People who learn to collaborate with AI systems effectively will not just write…

People who learn to collaborate with AI systems effectively will not just write code faster. They will fundamentally change how software is learned, designed, reviewed, and shipped."
""
"New ways of learning and working in the age of Gen AI advancements:"
""
"## For learning"
""
"Use agents like ChatGPT or Claude Code. Prompt them for a phase-wise learning plan on the topic of your interest, revealing only one concept at a time, with each concept followed by a small practice lesson."
""
"While learning each concept, it is IMPORTANT to sit with a pen and notebook and maintain handwritten notes. Not digital notes on tools like Obsidian, but literally pen and paper. There is something about physically writing things down that slows your thinking just enough to help concepts stay with you longer."
""
"For practice lesson, Use IDEs like VSCode or Cursor with AI-assisted auto-completion enabled. Describe your implementation ideas in plain English. Use tab-tab-tab... for implementation, and then review the generated code to understand how frameworks are abstracting or not abstracting your high-level thinking."
""
"## For implementing software features"
""
"Use agents like ChatGPT or Claude Code for ideation, generating developer specs (and checking them into the repo for future Spec-Driven Development), and creating code prompts or JIRA tickets to implement one atomic vertical slice at a time."
""
"Use Claude Code or GitHub Copilot CLI (along with CLAUDE.md or "https://lnkd.in/gMvynEgF") to actually implement the code by pointing them to the generated code prompt or JIRA ticket."
""
"Review the code before check-in. Raise the PR. Get reviews from reviewers before proceeding to the next vertical slice."
""
"The idea of implementing multiple independent vertical slices in parallel is exciting. But it also requires a different level of cognitive load management for both yourself and your reviewers. I believe most teams are still not fully prepared for this shift, which means we may not gain much by aggressively parallelizing everything from day one."
""
"Maybe move progressively."
"Maybe handle two vertical slices in parallel for now."
"And then gradually expand as the industry becomes more comfortable with managing this new style of working."
""
"At least to my knowledge, only a few companies like Anthropic or OpenAI currently operate comfortably at this level, where developers, reviewers, and product folks are all aligned around highly parallel AI-assisted execution."
""
"But we do not have to become them from day one at the cost of breaking our own peace of work, development quality, or introducing unacceptable production bugs."
""
"It is completely ok to acknowledge that we will eventually get there through better tooling, processes, and team maturity."
""
"But it does not have to happen all at once."
""
"#modern #software #development #genai #AI #learning