Jacob Jenne

I am extremely impressed by how quickly people who have never coded before can come up with solid, useful local SPA apps when I onboard them to Claude Code. The "Tier 3" of my training program CSC Generation is (previously) non-technical people live-build their first SPA in NextJS, run it locally, and get their own LLM API key. I instituted the pilot 1 month ago. So far just over 5% of the (non-engineering) corporate population at the company has built at least 1 fully functional tool to help them with their workflow, with a success rate of above 80% within 7 days of onboarding. They run the applications on their client, and have an API budget. Many of the apps are quite simple, and hyper-specific to the user. All that their apps have access to, are the same things their client laptops already had. One user, live, in front of the class, in a 1 hour session, finished a novel data normalization app, where the LLM proposed new normalization mappings and presented them to him, as a gamified "Yes/No" using the keys on the keyboard. After the class: added a feature where, once the Accept rate got above significance indicating the LLM could accurately predict at human level, we allows a mass accept. What would have been a labor intensive spreadsheeting exercise on 100's of thousands of items, became a 3 hour project building a game and playing it. Pretty neat example of "Personalized Software"! (PS, please come work with me in a technical or nontechnical capacity if this is the kind of thing you want to do or build enabling tools for)

63

6 Comments

As we’ve consistently emphasized whenever new tech emerges, new tools don't always exist to replace the old ones. Instead, they arise because traditional solutions no longer fit specific scenarios. Our job is to analyze the problem and choose the right tool for the task. While RAG remains a reliable staple for many use cases, GraphRAG is a superior alternative when dealing with complex data relationships—such as internal corporate knowledge management. You could still use traditional RAG for these tasks, but it often introduces unnecessary complexity to achieve the same result. It’s much like using an RDBMS to store data that inherently belongs in a graph database: it’s possible, but you’ll be forced to carry the burden of that extra, avoidable complexity. refer: https://lnkd.in/gMCShctQ

10

Cross-platform auth: Should you build it yourself or plug in a pre-built solution? We break down both paths: what it really takes, where it gets hard, and how studios are speeding up launch without losing control. 👉 Check it out: https://lnkd.in/gzZ8-BY8

8

Part 2 of my ongoing series From Vibe Code to Production-Ready is live! This one dives into Claude, Gemini, and MCPs (Zen + Serena) -- not just as shiny tools, but as a framework for moving from quick prototypes into reliable systems: Command patterns, context, and the practices that make "vibe coding" scale. Part 1 explored why vibe coding eventually breaks down, and Part 2 is the bridge: where experimentation starts becoming engineering. Part 3 will explore agent command patterns: Compacting, Context, and Claude Code Ops Read here: https://lnkd.in/e8-dnaYC #AI #MCP #ClaudeAI #Gemini #AgenticAI #Engineering #ProductDevelopment #VibeCoding #FromPrototypeToProduction #TechLeadership

10