Three AI tools for practicing, planning and collecting

A spoken answer on a screen, a coding task broken into an editable plan, and a collection of links that suggested what to explore next: these three tools made machine-learning assistance visible at different points in a person's work.

Interview Warmup: hear a question, review an answer

Google's Interview Warmup offered practice questions for people preparing for job interviews. You could speak an answer and see it transcribed in real time, then review patterns such as frequently used words, job-related terms and the talking points covered. Google said the answers were not graded: the feedback was there to help people reflect and try again. Sets included questions related to Google Career Certificates as well as general questions. The original Google introduction describes the experience; this is not a test we performed.

Google later retired the standalone tool. An archived retirement notice dated April 23, 2026 establishes that it was retired by that observation; it does not establish the final access day. Google’s continuing interview advice and certificates are separate. Interview Warmup archive.

Copilot Workspace: make the plan editable

GitHub Copilot Workspace began with a repository task, such as an issue. Its 2024 technical-preview introduction showed a workspace moving from the issue to a specification and a proposed plan. A developer could edit the plan and generated changes, use an integrated terminal to run checks, and then create a pull request. The useful object was not just a chat answer: it was a sequence of editable steps tied to a repository. GitHub's original walkthrough shows that sequence in context.

GitHub Next says the Workspace technical preview ended on May 30, 2025. The retirement concerned this hosted workspace, not GitHub Copilot as a whole or ordinary pull requests already created. Copilot Workspace archive.

Keen: make a collection that keeps looking

Keen was a Google Area 120 experiment for gathering things around an interest. A person could create a “keen,” save useful links, organize them and choose whether to share the collection. Google described using Search and machine learning to suggest more material based on what had been saved. In co-founder CJ Adams's introduction, birding resources and plans for a flower farm explain the kind of personal interest a collection could hold.

Keen closed in 2024; its collections were part of the discontinued service, not Google Search itself. Keen archive.

The inputs were different—a voice answer, a repository issue and a handful of saved links. In each interface, the person could look at an intermediate result and decide what to do next.