Survey the landscape vs. work through it
roadmap.sh's strength is breadth and orientation. It covers many domains, marks what is essential versus optional, and gives you a mental model of a whole field at a glance. When you do not yet know what you do not know, that overview is exactly right, and its community keeps the maps broadly current.
Its limit is that a node like 'transformers' is a label, not a lesson. You leave roadmap.sh and go hunting for the right resource, then for code, then decide if you have done enough to move on. path·ai compresses that: each module already carries a linked resource, runnable code, and a checkpoint, so 'work through transformers' is something you can actually start doing rather than start researching.
Being fair to both
Honesty about trade-offs: roadmap.sh covers far more fields than path·ai's focused ai/ml tracks, has a large community behind it, and is great for high-level planning. path·ai is narrower on purpose and adds the execution layer — ordering at the module level, a resource and code per step, and self-check checkpoints — for free with no account.
They are not really rivals. A common, sensible workflow is to use roadmap.sh to choose a direction and understand the terrain, then use path·ai to actually walk a track step by step. Neither is a course or a certificate, and neither replaces building your own projects, which is still where real skill comes from.
roadmap.sh vs. path·ai
| roadmap.sh | path·ai | |
|---|---|---|
| What it is | Community topic graph | Ordered learning path |
| Breadth | Many fields | Focused ai/ml tracks |
| Resource per step | You find it | Linked |
| Runnable code | Not included | Paired with each module |
| Checking understanding | On you | Checkpoint per module |
| Best for | Surveying the landscape | Working through it |
frequently asked
Is there a better alternative to roadmap.sh for machine learning?
It depends what you need. roadmap.sh is best for surveying the field; path·ai is better for working through it, because each step links a resource and runnable code with a checkpoint. Many people use both.
Does roadmap.sh include the actual learning resources?
It focuses on the topic graph and order, with some links, but you largely source the resources and code yourself. path·ai attaches a resource and runnable code to each module.
Is roadmap.sh enough to learn machine learning?
It's enough to know what to learn and in what order. To actually learn it you still need resources, code practice, and follow-through — which is the layer path·ai adds.
Can I use roadmap.sh and path·ai together?
Yes, and it's a good combination: roadmap.sh to pick a direction and see the terrain, path·ai to walk a track step by step with resources and code attached.
Last updated June 7, 2026