If you bought a Raspberry Pi and it’s been sitting in a drawer since 2020, I get it. Life happens. But the Pi 5 changed everything, and honestly 2026 is the best time to dust one off (or grab a new one). I’ve been running projects on these little boards for years, and the jump from Pi 4 to Pi 5 is the biggest generational leap I’ve seen. We’re talking 2-3x faster CPU, PCIe support for NVMe storage, and enough headroom to run things I never thought possible on a $80 board.
This post is an update to something I wrote back in 2020 when the Pi 4 was king. Some of those projects still hold up, but a lot has changed. So let’s go through my current favorite five things to do with a Raspberry Pi in 2026.
What You Need to Get Started
Before we get into the projects, here’s what I’d actually buy if I were starting fresh today.
A CanaKit Raspberry Pi 5 Starter Kit is the easiest way in. It comes with a power supply, a case, and a microSD card, so you’re not scrambling to figure out what adapter you need. CanaKit has been doing this for years and their kits are solid.
For storage, get a Samsung 256GB microSD. Cheap cards will wreck you with random corruption. Samsung’s endurance line is worth the few extra dollars.
If you want something a bit nicer than a bare board, the Argon ONE V3 case is my favorite Pi 5 enclosure. It routes the ports to the back, has a built-in fan controller, and looks like actual consumer hardware instead of a science fair project.
And if you’re planning to run Nextcloud or anything storage-heavy, grab the NVMe SSD base. The Pi 5’s PCIe slot makes this possible and the speed difference over microSD is not subtle.
Now let’s get into the fun stuff. If you want the bigger picture, check out our full list of Pi 5 projects for beginners in 2026.
1. Retro Gaming Console
This is how I got hooked on Raspberry Pi in the first place, and it’s still one of the best uses for the hardware. There’s something about playing Castlevania: Symphony of the Night on original PlayStation hardware at 60fps that hits different when you know you’re running it on a $80 computer the size of a credit card.
The two main options in 2026 are RetroPie and Batocera. I’ve used both, and my honest take is that Batocera has become the easier recommendation for most people. You flash it to an SD card, boot up, and you’re in a polished front-end within minutes. RetroPie gives you more control and a bigger community around customization, but the initial setup is more fiddly.
The Pi 5 can handle pretty much everything up through PlayStation 1 and N64 without breaking a sweat. Dreamcast emulation (Flycast) runs well for the majority of the library now, which was not the case on Pi 4. PlayStation 2 via PCSX2 works for simpler games but the more demanding titles will chug. I wouldn’t count on PS2 being your daily driver yet.
For systems that run great: NES, SNES, Game Boy Advance, Sega Genesis, Sega Saturn, Arcade (MAME), PS1, N64, Nintendo DS. That’s an enormous library to dig into.
The controller situation has gotten better too. 8BitDo makes fantastic controllers (the SN30 Pro is my go-to) and they connect over Bluetooth without any driver nonsense. I’ve also had good luck with wired Xbox controllers, which Linux picks up automatically.
One thing I always tell people: spend 20 minutes setting up a proper scraper in Batocera so you get box art and metadata for your games. The difference between a raw file list and a proper media-rich library is massive.
2. Network-Wide Ad Blocker with Pi-hole
I set up Pi-hole on my home network about four years ago and I will never go back to a network without it. Every device on your WiFi, including smart TVs, phones, and tablets, stops seeing ads at the DNS level. No browser extensions required, no app to install on individual devices.
The way it works is simple. Pi-hole becomes your local DNS server. When any device on your network asks “where is doubleclick.net?”, Pi-hole checks its blocklist and returns nothing instead of the real address. The ad never loads. You don’t even see a placeholder, it just doesn’t exist.
On the Pi 5, Pi-hole runs fast enough that you’ll never notice any DNS latency. I’ve got around 1.8 million domains on my blocklist and queries still resolve in under 5ms.
The setup is one of the easiest Linux installs I’ve done. You run a single curl command, answer a few prompts, then change one setting in your router to point DNS requests at the Pi’s IP address. That’s really it.
We have a full walkthrough that covers everything from installation through the router configuration step: How to Set Up Pi-hole on a Raspberry Pi 5. If you do one project from this list, make it this one.
The stats dashboard alone is addictive. You can watch in real time how many ad requests are getting blocked across your whole household. Last month mine blocked about 23% of all DNS queries on my network. That’s a lot of junk that wasn’t loading.
3. Home Automation Hub with Home Assistant
Home Assistant is the software I wish I’d found years before I did. I was deep into the Google Home ecosystem, then Alexa, and every time a company killed a feature or changed their privacy policy I’d lose something. Home Assistant runs locally on your Pi, talks to all your smart devices directly, and doesn’t phone home to anyone.
The Pi 5 handles this well. Home Assistant’s HAOS (Home Assistant Operating System) flashes to a microSD card, boots up, and you’re into the onboarding flow in under 10 minutes. It auto-discovers a lot of devices on your network immediately, which still kind of amazes me.
