# What is my AI doing when it says "building" or "running tests"?

Checked 2026-11-15 against GilbApp d5296208bbffa2ab70076e6194f5537c4be1ddf7.

Building makes the finished files your app is put live with; running tests runs small checks that try each part and say pass or fail.

## Deeper

Your AI coder, such as Claude Code or Codex, works in small acts, one after another: it reads files, writes and changes code, searches, and runs commands, the instructions a computer carries out. "Building" and "running tests" are two such commands.

Building. Your app's code is written in a form people can read and change. Building turns it into the finished folder of files a browser opens; that folder is what is put live. "Building" can also simply mean that your AI is writing the code for a step.

Running tests. A test is a small program that tries one part of your app and says whether it did what it should: pass or fail. For an app on GilbPlatform, Agent.Coder, the agent that writes your app's code, follows a testing standard, a written rule for its tests: every part that other parts use gets tests, and so do errors such as bad input or a lost connection. Tests that try the app in a browser are part of it. Then Agent.Tester, the agent that tests what others built, tests the step again as a separate agent that knows nothing of how it was built, and saves a PASS or FAIL. When a step changes a screen, Agent.Coder also opens that screen in a real browser at phone size, takes a screenshot and counts the errors. The answer [What counts as proof that something works?](http://support.gilb.com/help/what-counts-as-proof-that-something-works) says more.

What you see while it works. In a terminal, Claude Code shows each act as it goes; 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" \(a diff shows the lines a change added and removed\) \(https://learn.chatgpt.com/docs/codex/cli, read 13 November 2026\). In GilbApp's chat, with [Claude Code connected](http://support.gilb.com/help/how-do-i-connect-my-ai-to-gilbapp-chat) and [a code folder](http://support.gilb.com/help/how-do-i-let-my-ai-build-in-a-folder-on-my-computer-from-gilbapps-chat) \(the folder on your computer where your AI writes your app's code\) switched on, the line under the conversation names each act as it happens: "Writing" and a file, "Reading" and a file, "Running:" and a command, or "Searching". Without a code folder it reads "The agent is writing…", or "Ran" and the name of a tool. The answer [How do I know what my AI is working on right now, and whether it is stuck?](http://support.gilb.com/help/how-do-i-know-what-my-ai-is-working-on-right-now-and-whether-it-is-stuck) says more.

Do I need to do anything? Usually not: let it run. If you chose "Approve before each write" for the code folder, the chat asks "Allow this AI to run this command?" and shows the command, with "Allow" and "Decline"; Claude Code in a terminal can ask the same. A command that installs, builds or runs is normal while building. If you are not sure, press "Decline"; your AI goes on without that command, and you can then ask it what the command does; the answer [Can I trust what the AI suggests and builds?](http://support.gilb.com/help/can-i-trust-what-the-ai-suggests-and-builds) says what to allow.

This answer is written for a computer screen.

Source: GilbApp 49b1ae53 \(https://gilb.com/app/version.json, read 13 November 2026\): src/components/agent/CodeFolderParts.tsx:31-43 \("Writing", "Reading", "Running:", "Searching"\), :93 \("Approve before each write"\), :126-140 \("Allow this AI to run this command?", "Allow", "Decline"\); src/components/agent/AgentChat.tsx:716, :1248 \("The agent is writing…"\), :755 \("Ran" and a tool\). Standards: Rule\_Code\_testing \(RuleTestTools, RuleTestWhatGetsTested\), Proc\_v\_i\_c\_PublishPagesDev \(build the static folder, then publish it\), Agent.Coder \(AfterIncrement\), Kai-Zen.Cycle \(Check\). Makers, read 13 November 2026: https://code.claude.com/docs/en/interactive-mode, https://learn.chatgpt.com/docs/codex/cli. Live answers, sentences shared: How do I build an app with an AI coder?, What counts as proof that something works?, Kai-Zen says something I do not understand, or stopped. What now?.
