🎙️ How I AI: Claude Opus 5 Review + Browser use in Codex + How Cursor and a Raspberry Pi makes AI fun

2026-07-27 · Source: Lenny's Newsletter · Field: Technology & Digital — Artificial Intelligence & Machine Learning, Software Development & Engineering, Robotics & Autonomous Systems · Depth: Intermediate, medium

Summary

The content examines three distinct AI applications and evaluations. It details how Codex enables advanced browser automation for exhaustive web app QA testing, uncovering a months-old bug, and conducting persona-based user research. Codex also efficiently managed LinkedIn inboxes and remotely controlled an iPhone via screen mirroring, with human intervention for CAPTCHAs. Second, it highlights how Cursor, combined with a Raspberry Pi, empowers individuals with minimal coding experience, like Maddie Reese, to develop hardware projects such as a Twitter pager and a personal API, following a structured workflow. Lastly, a review of Claude Opus 5 shows its top performance in a seven-model blind benchmark, scoring 78, ahead of Claude Sonnet 5 (77) and GPT-5.6 Sol (76). Despite its capability, Opus 5 exhibits a timid personality and verbose output, suggesting optimal use in asynchronous agentic coding. Gemini 3.1 Pro scored lowest at 32.

Key takeaway

For AI Product Managers evaluating new models or designing agentic workflows, recognize that raw capability is becoming table stakes. Focus on a model's personality and operational efficiency, like Claude Opus 5's asynchronous strength, to optimize integration. You should also explore browser automation with tools like Codex for exhaustive QA and persona testing, and consider AI-assisted hardware development with Cursor for novel applications, even with limited coding expertise.

Key insights

AI's utility extends to browser automation, hardware development, and nuanced model personality assessment, revealing new operational efficiencies.

Principles

• Frontier models benefit from room to think.
• Match model effort level to the job.
• Model personality reflects company culture.

Method

For AI-assisted hardware projects, follow a sequence: brainstorm the idea with AI, conduct an AI interview, generate a shopping list, purchase components, then build.

In practice

• Use AI for exhaustive web app QA testing.
• Automate LinkedIn inbox triage with AI.
• Build hardware projects with AI, minimal code.

Topics

• AI Browser Automation
• Large Language Models
• AI-Assisted Hardware
• Software Quality Assurance
• Model Evaluation Benchmarks
• Agentic AI Workflows

Best for: Machine Learning Engineer, NLP Engineer, Product Manager, AI Engineer, AI Student, AI Product Manager

Related on AIssential

• Claude Opus 5 Is THE GREATEST AI Model EVER?! Beats Fable & CHEAPER! (Fully Tested) — Claude Opus 5 offers frontier-class AI capabilities at significantly lower cost than competitors, excelling in complex reasoning and autonomous code generation.
• OpenAI Codex Can Now Use Your Computer! (Full Breakdown + Testing) — OpenAI's Codex is an evolving AI agent designed for non-technical users, excelling in browser and computer control.
• That's my designer - Claude — AI product development focuses on enhanced multimodal capabilities, agentic workflows, and specialized design tools.
• How to Use Opus 4.7 and the New Codex — New AI models and applications enhance autonomous agent capabilities for diverse knowledge work tasks.
• I hate Opus 5. It’s the best model, anyway. — Opus 5 excels in output quality for design tasks despite a "neurotic" and verbose interaction style.
• Opus 4.7 Part 2: Capabilities and Reactions — Claude Opus 4.7 offers enhanced intelligence and autonomy but requires careful handling due to its distinct behavioral quirks.

See Counsel's argued verdicts on the open AI decisions leaders are weighing →

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by Lenny's Newsletter.