The one-tool rule
Pick one primary job for the next two weeks. Extra installs create fake progress. Depth beats a dock full of icons.
Match the job
Learn the language of AI → read Learn, then use ChatGPT, Claude.ai, or Gemini.
Run models on your machine → Ollama or LM Studio (see that comparison); optional Open WebUI for a browser UI.
Study from your PDFs → NotebookLM (when uploads are allowed).
Research with links → Perplexity, then Semantic Scholar / Ai2 Asta for papers.
Ship code or UI faster → Cursor, GitHub Copilot, or a generator like v0 / Bolt for frontend drafts.
Northwest open models → Ai2 / Olmo (see that note).
Before you call it experience
Can you explain what the tool did on your last task in two minutes?
Did you verify facts or run the code?
Did you follow data rules for your school or workplace?
If yes, write a short note. That is the portfolio, not the install.
FAQ
Should I learn prompting or Python first?
Do both in tiny loops: one clear prompt practice session, one tiny script or notebook. Neither replaces the other if you care about building software.
Where is the full list?
The tools index is the catalog. Comparisons (this section) are the decision layer on top.