I've been exploring the idea of narrow harnesses for AI agents. Coding agents such as Codex and Claude Code are good examples of agents operating inside highly specialised environments with tools, context, feedback and verification.
This made me wonder what happens when we apply the same idea outside coding. For example, a Kubernetes harness whose only responsibility is to observe deployments, investigate failures and verify recovery.
The article explores whether specialised harnesses could become an interesting alternative to increasingly general-purpose agents.
No actually. Shyell is like performance and observability shell. Its features are real-time CPU/memory monitoring, automatic project stack detection, and that benchmarks every command you run.
Following some feedback about the original name being already taken, I have rebranded the project and released v0.3.0. This release isn't just a rename, it is s a complete refinement of the installation and cross-platform experience.
Homebrew is almost ready. Preparing for a Homebrew release (coming in v0.4.0).
I would love to hear your thoughts on the new name and the tabbed approach to monitoring.
Perfmon essentially allows you to define commands you use daily, defined them in a toml file and tab them, ie customizable.
It could be docker or kubernetes commands and so on. This is different from btm.
I built Perfmon because I found myself constantly switching between multiple terminal tabs to check different system metrics (top for processes, vmstat for memory, netstat for traffic, etc.). I wanted a single, unified view that didn't just show a hardcoded set of stats, but let me bring my own tools.
What makes it different:
Unlike static monitors like htop or btop, Perfmon is designed around a tabbed interface where each tab is just a shell command defined in a TOML config. If you have a specific script or a grep command you run every 5 seconds to check a log or a metric, you can just drop it into the config, and it becomes a tab in your dashboard.
Key Features:
- Extensible: Define any shell command as a tab.
- Live Metrics: Real-time sparklines for CPU, Mem, Load, and Network (works on Linux, macOS, and Windows).
- Modern TUI: Built using the Go bubbletea framework for a clean, responsive feel.
- Lightweight: Minimal resource footprint.
good q! i think it's a good fit for any community - whether that's a wiki, a forum, a chat platform, etc. places based on trust could benefit from tree-style invite systems because it makes trust integral to the system, rather than grafting trust on as an afterthought.
BookGraph demonstrates that the next leap in AI isn't just "smarter models", it's better context. By combining the reasoning power of LLMs with the structural integrity of Graph Databases, we move from a world where we "search" for information to a world where we "interact" with intelligence.
I built HN Status Bar, a small VS Code extension that shows Hacker News stories in the status bar while coding.
What it does:
- Shows one HN story at a time
- Rotates automatically (default 20s)
- Refreshes from the HN API
- Click current story to open it
- Configurable settings (timing, item count, score visibility)
This made me wonder what happens when we apply the same idea outside coding. For example, a Kubernetes harness whose only responsibility is to observe deployments, investigate failures and verify recovery.
The article explores whether specialised harnesses could become an interesting alternative to increasingly general-purpose agents.
reply