┆dev system
┆dev web
┆dev agentic (2026+)
Develop a minimalist agentic system (slimmer than Openclaw) to minimize the attack surface. Enhance GUI interaction capabilities, as CLI is already set. Create a 'skills.md' document.
Software engineer (SWE) is basically playing Factorio in a terminal right now.
If you use agentic AI, you know exactly what this means. We went from manual artisans to automation supervisors.
In Factorio, you build machines and conveyor belts to automate everything.
In our terminals, we now do the same: we give a prompt and watch the AI autonomously grep, read, edit, and run tests. Terminal logs are the new conveyor belts.
Our job has completely changed :
⤫ Typing code manually
✓ Managing AI context & API budgets
✓ Unblocking the AI when it loops
✓ Breaking down big tasks into automated agent pipelines
We are no longer coders.
We are algorithmic factory managers
Equipment
Scope : Askar 103APO
Lens : Askar 0.6x Reducer
Camera : ZWO ASI 294MC Pro
Filter : Antlia TriBand RGB Ultra
Mount : Skywatcher AZ-EQ5 GT
Guiding : Skywatcher EvoGuide 50ED with ZWO ASI 224MC
Controller : ZWO ASIAir Pro
Focusser : ZWO EAF
Color Palette : HSO
Package Manager
Run Claude Code, Codex, or any AI coding agent in "yolo mode" without nuking your home directory.
Runtime
macOS
Docker Desktop, OrbStack, Colima, or Apple container
Linux
Docker or Podman
Anatomy of a Linux Desktop
An agent operating in a sandbox runs in full YOLO mode 100% autonomous, without asking the engineer for guidance. That’s why a sandbox environment is so crucial for agents.
Devboxes, hot and ready
Minions : Stripe’s one-shot, end-to-end coding agents
Kernel ‒ WM ‒ Shell
---------------------
nixos
Ϟ Linux ‒ Niri ‒ nushell
macos
Ϛ XNU ‒ Quartz Compositor ‒ zsh
windows
Ͻ Windows NT ‒ Desktop Window Manager ‒ powershell
Claude Code Take the Wheel, I prompt, you code, that's the Deal. ♪
Containerized hacking environments for professionals, students, CTF players, bug hunters, researchers, ... and for you
stereOS is a Linux OS purpose-built for AI agents, offering better scaling and security than Docker or VMs. It uses NixOS and gVisor to run agents in isolated sandboxes with virtual kernels.
This architecture protects the host hardware from breaches and allows agents to safely spawn sub-agents in a read-only environment.
