How to decide what to learn next
Let the job market tell you, rather than guessing from a course catalogue.
Learning plans usually come from what looks interesting or what a course provider is promoting. A better source is the set of jobs you actually want.
The method
Collect ten to fifteen postings for roles you would take tomorrow. Paste them into one document. Count how many mention each skill. The distribution is your curriculum, in priority order, specific to your market and your level.
We did this for one engineer across eleven postings. Docker appeared in eight, Python in seven, shipping LLMs to production in seven, Kubernetes in six. That told him his gap was infrastructure rather than AI, which was the opposite of what he assumed.
Separate a gap from a preference
Something that appears in most postings is a gap and worth closing. Something that appears in two is a preference of those employers. Do not build a study plan around a single job you liked the look of.
Pick the one that unlocks the most
Some skills are gates: without them a whole category of role is closed. Others are nice to have. Closing one gate is worth more than improving three things you already do adequately.
Then prove it publicly
A completed course is a claim. A deployed thing using the skill is evidence. Budget the same amount of time for building something small as for the learning itself.
Find roles that actually fit
Upload your CV and AgentFlow AI reads your field, level and skills from it, then matches you against thousands of live postings. The file is parsed in your browser and never uploaded.
Try it