain receipts to Claude for processing (but don’t want to give it blanket access to my email), and because this way it can receive the book order confirmations without me accidentally seeing them and spoiling the surprise!

(4) Connect your Go Note Gos to your Claude

Go Note Go is a screenless keyboard I designed for capturing my thoughts as I drift off to sleep (among other situations). I have a bunch of these around my apartment. You probably don’t. But if you do, set up a trigger word by which your Claude is directed to read your notes.

(5) Give your Claude access to your photos

I use PhotoSync to automatically back my photos up to a Google Cloud Storage bucket, and I give my Claude access to the bucket. This lets me reference any photos I’ve taken in my voice notes to Claude.

(6) Give your Claude access to your printer

Wake up to the brrr… of Claude printing something for you – a silly comic, or a morning routine. Sticker printers ($33) and receipt printers ($11) are fun too.

Making your setup robust

These next recommendations are a bit more in the weeds, and will help you make your hackpacking setup more robust if you’re setting it up in fall 2026.

(1) Use a watchdog process to restart your Claude if it ever falls over

Just ask your Claude to set this up for you. Have the watchdog message you when this happens so you’re not in the dark when things go wrong.

(2) Turn on swap space on your machine

Especially important if you use a low RAM machine. Just ask your Claude to do this for you.

(3) Make your iOS Shortcut work offline

To make your iOS Shortcut work offline, have it use the “Mail” app to send Claude your voice recordings or save to iCloud Storage. Saving to Dropbox doesn’t work offline. Saving to Google Cloud Storage from an iOS shortcut doesn’t seem to work at all. Be careful that with Mail, sometimes messages get stuck in the Outbox and need to be released manually.

(4) Prefer Action Button over Triple Tap if available

Newer iPhones (15 Pro and newer) offer the Action Button that I so love. Older iPhones support triggering an iOS shortcut with either 2 or 3 taps on the back of the phone. The Action Button is more robust than Triple Tap.

(5) Use Termius for Troubleshooting

If ever your setup does fall over, Termius lets you ssh into your VM from your phone so you can repair your setup without needing to wait until you’re at a computer again. I haven’t had to do this in months, but it was common while I was still getting set up.

(6) Enable “remote control”

Then you can see your Claude Code session in the Claude app. Sometimes – especially early on in getting set up – Claude will require you to give explicit approvals, and it won’t take your audio or Slack word as a sufficient approval. Give these approvals in the app.

Going Hackpacking

I love my hackpacking setup. I’ll walk around Central Park dictating ideas to Claude as they come to me. On the subway – even without internet access – I’ll share ideas, knowing Claude will get to work as soon as my internet connection is restored. This dream I’ve had for years – of being able to build software truly screen free – has now arrived.

Because of hackpacking, I now have a custom suite of apps that suit my quirks, and I build new ones on a whim on my commute. Just this past week I thought “how fun it would be to have an Ezra Klein simulator”, where I can pretend I’m a guest on his podcast. By the time I got home the simulator was ready.

So give hackpacking a try. If you’re already hackpacking, let me know your hackpacking toolchain. And if you want to join me on a side-projecting hike, have your Claude get in touch.