I write JS, sometimes Rust & Go, I also build stuff at @techulus
Joined May 2009
- Tweets 5,238
- Following 191
- Followers 541
- Likes 2,408
665 Photos and videos
So why am I building a PaaS (platform as a service)?
Short answer: It's because I love @Railway and I want the same experience with my own servers.
Long answer: I self-host several services on my home lab servers. This was a very manual task with minimal automation; I pulled code into servers, set up Docker Compose files, Tailscale, Traefik, DNS, pretty much everything was manual. On the other hand, when I deploy my side projects to services like Railway or @vercel, everything is just a few clicks away and feels magical. I want to bring this seamless experience to my infrastructure.
In case you're wondering, yes, I've tried @coolifyio / @getdokploy / @dokku / @portainerio and some other self-hosted PaaS solutions, but they didn't work well for me. There were some flaws that I couldn't fix (private networking, load balancing workloads across servers, ease of swapping servers, self-healing workloads etc.).
So I started to rethink how I could build the perfect self-hosted PaaS. FYI, it's heavily inspired by platforms like Railway, Fly.io, etc. My research was mostly reading their engineering blogs and chatting with LLMs.
Here's how my solution differs:
> Coolify/Dokploy = centralised control which is SSH-driven, push-based orchestration, not multiple server native.
> What I'm building = a distributed peer-to-peer container fabric: agent-driven, pull-based orchestration, private-first networking, disposable workloads, proxy nodes to handle ingress, and finally, a touch of magic.
2
5
394
Jun 13
The US government, citing national security authorities, has issued an export control directive to suspend all access to Fable 5 and Mythos 5 by any foreign national, whether inside or outside the United States, including foreign national Anthropic employees.
The net effect of this order is that we must abruptly disable Fable 5 and Mythos 5 for all our customers to ensure compliance.
Access to all other Claude models is not affected.
We apologize for this disruption to our customers. We believe this is a misunderstanding and are working to restore access as soon as possible.
Read our full statement: anthropic.com/news/fable-myt…
3
hey @jasonfried it would nice to add an agent as a basecamp user, do you have a recommended flow for settings this up?
My workflow currently is asking an agent to go through a task list and and add comments, but I don't want the comments to show under my name.
3
1
17
12,338
Ferrari Luce was good
Jun 4
Audi has officially unveiled its first hybrid supercar called the Audi Nuvolari.
• Starting price: $697,000
• 0-60mph: 2.5s
• 978 hp
• 499 units will be built
• Fastest production vehicle in Audi history
• Top speed: Over 217 mph (350 km/h)
• New Audi signature paint color: Titanium
• Three electric motors each produce 110 kW
• Two oil-cooled electric motors mounted on the front axle
• 7.3 kWh lithium-ion battery
• Fully electric driving capability in E-Hybrid mode
• Hybrid powertrain combines a 4-liter twin-turbo V8 with three axial-flux electric motors
• Formula 1-derived prepreg autoclave carbon manufacturing process
• Brake-by-wire braking system
• Braking system capable of absorbing up to 2.8 megawatts of energy
Deliveries begin in the first half of 2027. More photos in the thread below:
102
Off to a productive weekend:
- migrated capture database from @Railway to @PlanetScale , those NVMe drives are fast 🏎️
- setup failover instances in GCP
- testing failover load balancing with @Cloudflare
28
How do I get this in Sydney 🥹
May 21
Introducing: UniFi 5G Backup
🔹 Carrier Unlocked
🔹 Plug & Play via any PoE port
🔹 Versatile mounting options
🔹 Granular failover controls
Only $99
ui.social/U5G-Backup
16
Arjun retweeted
May 19
UPDATE: So far we've identified 639 compromised npm package versions across 323 unique packages in tonight’s Mini Shai-Hulud wave.
That includes 558 versions across 279 unique @antv packages. Most were detected within ~6 minutes of publication.
socket.dev/blog/antv-package…
37
168
1,033
1,025,020
I've been thinking a lot lately on HA for the container deployment platform I'm building. There are two big problems:
- central control plane: apps run, but you don't have control over the cluster
- stateful volumes: they're mounts to local disk, if the machine dies, backup recovery is the only option (aka we lost data)
basically, control plane and storage lacking HA
I want to tackle control plane first (because storage might be even harder)
I'll start exploring if I can move the control plane state to a replicated state machine, an RSM as the source of truth for cluster coordination state.
22
Vercel labs is cooking!
Great to see lot of fun projects
May 15
Introducing Zero
The programming language for agents.
I wanted a systems language that was faster, smaller, and easier for agents to use and repair.
Explicit capabilities. JSON diagnostics. Typed safe fixes.
Made for agents on day zero.
37
How does things like this happen? Btw, “This wasn’t caused by AI”
May 15
This was our mistake, and we’re sorry.
What happened: our team accidentally cancelled a large number of subscriptions (rather than a single subscription as was intended). This wasn’t caused by AI. We’re working with Screen Studio to restore impacted subscriptions and we’ve tightened our internal systems and guardrails to help prevent this from happening again.
28