I don't know how I stumbled on MAGE, but I found the Medium article open in a tab on my phone and I read it. It resonates with me (20+ year software developer, leading the migration to AI usage in my company).
I have a personal project repo I use as a testbed for AI tools/workflows, so I decided to put MAGE to work there and see how it goes. If things go well I'll work with my project team to apply it to our repo, then if things go well there we'll spread it to other projects in the company.
I started a couple of days ago, got through step 4 of the "Quick Start" steps, then decided to pause and read some of the book to get a grounding in what MAGE does and how to make it work. I've read several chapters of the book now. Then I returned to find that the quick start had been removed and replaced with something else that's completely different. (This also left me with a different question -- had I gotten my repo into a weird state that doesn't apply anymore?)
I admit the previous quick start wasn't the easiest to understand (obviously I felt I needed to ground myself in "how to use MAGE" before finishing it). But at least it had some concrete steps to apply. I assume, though I haven't done a side-by-side comparison, that those steps are related to or based on the chapters on applying MAGE to a brownfield project. Frankly a checklist-style migration guide would be very useful -- I'd rather not have to have the book open and derive the steps myself.
The new quick start seems to focus entirely on three skills. That's good -- I don't recall seeing much of anything about the skills in the old one -- but I feel that it's more of a hand-wavy description of what they are. There is some text for each one that sort of resembles a prompt, but it doesn't look much like a prompt I would type -- and in any case one of the points of having a skill is so I don't have to type a long prompt. 🙂
After reading and re-reading it, I still don't grasp where or how I would apply the skills in my day-to-day workflow. In a broader sense, I don't see anything that tells me how (in a practical "do this, do that" sense) MAGE fits into my development activities. I also don't have a good handle on whether MAGE assumes/relies on any specific configuration or setup (the previous quick start talked about updating CLAUDE.md; I think I still some some links to that as I was poking around trying to understand the changes, but I may be wrong). Many AI workflow "frameworks" do (the one I'm most familiar with is Matt Pocock's skills) -- MAGE doesn't seem to be a 1:1 match for those, though it does have some overlap, and because that's what I've seen it was the mental model I was going in with.
I found the "Apply the MAGE method" page to be a good one-pager overview of the method, and the "Start where you are" section seems to pull some things out of the "brownfield" chapter of the book that are a good high-level view of what adoption would look like. But then it ends without taking it to the next "hands-on" level, and points to Part 4 of the book (which I'm more than halfway through without feeling I'm able to actually get started) and Appendix G (which I skimmed). I've also skimmed over Appendix D "Operator's Reference." Both have some good concepts, and even some things that feel like actual application (e.g. questions I should be asking myself about my work) but are still too opaque for me to feel like I have a handle on how to use them.
And again, maybe I'm just misunderstanding -- maybe MAGE is more about the principles, and not so much a cookie-cutter recipe to follow. If so, that's fair, and I will continue reading and studying to figure out how I can try to apply those principles.
Please don't take this long missive the wrong way -- I wouldn't be commenting at all if I didn't see a lot of promise here and genuinely want to try it out. I'm afraid I just need something less academic and more practical. I don't even mind writing it myself -- that's part of what I would/will do if we start adopting it in my work. But of course I need to have a good "model" for myself first 😉.
I don't know how I stumbled on MAGE, but I found the Medium article open in a tab on my phone and I read it. It resonates with me (20+ year software developer, leading the migration to AI usage in my company).
I have a personal project repo I use as a testbed for AI tools/workflows, so I decided to put MAGE to work there and see how it goes. If things go well I'll work with my project team to apply it to our repo, then if things go well there we'll spread it to other projects in the company.
I started a couple of days ago, got through step 4 of the "Quick Start" steps, then decided to pause and read some of the book to get a grounding in what MAGE does and how to make it work. I've read several chapters of the book now. Then I returned to find that the quick start had been removed and replaced with something else that's completely different. (This also left me with a different question -- had I gotten my repo into a weird state that doesn't apply anymore?)
I admit the previous quick start wasn't the easiest to understand (obviously I felt I needed to ground myself in "how to use MAGE" before finishing it). But at least it had some concrete steps to apply. I assume, though I haven't done a side-by-side comparison, that those steps are related to or based on the chapters on applying MAGE to a brownfield project. Frankly a checklist-style migration guide would be very useful -- I'd rather not have to have the book open and derive the steps myself.
The new quick start seems to focus entirely on three skills. That's good -- I don't recall seeing much of anything about the skills in the old one -- but I feel that it's more of a hand-wavy description of what they are. There is some text for each one that sort of resembles a prompt, but it doesn't look much like a prompt I would type -- and in any case one of the points of having a skill is so I don't have to type a long prompt. 🙂
After reading and re-reading it, I still don't grasp where or how I would apply the skills in my day-to-day workflow. In a broader sense, I don't see anything that tells me how (in a practical "do this, do that" sense) MAGE fits into my development activities. I also don't have a good handle on whether MAGE assumes/relies on any specific configuration or setup (the previous quick start talked about updating CLAUDE.md; I think I still some some links to that as I was poking around trying to understand the changes, but I may be wrong). Many AI workflow "frameworks" do (the one I'm most familiar with is Matt Pocock's skills) -- MAGE doesn't seem to be a 1:1 match for those, though it does have some overlap, and because that's what I've seen it was the mental model I was going in with.
I found the "Apply the MAGE method" page to be a good one-pager overview of the method, and the "Start where you are" section seems to pull some things out of the "brownfield" chapter of the book that are a good high-level view of what adoption would look like. But then it ends without taking it to the next "hands-on" level, and points to Part 4 of the book (which I'm more than halfway through without feeling I'm able to actually get started) and Appendix G (which I skimmed). I've also skimmed over Appendix D "Operator's Reference." Both have some good concepts, and even some things that feel like actual application (e.g. questions I should be asking myself about my work) but are still too opaque for me to feel like I have a handle on how to use them.
And again, maybe I'm just misunderstanding -- maybe MAGE is more about the principles, and not so much a cookie-cutter recipe to follow. If so, that's fair, and I will continue reading and studying to figure out how I can try to apply those principles.
Please don't take this long missive the wrong way -- I wouldn't be commenting at all if I didn't see a lot of promise here and genuinely want to try it out. I'm afraid I just need something less academic and more practical. I don't even mind writing it myself -- that's part of what I would/will do if we start adopting it in my work. But of course I need to have a good "model" for myself first 😉.