Congrats to the brave adventurers of the 1st cohort for completing my full-day Advanced Claude Code...
I'm grateful to welcome many alumni of my Claude Code foundations workshop returning for the 2nd course.
I’m grateful to welcome many alumni of my Claude Code foundations workshop returning for the 2nd course.
Kelvin, Mona, Damien, Ben, Caleb, Justin, Kah Meng, Kie Chuan, Wai Soon, Andrea, Lynn, Colin, Veronica, Natalie, Terence, chee shen.
There is no “puking blood” setup this time round as we did that in the 1st workshop. But I’m not sure about other symptoms like “brain cramp”.
The main takeaway of this workshop is the advanced knowledge of how to use Claude Code (or similar AI agents like Codex) for big and long-running tasks, from agentic loop to loop engineering.
Big and long-running tasks challenge two structural limitations of LLMs: the context window limit and accumulated quality loss. The context window naturally fills up on a big and long-running task. As it fills up, instruction following starts to degrade. And as the task goes from step to step, quality loss accumulates without an expert-in-the-loop between steps.
We deep dive into how to use context engineering, skills, and workflow patterns to solve these limitations. It is a compression of my practical insights after using Claude Code daily for 11 months, for coding and non-coding tasks.
At the end of the workshop, I showed the class one of my skills that combines every concept: my /plan-implementation skill. It takes a PRD as input, scopes it, then orchestrates multiple specialist subagents (landing page, db, coding, security, ux/ui, graphql, react and more), each running its own reflection loop, to work together on a concrete implementation. I can then pass that to my other skill, /implement-plan, which could run for 5 hours to churn out a production-grade app.
While the ending demo is a software development workflow, the same knowledge can be applied to design workflows that tackle other big and complex jobs.
I’m looking forward to seeing what workflows my alumni will automate in 1 week’s time, as a challenge to win a portable monitor.
#AI #ClaudeCode #VibeCoding #AIAgents #BuildInPublic
If you are already familiar with Claude Code and want to take it to the next level, the next cohort of this workshop is on 24 Jul, 10am to 6pm. It is designed for business professionals who have basic knowledge of using Claude Code.
Details: https://lnkd.in/eR6WPQnC
If you are new to Claude Code and the Terminal, the next cohort of my foundations workshop is on 16 Jul, 1pm to 6pm. It is designed for anyone with zero knowledge of Claude Code and the Terminal, or who wants to get the foundations right.
Details: https://lnkd.in/e9r8kScw
Enjoyed this? Subscribe for more.
Practical insights on AI, growth, and independent learning. No spam.
More in AI Agents
Can LLMs like ChatGPT do reasoning? It failed my casual tests in under 30 minutes.
… more
Range vs Credit Anxiety. Anyone? 🤪
If you have read my previous posts, I have shared how Claude Code performance tanks once your code and best practices hit a certain size due to context bloat...
GenAI Design Thinking Workshop
Helping participants break down their business processes to identify opportunities for adopting agentic AI workflows using the 5I framework.
❌ One common mistake we keep seeing business owners make in marketing:
They hire a junior marketer to run ads, create content and post on social media... And think their marketing is covered.
Your RAG System's Real Problem Isn't Hallucination
Two top common root causes are:
No, Karpathy Didn’t Say Vibe Coding Doesn’t Work
This starkly contrasts with my own experience.
Can LLMs like ChatGPT do reasoning? It failed my casual tests in under 30 minutes.
… more
GenAI Design Thinking Workshop
Helping participants break down their business processes to identify opportunities for adopting agentic AI workflows using the 5I framework.
Your RAG System's Real Problem Isn't Hallucination
Two top common root causes are:
Range vs Credit Anxiety. Anyone? 🤪
If you have read my previous posts, I have shared how Claude Code performance tanks once your code and best practices hit a certain size due to context bloat...
❌ One common mistake we keep seeing business owners make in marketing:
They hire a junior marketer to run ads, create content and post on social media... And think their marketing is covered.
No, Karpathy Didn’t Say Vibe Coding Doesn’t Work
This starkly contrasts with my own experience.