Claude Opus 5 and Codex Browser Automation Experiences
Discover how AI is transforming daily workflows through automation.
1Comprehensive Testing with AI
Claire achieved significantly more thorough results in application testing using Codex compared to human testing. While users typically follow the 'happy path' in traditional tests, AI can test from both individual and team perspectives. By testing edge cases in required fields, it uncovered a critical bug that had gone unnoticed for months. This approach has the potential to revolutionize QA processes, offering more robust testing coverage than manual methods could provide.
2Models with Thinking Space
Experiences show that AI models perform better when given more thinking space. Initially providing 25 specific test items for browser automation, Claire now gives general instructions like 'QA the onboarding flow,' allowing the model to develop its own testing strategy. This freedom enables broader test coverage without the limitations of human assumptions, often revealing issues that would otherwise remain hidden.
3Power of Persona Testing
EJ's persona testing approach uncovers friction points that synthetic user research misses. By directing AI to use products as specific roles (a PM post-meeting, an engineer reviewing PRDs), structural UX issues become apparent. In ChatPRD, this method clearly revealed a cross-thread reference flow problem that Claire knew existed but had never experienced so vividly from the user perspective.
4AI for LinkedIn Management
Codex offers practical solutions for managing LinkedIn inboxes. While Claire initially used high-compute models, she discovered medium-level models could effectively process unread messages, draft context-aware replies, and flag items needing personal responses. This proves particularly valuable for users with hundreds of messages and no official API connection, saving significant time on routine communications.
5Remote iPhone Control
AI browser automation can even control iPhones through screen mirroring. While traveling, Claire needed to manage home Wi-Fi settings installed on her California phone. Codex mirrored the iPhone, updated router settings, completed SSH setup, and closed ports afterward - a series of complex tasks nearly impossible to perform manually from a hotel room, demonstrating AI's potential for remote device management.
6Matching Models to Tasks
Different tasks require varying model capabilities and compute levels. Sorting LinkedIn messages demands different reasoning than writing production code or conducting QA tests. Selecting appropriate compute resources makes workflows faster and more cost-effective, especially crucial for all-day browser automation. This optimization becomes increasingly important as AI integrates deeper into daily operations.
7Human-AI Collaboration
When websites block the AI agent, humans can briefly intervene. During a Free People shopping session, when the site flagged Codex as a bot with CAPTCHA, Claire completed verification before returning control. This demonstrates efficient division of labor - AI handles tedious browsing while humans manage identity verification moments, creating a symbiotic workflow that plays to each party's strengths.
8Hardware Projects with Cursor and Raspberry Pi
Maddie Reese built a working Twitter pager, AI-powered receipt printer, and personal API with zero coding background, using Cursor and Raspberry Pi. Comparing her coding ability to basic Spanish survival skills, Maddie shows how technical depth isn't prerequisite for hardware projects. She can review code direction and spot errors like incorrect wire specs without writing functions from scratch.
9From Idea to Implementation
Maddie's process begins with explaining ideas to Cursor for interrogation before purchasing components. Only after resolving major questions does Cursor generate shopping lists. When it recommended wrong wires once, Maddie caught the mistake during final review, preventing wasted purchase and troubleshooting. This workflow - idea, interview, shopping, build - successfully guided her thermal printer, pager, and API projects from concept to reality.
10Future of AI-Augmented Creativity
Claire and Maddie's experiences demonstrate how AI democratizes technical fields while transforming creative processes. From browser automation to hardware projects, AI-enhanced workflows boost efficiency while redefining human focus areas. As these technologies proliferate, they promise fundamental changes in how we work and create, potentially reshaping entire industries through accessible, intelligent augmentation.
Original article: Lenny's Newsletter — Claude Opus 5 ve Codex ile Tarayıcı Otomasyonu Deneyimleri