Skip to content
How I AIHow I AI

I let Codex control my browser so I don't have to

Today I’m walking you through one of my absolute favorite AI features right now: browser and computer use via Codex (the ChatGPT desktop app). I use this every single day, personally and professionally, and I wanted to share the specific workflows I’ve built, the moments that surprised me, and the mental model that makes it actually click. *What you’ll learn:* 1. How browser use and computer use work, and why the Codex desktop app plus Chrome extension is the combo I rely on 2. How I use Codex to QA my onboarding flow, including exhaustive mobile testing I would never do manually 3. Why under-prompting frontier models gets better results than detailed step-by-step instructions 4. How my husband EJ Lawless’s persona-impersonation trick surfaces friction points I can’t see as the builder 5. How I use browser use to get through my LinkedIn inbox without touching it myself 6. How I had Codex shop Free People’s sale and add 10 medium items to my cart (breastfeeding-friendly and Hawaii-ready) 7. How computer use can control iPhone mirroring so your Mac can technically operate your phone 8. Three more computer-use shortcuts: filling annoying forms, creating Google Sheets mid-workflow, and managing router *Brought to you by:* Runway—The creative AI platform for images, video and more: https://runwayml.com/howIAI Hyperagent—Deploy fleets of agents that handle real work: https://www.hyperagent.com/howiai *In this episode, we cover:* (00:00) Intro (01:46) What browser use and computer use actually are (03:08) Why I use Codex specifically and how the desktop app plus Chrome extension works (04:15) Use case 1: QA testing my onboarding flow (10:41) Results: 11 issues, one high-severity blocker, one Google Sheet with screenshots (12:10) Use case 2: persona testing (18:20) Use case 3: LinkedIn inbox, hands-free (20:37) Use case 4: AI personal shopper (23:47) Rapid-fire uses: forms, iPhone mirroring, router access from out of state, Google Docs (26:50) Wrap-up *Tools referenced:* • Codex (ChatGPT desktop app): https://openai.com/codex • Claude desktop app: https://claude.ai/download • Monologue (voice dictation for AI): https://monologue.app • iPhone mirroring (Apple): https://support.apple.com/en-us/111775 • Google Sheets: https://sheets.google.com *Other references:* • Jesse Genet episode (How I AI): https://www.lennysnewsletter.com/p/5-openclaw-agents-run-my-home-finances?utm_source=publication-search *Where to find Claire Vo:* ChatPRD: https://www.chatprd.ai/ Website: https://clairevo.com/ LinkedIn: https://www.linkedin.com/in/clairevo/ X: https://x.com/clairevo _Production and marketing by https://penname.co/._ _For inquiries about sponsoring the podcast, email jordan@penname.co._

Claire Vohost
Jul 22, 202627mWatch on YouTube ↗

CHAPTERS

  1. 0:00 – 1:01

    Hands-free work: letting AI drive the browser and computer

    Claire introduces the idea of “Look Ma, No Hands”: using AI to navigate websites and your computer when your hands (and attention) are busy. She previews three categories of examples—builder, productivity, and personal/fun use cases—powered by recent improvements in browser/computer control.

    • Motivation: reduce digital toil by delegating clicking/typing to AI
    • Browser/computer control has improved significantly in her experience (Codex + newer models)
    • Preview of upcoming demos: QA, productivity workflows, and personal tasks
  2. 1:01 – 1:31

    Sponsor: Runway (creative platform for images/video/content)

    A brief sponsorship segment explaining Runway as an end-to-end AI creative platform. Claire highlights speed from concept to deliverable and mentions enterprise adoption and a promo link/code.

    • Runway supports generating images, video, and creative deliverables
    • Positioned as fast and scalable for teams without ballooning budgets/timelines
    • Mentions notable customers/studios and the promo URL/code
  3. 1:31 – 3:02

    Browser use vs. AI-native browsers—and why Claire picks Codex

    Claire defines what “browser use” and “computer use” are: LLMs controlling mouse/keyboard to operate apps and websites. She contrasts this with AI-native browsers and explains why Codex (ChatGPT desktop app + extension) feels best for reliable computer control.

    • LLMs can control mouse/keyboard to operate your machine like a “virtual coworker”
    • AI-native browsers exist (e.g., Comet/Atlas), but she prefers using tools via Codex/Claude
    • Codex is framed as the most capable at full computer control in her workflow
  4. 3:02 – 4:02

    Setup and invocation: desktop app + Chrome extension, plus @browser/@chrome/@computer

    She explains the practical setup: install the desktop app and the official Chrome extension to enable control. Then she breaks down the three invocation modes in Codex—side-browser, Chrome-control, or full-computer control—and how choosing the right use case is the real key.

    • Requirements: desktop app installed + official Chrome extension in Chrome
    • Three control modes: @browser (side window), @chrome (controls Chrome), @computer (whole machine)
    • Success depends on selecting strong use cases for delegation
  5. 4:02 – 6:34

    Use case #1: AI-driven QA of a web onboarding flow (desktop + mobile)

    Claire demonstrates using browser control as a QA companion for a product onboarding flow. She asks the agent to test usability and mobile responsiveness, take screenshots, and log issues into a Google Sheet—mirroring how a thorough human tester would work, but more exhaustively.

