MyJackfruit

Help
Browse topics · Watching and following the work

Watching and following the work

← Watching and following the work

How do I know what my AI is working on right now, and whether it is stuck?

Short answer

Read the working line in GilbApp's chat or your terminal; answer any question waiting for you; if nothing moves for long, ask it.

Before you start

Words used here (1)
GilbApp
GilbApp is the app, opened in a web browser at https://gilb.com/app, where you open, fill and share your vaults.
Deeper

Also asked:

My AI has been working for a long time. Is it stuck? Most often it waits for you: look for a question or an Approve card. A hand-off between agents may run up to 15 minutes; if nothing moves, ask it with Slip in now in the chat, or press Esc in Claude Code.

In GilbApp's chat. While your AI works, a line with three moving dots stays under the conversation and names what it does at that moment: The agent is writing…, or Ran and the name of a tool, one action your AI can use, such as reading a record. With Claude Code connected and a code folder (the folder on your computer where your AI writes your app's code) switched on with Code in, the dots give way to a line that starts Working and shows the time and the last act on your computer: Writing and a file, Reading and a file, Running: and a command, or Searching. Stop under that line stops the answer that is running. While the line shows, the AI is still working, and Send reads Queue. A message you send then waits under Waiting to send until the agent is done; Slip in now hands it to the agent at its next step, so you can ask what it is doing.

In a terminal, the window where you type commands. Claude Code shows each act as it goes, and its to-do checklist, the list of steps it plans; Anthropic says Ctrl+O Shows detailed tool usage and execution (https://code.claude.com/docs/en/interactive-mode, read 13 November 2026). OpenAI says Codex lets you inspect commands and diffs as they appear (https://learn.chatgpt.com/docs/codex/cli, read 13 November 2026).

Which step. Kai-Zen, the planning agent (an agent is your AI following one written procedure for one kind of work), hands each step to Agent.Coder, the agent that writes your app's code, and reports to you when the work you asked for ends. When the plan is kept in a vault, the step's card in GilbApp's Plan shows its Status, where it stands: Building while it is being built. The answer How do I write a step? shows the card. To know where the project stands, ask Kai-Zen.

Stuck, or waiting for you? Most often it waits for you. In the chat, before the first answer in a code folder, a card asks Let this AI read, change and run commands in the folder for this session?, and your message is not sent until you press Approve (the AI may work in the folder for the whole session), Approve before each write (it asks before each write and command) or Decline (the message goes as a plain chat message and Code in is switched off). With Approve before each write, a card asks before each write or command; one not answered in 10 minutes counts as declined. In a terminal, Claude Code or Codex may ask to allow a tool or a command, and waits. Kai-Zen's message may end with a question: the answer Kai-Zen says something I do not understand, or stopped. What now? lists when it waits.

Long is not always stuck. When one agent hands work to another, the hand-off may run up to 15 minutes, and up to 30 for some planning agents; if it runs out of time, what it wrote before then is saved. With Claude Code, a long job can run out of tool rounds, the steps it may take in one turn. The chat then goes on by itself, up to 15 times, and says "The turn ran out of tool rounds, so I am going on by myself (1 of 15). Press Stop to stop." Stop stops it after the turn that is running; send continue to go on.

If it seems stuck. In Claude Code, press Esc: Anthropic says it stops "the current response or tool call mid-turn so you can redirect. Claude keeps the work done so far." Then ask what it was doing. In the chat, with Claude Code and Code in switched on, press Stop under the Working line: it stops the answer that is running, and "You pressed Stop. Send continue to go on." shows above the box. Otherwise no button stops an answer already running; ask with Slip in now. If an error line shows above the box, such as The agent could not answer., send your message again, and if the line stays, write to [email protected] with what you sent, the line, and the time.

This answer is written for a computer screen.

Source: GilbApp d5296208 (https://gilb.com/app/version.json, read 15 November 2026): src/components/agent/AgentChat.tsx:2072-2076 (The agent is writing…; a code turn shows Working and Stop), :1506-1517 (Stop ends the running turn), :2127 ('You pressed Stop. Send continue to go on.'), :2151-2152 (Waiting to send), :2240 (Slip in now), :2333 (Queue, Send), :260 and :2104-2128 (15 automatic turns, Stop); src/components/agent/turnActivity.ts:34 (Ran), :75 (Working · and the time); src/components/agent/CodeFolderParts.tsx:35-42 (act lines), :86-108 (Let this AI read, change and run commands in … for this session?, Approve, Approve before each write, Decline); src/components/agent/AgentWorkingIndicator.tsx:60-80 (three dots, the Working line, Stop); src/lib/base/planCard.ts:31 (Building). Claude Code bridge 0.2.0 README (a write or command question not answered in 10 minutes is declined). gilb-connector 0.6.1 (https://gilb.com/app/gilb-connector.mjs, read 13 November 2026): hand-off limits 15 and 30 minutes; What it wrote before then is saved. Standards: Kai-Zen.Cycle (Build), Kai-Zen.ValueDeliveryStepRules (Status). Makers, read 13 November 2026: https://code.claude.com/docs/en/interactive-mode, https://learn.chatgpt.com/docs/codex/cli. Live answers, sentences shared: Kai-Zen says something I do not understand, or stopped. What now?, How does Kai-Zen run a whole delivery cycle?.

Checked against GilbApp d529620.