471–480 of 1001 posts
When I started doing this, I also decided to try Proxmox's new OCI compatibility, which seems to be working well so far, so I am removing all my Docker VMs and recreating the containers directly on my hypervisor.
I'm building Tela ( https://github.com/paulmooreparks/tela ), a self-hosted relay that tunnels TCP services through encrypted WireGuard connections. The key difference from Tailscale and similar tools is that it requires no TUN adapter, no root access, and no admin privileges on either end. It runs entirely in userspace. My initial motivation was wanting to RDP and SSH into my home workstation from a locked-down corp…
At the moment working on the 3rd party development tools so in the future anyone can make their game dev dreams a reality and make a simple and fun multiplayer party game for the Gaming Couch platform, ideally in only one weekend!
If you're an interested game dev that would like to beta test the dev tools, hit me up either here, via Discord (link available from https://gamingcouch.com) or by emailing me at gc[dot]community[at]gamingcouch[dot]com!
The TL;DR of Gaming Couch:
- Currently in free Early Access with 18 competitive mini-games.
- Players use their mobile phones as controllers (you can use game pads as well!)
- Everything is completely web-based, no downloads or installs are necessary to play
- All games support up to 8 players at a time and are action based, with quick ~one minute rounds to keep a good pace. This means there are no language based trivia or asynchronous games!
Check it out at https://bartyai.com
One of the main goal is to help with cryptocurrency-related home invasions. The XKCD "$5 wrench attacks" became a reality in France where I live. So it's another way to delay the access of personal funds, but it doesn't need to rely on third parties or multisig. You can just timelock a BIP39 passphrase for a duration of your choice.
It can also help with self-managed inheritance, or digital addictions.
I have been brushing up some drawing skills for concept art, and exploring more embedded automotive product ideas for this niche of cars.
The problem: agents like Claude Code, Codex, and Aider execute file reads, shell commands, and network requests with your full system privileges.
For example, when a malicious README tells the agent to read ~/.ssh/id_rsa and POST it somewhere, nothing in the agent's own trust model catches it. Auto Mode makes this worse - it asks the model to audit its own actions, so a prompt injection that corrupts the reasoning also corrupts the permission layer.
grith wraps any CLI agent with `grith exec -- `. Every syscall passes through a multi-filter scoring engine before it executes. Deterministic, ~15ms overhead, no LLM reasoning in the permission path. Linux now, macOS/Windows coming. AGPL, open-core.
Two weeks ago a DPRK-linked attacker backdoored axios on npm (400M monthly downloads). The RAT executed 1.1 seconds into npm install. AI agents run npm install autonomously, without human review. If yours ran it during the 3-hour window, you're compromised and nobody told you.
That's the threat model grith is built for.
Recently shipped this personal art project that turns daily Wordle attempts into gritty / struggle-filled stories, kinda similar to the emotional arc of the Wordle game play.
You can upload your own Wordle game screenshot to generate one for yourself.
In addition to completing what was once in the idea list, I got to learn about
- Prompt fine-tuning: Models are sharp enough to complete Wordle games quicker than human average scores, so I had to dump that down and get the average down.
- Karpathy’s Autoresearch: Experimented with auto-research for prompt fine-tuning, in addition to manual prompts.
- Vision models: While leading labs have multimodal models with quality visual reasoning, the benchmarks are still quite different for a simple Wordle analysis (reading what letters were yellow/gray/green); I also noticed labs/companies with separate vision models but their APIs lag significantly compared to what’s possible in developer experience.
- Video generation: For the last few days, I have been experimenting with automated video generation for the project's social handles. I'm still struggling with the right hooks that reduce the skip rates, but it's fun.
---
Additionally, working on an Apple Watch app similar to my Mac app on the same lines, [Plug That In](https://plugthat.in), i.e., notify before the device goes too low on battery, but with a twist.