    • Delegating manual UI QA to an agent for usability and responsiveness checks
    • Agent clicks through flows, tries dropdowns, error states, and viewport resizing
    • Goal output: screenshots + structured issue list in a Google Sheet
  6. 6:34 – 9:05

    QA philosophy: exhaustive edge cases and the power of under-prompting

    While the agent runs, Claire explains why this approach beats her typical “happy path” testing. She notes that newer frontier models often perform better with minimal instruction, planning their own coverage rather than following an overly prescriptive checklist.

    • Humans tend to test the happy path; agents can systematically probe failure states
    • Exhaustiveness: tries alternate paths (team vs. individual), required fields, and errors
    • Tip: under-prompt newer models for better autonomous planning and coverage
  7. 9:05 – 12:08

    QA results: 11 issues found, including a high-severity blocker—with a screenshot-backed spreadsheet

    The run surfaces a blocking validation problem and additional issues across desktop and mobile. Claire shows the resulting Google Sheet: prioritized findings, reproduction steps, remediation notes, viewport context, and embedded screenshots for tracking fixes and delegating follow-up work.

    • Finds a key bug: Continue clickable without required selection/validation (blocker)
    • Completes mobile testing and flags UI/UX issues (overflow, touch targets, accessibility cues)
    • Outputs a practical Google Sheet with steps, severity, and screenshots for each issue
  8. 12:08 – 16:10

    Use case #2: Persona testing your product with agent “fresh eyes”

    Claire shares a persona-driven testing workflow suggested by her husband: have the agent use the product as distinct user types, then write a research-style critique. The personas include a PM creating a PRD, an engineer turning it into a technical spec/prototype, and a team leader assessing adoption.

    • Prompting the agent to behave like multiple personas while navigating the app
    • Three personas: PM (create PRD fast), engineer (tech spec/prototype), team leader (usage oversight)
    • Deliverable: critique with friction points, delight moments, and improvement ideas
  9. 16:10 – 18:11

    Persona test findings: handoff friction, missing cross-reference patterns, and unclear loading/error states

    As the agent transitions from PM to engineer, it reveals a key product friction: difficulty referencing one created document in another thread. The run also surfaces UX feedback around slow/opaque loading and unclear system status during generation, especially when errors occur.

    • Persona handoff exposes structural expectation gaps (document referencing/mentioning)
    • Agent behavior highlights what a new user assumes will work vs. what actually works
    • Feedback includes loading transparency, performance perception, and failure-state clarity
  10. 18:11 – 20:42

    Use case #3: LinkedIn inbox triage and drafting replies without an official API

    Claire shows a hands-free workflow for processing LinkedIn DMs: prioritize critical messages, draft friendly replies for simple notes, and leave guidance for the rest. She notes this is useful because LinkedIn lacks a convenient official API/MCP workflow, and she adjusts model “effort” for speed vs. quality.

    • Automates inbox triage: respond to critical items, annotate others with suggested replies
    • Works around lack of official LinkedIn integrations by using browser control
    • Model calibration: lower effort/faster models may be sufficient for routine replies
  11. 20:42 – 23:43

    Use case #4: AI personal shopper in Chrome (with human verification when challenged)

    Using @chrome, Claire asks the agent to shop Free People’s sale for Hawaii-appropriate, breastfeeding-friendly items and add 10 options to her cart without checking out. The demo highlights both the convenience and real-world friction like bot/agent verification steps where the human must intervene.

    • Shopping prompt constraints: size medium, comfort, nursing-friendly, Hawaii weather
    • Agent navigates sale pages, selects items, and populates the cart automatically
    • Caveat: bot checks/verification can require the human to step in as the “hands”
  12. 23:43 – 26:16

    Rapid-fire additional uses: forms, controlling your phone via mirroring, remote router fixes, and creating Docs/Sheets via the browser

    Claire closes with practical miscellaneous scenarios where computer use shines: filling tedious forms, operating an iPhone through Mac mirroring, and handling remote network/router tasks while out of state. She also notes that when plug-ins/MCPs fail, letting the agent do the task directly in the browser can be the most reliable path.

    • Automate painful form-filling (procurement, camps, legacy websites)
    • Use iPhone mirroring + computer control to operate phone apps from the desktop
    • Remote ops story: adjusting router/firewall settings, SSH access, then closing ports
    • Fallback tactic: create Google Docs/Sheets directly via browser when integrations fail
  13. 26:16 – 27:40

    Wrap-up: why computer use changes digital life (and a call for privacy/security questions)

    Claire reflects on the arc from QA debugging to personal shopping, arguing that AI-driven computer control will reshape how we interact with digital systems. She invites viewers to share their own use cases and asks for questions about privacy and security for future coverage.

    • Computer use reduces time-on-keyboard and enables work to happen in the background
    • Framing: a major shift in human-computer interaction
    • Invites comments: use cases plus privacy/security concerns and follow-up topics

Get more out of YouTube videos.

High quality summaries for YouTube videos. Accurate transcripts to search & find moments. Powered by ChatGPT & Claude AI.