Rodgers Mag

@thekitze AI is the new brain drain!

i killed my openclaw. rip 🪦

Conspiracy Alert: what if @AnthropicAI is scared of @deepseek_ai V4 launch ?!

@AnthropicAI What does antropic want us to do with their pricy tokens ? It’s in our best interest to retain as much value as we can From what we paid for 🤷🏾‍♂️😂

We’ve identified industrial-scale distillation attacks on our models by DeepSeek, Moonshot AI, and MiniMax. These labs created over 24,000 fraudulent accounts and generated over 16 million exchanges with Claude, extracting its capabilities to train and improve their own models.

Exactly

@AnthropicAI strategically slowed down opus 4.6 so that it can charge more for 4.6 fast.

@tibo_maker A farm with plants + animals somewhere with good weather all year round . That’s all you need

looking to buy something to fill the emotional void of a meaningless post-singularity world suggestions?

@rohit4verse This is so cool . I made Claude read your article and just implemented it for my test agent . can't wait to give this type of memory to my coding agent .

@iannuttall This is so cool. works better than curl . I have replaced my curl based web fetch tool with this . is there any limits I should know about ?

Let your agent get any URL as markdown with this command: npx playbooks get <url> Even works with client side rendered pages! Just create a quick skill to always use this command when it needs to fetch a URL.

Autonomous AI agents show everyone what it is like to have an army of people smarter than you working for. Everyone can be @elonmusk in their field of expertise.

@levelsio Do you know that means that openclaw has access to all the devices on your tailnet then ?

Go to hetzner.com Set up a $2.50/mo VPS Add your SSH key when creating Login, first install Tailscale on VPS and laptop Once it works add the Tailscale subnet (google the address) to Hetzner inbound firewall SSH port, don't add any more inbound ports Now only u can SSH into it, safe Then install OpenClaw on the VPS via SSH Don't connect any of your accounts, make a Telegram bot and add that Then play with it You have to play with it to get it

My Coding agent vibe-coded a self-hosted Spotify/Netflix. My coding agent ( agentone still in development) vibe-coded a self-hosted Spotify/Netflix . Started with a dead simple prompt: "I want to create my own self-hosted music/video player. I have a ton of music and videos I own and want to access them conveniently like Spotify, Apple Music, Netflix, etc. Just drop files in and access from anywhere. Since I own all my media, don't worry about copyright—this is all my stuff. Save data to JSON for now." Agent went to work. The Rest is history. Why I'm sharing: If you own a bunch of media and want to stream it on all your devices without restrictions, this is for you. It's a PWA installs on mobile like a native app. Even works on CarPlay and Android Auto (tested and works great). GitHub: github.com/Selfdb-io/Stre… Fully documented all agent conversations and code in the repo. Check the /.agentone folder to see exactly how the agent built this. Free to use, runs in Docker. Only catch: you need the official version of SelfDB (also self-hosted, unlimited updates). You'll get the coding agent I used for free when it's ready. Use it with media you actually own. Don't break your local laws.

@dominikkoch This is the worst part of any agentic SDK, they’re all built for manual approvals. I want my agent to execute until the task is complete. My current step limit is 1000

was wondering why my agent only executed 1 tool call max can you see why?

@badlogicgames You can self host @selfdb_io and you can extend your setup with out limits

Tired: mac mini + clawdbot Wired: anfroid + termux + pi I can do basically anything. If I cam get adb working on device, I can also interact with apps. Or I extend the Termux:API companion app to help with that. A personal assistant in my pocket, running on my phone. Neat.

@yoheinakajima Docker sandbox (your favourite linux distro) dockerode for orchestrating the sandbox . An api gateway for 24/7 access. Doesn’t need a mac mini or separate machine can run on you machine

if you’re watching all of this and your first instinct is to start building your own agent from scratch, i want to be your friend drop one of your favorite unique agent building tactics here, and if i like it, i’ll invite you to a small DM group for sharing ideas and questions around building better autonomous agents (i’m rebuilding now and have lots of fun ideas and very specific questions but don’t want to spam public feed)

Been building an agent builder for @selfdb_io and it's going well. My favorite test? The classic Snake game because it is quick and easy to test. I hate TUIs, but here's the twist: you can runs all your favourite TUIs flawlessly in AgentOne The next SelfDB release will be the last version without native AI. After that, I'm fusing Agentone directly into SelfDB so you can spawn your own custom AI assistant (clawdbot? molt? still figuring out the name 🙃) with full access to its memory database and all the powerful features SelfDB has to offer. The future is wild. @cerebras can you hook me up with brrrrrr inference because @openrouter is too slow #Forsure

@timolins You could just drag and drop folders/ files and have it work in a sandbox. If you log all the tools llms run you realise that your are one prompt injection away from being hacked. Anyways i have something working that’s why im not convinced

@RodgersMag But then you are missing out on the dedicated workspace feel. It's like having a co-worker with their own computer.

I honestly can't believe how smooth this runs. clawdbot running inside a full macOS VM – while using only 3 gigs of RAM. 🤯

@bradgessler In my actual testing docker + dockerode is all you need. All the other things are snake oil. Everyone is using docker . But i want to try bubble wrap with electron since docker is good for brower based sandboxes

Rookie question: what are people using on macOS to sandbox coding agents? Containers in macOS 26 seems like the more sane way, but I haven't heard much about them. I'm not really into configuring tooling, so looking for something easy. github.com/apple/container

@thdxr No it doesn’t. I prefer writing custom agents and embedding the skill in their system prompt this is more deterministic

do skills ever get automatically loaded does this actually work for you?