The progression here is exactly what's been missing from most Claude content. Everything else online is either "here's how to write a prompt" or "here's a 10-minute demo." This is the first time I've seen someone map the actual skill ladder, from file creation reflexes in Tier 1 to encoding your own workflows in Tier 3, with honest time estimates attached.
The Kieran Healy-style note about doing Tier 1 slowly enough to actually learn the reflexes, not just watch Claude do it, is the kind of thing that only comes from having seen people rush past it and stall later. The "do the workflow manually 10 times before automating it" rule in Tier 3 is going to save a lot of people from building systems around habits they don't actually have yet.
We're building a course called "Master Claude in the Real World" that covers a similar progression: Claude chat and prompting fundamentals, Claude Code for agentic workflows, Cowork for desktop automation, and integrations with real tools. The gap between casual Claude users and people running it across their entire stack is exactly what we're trying to close. Just launched on Kickstarter for anyone who wants a structured path through this: https://shorturl.at/ZrG8p
Project 9 is where I'm pointing everyone who asks where to start.
This is the first Claude Code project list that feels real, start local, then APIs, then systems that actually stick. But most people will jump to the shiny Tier 3 stuff and skip the boring parts like backups, permissions, and who pays when the API bill shows up. I’d start with the expense tracker or photo organizer, if it saves you time weekly you’ll keep building. Long term the edge won’t be smartest agent, it’ll be the person with clean data and simple automations that don’t break every Tuesday.
The 'they compound' framing is exactly right, and it's also why most tutorials fail to land.
They jump straight to 'build an agent' without the foundation work. I made this mistake too. My current autonomous setup is effectively your tier 1 + tier 2 + tier 3 stacked - but built in order, each layer teaching me something the next one needed.
The tier 2 to tier 3 transition is where most people stall. State management and context handoff have to be solid before autonomous workflows actually work reliably. The API skills at tier 2 are how you learn that.
Jenny well done. I love the progression here with minute details of the time it may take. You made it seem like it's gonna be a breeze going from the first to the advanced level. Kudos!
Super helpful project ideas, Jenny! I’m already 'stealing' a few to build in my own Claude. Also great to assess our way of approaching these projects without feeling too overwhelmed. Your frameworks are well thought out and you made it look so much easier to try them.
bookmarked, need to revisit this when I have more time to dig in
The progression here is exactly what's been missing from most Claude content. Everything else online is either "here's how to write a prompt" or "here's a 10-minute demo." This is the first time I've seen someone map the actual skill ladder, from file creation reflexes in Tier 1 to encoding your own workflows in Tier 3, with honest time estimates attached.
The Kieran Healy-style note about doing Tier 1 slowly enough to actually learn the reflexes, not just watch Claude do it, is the kind of thing that only comes from having seen people rush past it and stall later. The "do the workflow manually 10 times before automating it" rule in Tier 3 is going to save a lot of people from building systems around habits they don't actually have yet.
We're building a course called "Master Claude in the Real World" that covers a similar progression: Claude chat and prompting fundamentals, Claude Code for agentic workflows, Cowork for desktop automation, and integrations with real tools. The gap between casual Claude users and people running it across their entire stack is exactly what we're trying to close. Just launched on Kickstarter for anyone who wants a structured path through this: https://shorturl.at/ZrG8p
Project 9 is where I'm pointing everyone who asks where to start.
This is super cool and a good guide to understanding level of difficulties. I have yet to play with APIs, so this is a good push for me to do it next!
Really love to hear this! APIs sounds like difficult, but it gets a lot easier after playing with for a few times.
Thanks for sharing it is useful
Project 12 caught my eye. I built exactly that sort of autonomous research pipeline. Three search engines (Firecrawl, Synthetic.new, Exa) orchestrated by a single Claude Code skill, all running in parallel. Fifteen minutes of setup. Wrote up every file you need here: https://reading.sh/how-to-build-a-solid-research-pipeline-in-claude-code-ff7878c5e2b5
Thanks for sharing this, JP. I strongly believe in MCP and eliminating context switching. Your work really resonates with me.
Hi Jennie, nice to see this amazing content from you
Thank you very much really inspired me
You are welcome Michael! This is what I love to see :)
This was so much fun! Thank you for sharing Jenny!!
You are welcome! So glad you enjoyed it 🤗
This is awesome thanks Jenny
I love this! But I love it more how much I - the not technical person - understand !
Loved this. Especially the personal command center 💡
Jenny - the best quality advice for builders of all levels. Thanks for sharing this one 🥳
This is the first Claude Code project list that feels real, start local, then APIs, then systems that actually stick. But most people will jump to the shiny Tier 3 stuff and skip the boring parts like backups, permissions, and who pays when the API bill shows up. I’d start with the expense tracker or photo organizer, if it saves you time weekly you’ll keep building. Long term the edge won’t be smartest agent, it’ll be the person with clean data and simple automations that don’t break every Tuesday.
The progression from simple builds to systems is a solid way to learn.
Thank you John :)
The 'they compound' framing is exactly right, and it's also why most tutorials fail to land.
They jump straight to 'build an agent' without the foundation work. I made this mistake too. My current autonomous setup is effectively your tier 1 + tier 2 + tier 3 stacked - but built in order, each layer teaching me something the next one needed.
The tier 2 to tier 3 transition is where most people stall. State management and context handoff have to be solid before autonomous workflows actually work reliably. The API skills at tier 2 are how you learn that.
How mine evolved overnight: https://thoughts.jock.pl/p/building-ai-agent-night-shifts-ep1
Jenny well done. I love the progression here with minute details of the time it may take. You made it seem like it's gonna be a breeze going from the first to the advanced level. Kudos!
Super helpful project ideas, Jenny! I’m already 'stealing' a few to build in my own Claude. Also great to assess our way of approaching these projects without feeling too overwhelmed. Your frameworks are well thought out and you made it look so much easier to try them.