On July 8, OpenAI shipped GPT-Live.

The biggest change from the old voice mode: it can listen and talk at the same time. You can interrupt mid-sentence. You can pause to think without it jumping in. Conversations feel like talking to a person, not waiting for a turn.

I tested it by learning Jason's verifier-setup Skill. I dropped the Skill file into ChatGPT, switched to GPT-Live, and said:

"You're my tutor. Walk me through this from the top. Start with the purpose, then cover the full workflow. Don't skip important details. If something is abstract, give me an example."

It explained that verifier-setup scaffolds a verification system for a project. Then it walked me through the full flow: investigate the project, confirm the start command and Driver, choose local or Sandbox, then generate the /verify Skill.

I interrupted whenever I needed to.

Later I asked:

"Say I just finished building a Login feature. What happens next?"

It immediately reframed the explanation around that scenario:

Build Login
-> Run /verify
-> Start the project
-> Playwright tests the login flow
-> Confirm it lands on Dashboard
-> Capture screenshots and recordings
-> Attach evidence to the PR

Five minutes in, I understood what the Skill does and how it runs in a real project.

GPT-Live works like a private voice tutor you can talk to anytime.

My workflow now: drop a new concept or long article into ChatGPT, let GPT-Live explain it start to finish. When something doesn't click, I interrupt, ask for a different example, or tell it to connect the explanation to my current project. When I need exact commands, I check the source doc.

Information keeps growing. Understanding is the bottleneck. GPT-Live turns dense material into a conversation where you can ask questions in real time.

Favorite way I've found to learn so far.

Links: