cd /news/ai-agents/the-terminal-should-not-own-the-work · home topics ai-agents article
[ARTICLE · art-130031] src=lezli01.is-a.dev ↗ pub= topic=ai-agents verified=true sentiment=↑ positive

The terminal should not own the work

Vincent moves agentic task ownership from the terminal into a background daemon that owns task state, workflow execution, agent processes, scheduling, the database, and git worktrees, with the TUI, CLI, and API acting only as clients. The daemon persists each state transition before acting, so a restart records the interrupted attempt, verifies any orphaned process before stopping it, and reruns the step without consuming a failure retry. Installing the daemon as a user service lets the operating system start the control plane across logins, decoupling long-running agentic coding workloads from the lifetime of the interface that launched them.

by read2 min views1 publishedSep 15, 2026
The terminal should not own the work
Image: source

A terminal is a good place to start a process. It is a poor place to store the truth about that process.

When an agentic task belongs to one terminal tab, the tab becomes an accidental control plane. Closing it can stop the work. Losing scrollback can erase the useful explanation. Reopening the project means reconstructing what ran, which branch it changed, and whether it was waiting, finished, or quietly stuck. The longer the task runs, the more fragile that arrangement feels.

Vincent moves ownership into a background daemon. The daemon owns task state, workflow execution, agent processes, scheduling, the database, and git worktrees. The TUI, CLI, and API are clients of that state. Closing any client changes nothing about the work behind it.

That architectural choice has consequences that are easy to feel. I can start a task, close the TUI, use the terminal for something else, and return later to the same state and history. Several clients can inspect the same daemon without becoming competing writers. The scheduler can admit tasks by priority and concurrency limits even when nobody has a dashboard open.

Durability also changes how interruption is handled. Vincent persists a state transition before acting on it. If the daemon stops during a step, restart recovery records the interrupted attempt, verifies any orphaned process before stopping it, and runs the step again without consuming a failure retry. The system does not pretend that a process survives a machine restart; it makes the interruption explicit and recovers from known state.

Installing the daemon as a user service extends the same model across logins. The operating system starts the control plane, and the terminal returns to being what it should be: one optional window into the work.

This matters because agentic coding is increasingly a workload rather than a conversation. Tasks wait on quotas, gates, retries, child branches, and external checks. Their lifetime should not be coupled to the lifetime of the interface that launched them.

The terminal can show the work. It should not own the work.

── more in #ai-agents 4 stories · sorted by recency
── more on @vincent 3 stories trending now
sponsored brought to you by zahid.host 4,200+ EU-deployed projects
reading about agents? ship yours in a single git push.

Run your AI side-project on zahid.host

EU-based hosting, git-push deploys, automatic HTTPS, no cold starts. Free tier with a custom domain — perfect for shipping the agent you just read about.

$git push zahid main
Live at https://your-agent.zahid.host
Get free account → Pricing
from €0/mo · no card required
LIVE [news/the-terminal-should-…] indexed:0 read:2min 2026-09-15 ·