They are approximately the right levels, although:
- Orchestrate is situationally useful. Quite often you're better off using agents in series rather than in parallel. Most people bragging about parallel agents or orchestrated agent workflows are just bragging.
- Automate is hard to pin down. Not every process requires automation. But anything that is guarded by mechanical (deterministic) process is VERY helpful for keeping agents in line.
- "loop" is not quite that. Loop as it's currently being described in the field is about automating the process so that it remains entirely within the LLMs (which is a pipe dream, btw). They're actually saying that you don't have to write prompts anymore! That's a dead end. However, "loop" in the sense you're describing (feeding results back into the documentation) is valid.
"Automate" and "Loop" are definitely poor names, given that all AI automates and the cheap subscriptions do loop to get things done even with what you call L2.
I wouldn't call the steps on this ladder "proficiency", to my mind they're too close to each other. Someone can go 0-2 in a few minutes if they're comfortable and enthused about reading the basic chat UIs.
3/4/5 are like, if this was a tech tree in a video game, all three would be unlocked at the same time by researching 2.
It doesn’t make sense to me to see that as a ladder. You’re not improving by going up. It’s different techniques and ways to implement, integrate LLMs. They each have their use cases and pros-cons, similar to any other topic in software and system design. The role of an engineer is to understand and manage trade-offs
Simply knowing how to use AI is already quite impressive; you don't need the ability to develop AI. I believe you should quickly apply AI to areas you are familiar with. Soon you'll discover what AI can and cannot do for you. After that, you can find the areas you need to think about and explore in depth, while those should be left to AI.
I don't care what my level is. I fire it up, try to prompt what I need to and get things done. If it doesn't do what I want, I update my prompt accordingly.
Those levels seem incorrect to me. They are written presuming that more AI == right answer, which is not always true.
I think the ladders would be less about how you use it, and more about when you use it. Do you understand when it absolutely should be used, when it absolutely should not, and when it is a grey area? Do you know how to make decisions in that grey area? Do you know how to identify the places where you need a human in the loop and optimize those processes? Do you know how to minimize those processes at the same time? Can you build department-wide systems that integrate institutional knowledge into everyone's usage? Division-wide? Company-wide? Can you create systems that keep that institutional knowledge updated and feed it back to the rest of the department/division/company?
Basically, approach it like any other job levels - it is not about how you do the job you are in, it is about whether or not the impact of what you do improves those around you, and how far that impact is felt.
They are approximately the right levels, although:
- Orchestrate is situationally useful. Quite often you're better off using agents in series rather than in parallel. Most people bragging about parallel agents or orchestrated agent workflows are just bragging.
- Automate is hard to pin down. Not every process requires automation. But anything that is guarded by mechanical (deterministic) process is VERY helpful for keeping agents in line.
- "loop" is not quite that. Loop as it's currently being described in the field is about automating the process so that it remains entirely within the LLMs (which is a pipe dream, btw). They're actually saying that you don't have to write prompts anymore! That's a dead end. However, "loop" in the sense you're describing (feeding results back into the documentation) is valid.
"Automate" and "Loop" are definitely poor names, given that all AI automates and the cheap subscriptions do loop to get things done even with what you call L2.
I wouldn't call the steps on this ladder "proficiency", to my mind they're too close to each other. Someone can go 0-2 in a few minutes if they're comfortable and enthused about reading the basic chat UIs.
3/4/5 are like, if this was a tech tree in a video game, all three would be unlocked at the same time by researching 2.
It doesn’t make sense to me to see that as a ladder. You’re not improving by going up. It’s different techniques and ways to implement, integrate LLMs. They each have their use cases and pros-cons, similar to any other topic in software and system design. The role of an engineer is to understand and manage trade-offs
Simply knowing how to use AI is already quite impressive; you don't need the ability to develop AI. I believe you should quickly apply AI to areas you are familiar with. Soon you'll discover what AI can and cannot do for you. After that, you can find the areas you need to think about and explore in depth, while those should be left to AI.
I don't care what my level is. I fire it up, try to prompt what I need to and get things done. If it doesn't do what I want, I update my prompt accordingly.
Those levels seem incorrect to me. They are written presuming that more AI == right answer, which is not always true.
I think the ladders would be less about how you use it, and more about when you use it. Do you understand when it absolutely should be used, when it absolutely should not, and when it is a grey area? Do you know how to make decisions in that grey area? Do you know how to identify the places where you need a human in the loop and optimize those processes? Do you know how to minimize those processes at the same time? Can you build department-wide systems that integrate institutional knowledge into everyone's usage? Division-wide? Company-wide? Can you create systems that keep that institutional knowledge updated and feed it back to the rest of the department/division/company?
Basically, approach it like any other job levels - it is not about how you do the job you are in, it is about whether or not the impact of what you do improves those around you, and how far that impact is felt.
L9: solving actual problems with A"I", not just increasing the speed of a crappy loop.