Where it gets really powerful is adding Zigbee or Z-Wave support. I use a ConBee III USB dongle for Zigbee, which lets me add devices from IKEA Tradfri, Aqara, Sonoff, and a bunch of other brands without any of their cloud apps. Everything talks through the Pi directly. The HUSBZB-1 is a popular combo dongle that does both Zigbee and Z-Wave if you have a mixed setup.
Some specific devices I’ve had great results with in this setup:
- Aqara door/window sensors (cheap, reliable, battery lasts forever)
- IKEA Tradfri bulbs and outlets (affordable, well-supported)
- Sonoff ZBMINI for in-wall switching (requires neutral wire, but works great)
- Shelly Plus 1 for automating dumber devices (these are brilliant little modules)
The automation side is where you end up spending most of your time. I have things like “turn on the porch light 30 minutes before sunset,” “turn off everything except the bedroom when we leave home,” and “don’t let the thermostat run if a window has been open for more than 10 minutes.” None of that required a cloud service or a subscription.
Fair warning: the learning curve is real. Home Assistant has a lot of surface area. Give yourself a weekend to get comfortable with the concepts before you expect to be building complex automations.
4. Personal Cloud Storage with Nextcloud
I started using Nextcloud seriously about two years ago after getting annoyed at Google Photos changing its storage policies for the third time. The pitch is simple: it’s like having your own Google Drive that lives in your house, on your hardware, under your control.
The Pi 5 with an NVMe SSD is what actually makes this practical. I tried running Nextcloud on a Pi 4 with a USB hard drive and while it worked, the performance was just bad enough to be frustrating. With an NVMe base and a decent SSD, Nextcloud on Pi 5 feels like a real cloud service. File syncing is fast, the photo gallery loads quickly, and the mobile apps stay in sync without me thinking about it.
I use Nextcloud for a few things: photo backup from my phone, syncing documents between my laptop and desktop, and sharing large files with clients. That last one used to mean Dropbox or WeTransfer. Now I just generate a share link from my own server.
The privacy angle is the reason I stick with it. My photos don’t train anyone’s AI. My documents don’t get scanned for ad targeting. My files are in my closet, on my hardware, encrypted with keys I control.
Setup-wise, I’d recommend using the Nextcloud AIO (All-in-One) Docker image. It handles the database, web server, and all the moving parts for you. You basically need Docker installed, then you run one command and follow the web installer. The documentation has gotten a lot better in recent versions.
One practical note: if you want to access your Nextcloud from outside your home network, you’ll need to set up port forwarding on your router and ideally a dynamic DNS service. Cloudflare Tunnels are another option if you don’t want to open ports. I use DuckDNS with a Cloudflare tunnel and it works reliably.
5. AI Projects and Local Language Models
This is the one that would have seemed absurd when I wrote the original version of this article in 2020. Running actual AI models on a $80 board. In your house. For free.
The Pi 5’s 8GB RAM variant (go for 8GB, seriously) can run small language models locally using Ollama. I’ve been running Qwen3 at the 0.6B and 1.7B sizes for lightweight stuff, and the 4B models are usable for summarization and simple Q&A tasks. You’re not going to run a 70B model, but for local automation tasks and experimentation, the smaller models can do more than you’d think.
What I actually use this for: I have a setup where the Pi monitors certain conditions at home and can generate a natural language summary when I ask it something. It’s way more interesting than it sounds when everything is running on hardware I own.
For image generation, the Pi 5 is honestly too limited for anything practical. I wouldn’t go down that road expecting good results. But for text tasks, agent frameworks, and local API endpoints? It’s more capable than I expected.
Ollama is the easiest way to get started. Install it with one command, pull a model like ollama pull qwen2.5:1.5b, and you have a local API endpoint running. From there you can connect it to things like Home Assistant’s conversation agent, or build small scripts that use local AI for classification or summarization.
I’ve also been tinkering with OpenClaw for building more structured AI agent workflows. The idea is that you can string together small models and tools into agents that do useful things automatically, all running locally. If you’ve been curious about the agent AI space but don’t want to pay OpenAI rates for every experiment, the Pi 5 is a good sandbox for that.
For robotics with an AI angle, I did a project a while back using a Pi as the brain for a tank robot. If you want to see what that looks like in practice, the DFRobot Devastator tank build covers the Python control code side of things.
The honest take on AI projects on Pi: it’s experimental, it requires more Linux comfort than the other projects on this list, and you’ll hit limitations. But if you want to learn how this stuff actually works at a low level without spending money on cloud APIs, there’s no better playground.
Which Project Should You Start With?
If you’ve never set up a Pi before, start with Pi-hole. It’s the fastest path from “unboxed the kit” to “this is actually useful in my daily life.” The setup is forgiving, the payoff is immediate, and you’ll feel the difference every time you load a webpage.
If you want the most fun per hour of effort, go with the retro gaming console. Batocera gets you playing old games in under an hour and there’s really nothing like it.
Home Assistant is the one with the highest ceiling. If you’re into home automation at all, you’ll end up going deep on this and not regretting it.
Nextcloud and the AI projects are more advanced setups that reward patience. Both are worth it, but give yourself a real weekend rather than expecting to bang either out in an evening.
If you’ve had a Pi collecting dust, the Pi 5 is worth the upgrade. And if you’ve never owned one, $80 gets you a computer that can do all five of these things without complaining.