From Research to Draft:
AI Workflow Patterns
Runners and Chat, alone or together. Mix them however you want.
THE TOOLS
Two Tools, Many Combinations
Runners process new notes automatically as you add them. Set up a context and prompt once, drop stuff on it forever.
Chat is for sit-down sessions. Select notes, ask questions, curate the responses.
You can use either one alone, or chain them together, or alternate between them. Here are a few patterns. Steal what works.
PATTERN 1
Chat Only: The Quick Synthesis
You've been collecting articles on a topic for a week. You need to write something by Friday. Select all your research notes, open Chat, ask "what are the three biggest themes across these?" Chat returns each theme as a separate note. Keep the good ones, discard the rest. Then select those theme notes and ask "draft an outline organized around these themes." You've got a draft in 10 minutes.
Best when: You're writing a one-off and already have your research.
PATTERN 2
Runner Only: The Background Processor
You're tracking a beat - every week, new articles show up on your topic. Set up a Runner with your existing position as context ("here's what I currently believe about X") and a prompt like "what's new here that challenges my view?" Drop each new article on it. The runner fires automatically and builds up a list of challenges over time.
Best when: You're continuously collecting and want processed insights ready whenever you sit down to write.
PATTERN 3
Runner Then Chat: Accumulate, Then Synthesize
Run a Runner over your research as you collect it. Each article gets processed into a summary or key-points note. When you're ready to write, select all the runner outputs plus your original notes and Chat to find the through-line: "pull the common argument across all these."
Best when: You're researching over weeks and don't want to re-read everything before writing.
PATTERN 4
Chat Then Runner: Find the Angle, Then Process
Start with Chat to explore your existing research and find the angle you want to take. Once you've landed on a thesis, set up a Runner with that thesis as context and a prompt like "does this support, contradict, or complicate the thesis?" From that point on, every new article you drop gets evaluated against your argument.
Best when: You need to find your angle first, then sharpen it over time.
MAKE YOUR OWN
These Are Starting Points, Not Rules
These patterns are examples, not prescriptions. Chain runners to other runners. Switch between chat and runners mid-project. Set up five runners with different prompts and see which outputs are most useful. The point is that both tools turn AI output into notes you can keep, curate, and reuse - so you can build workflows that fit how you actually work.