The optional cross-check, and the Co-Scientist demo
The one tool in this course with a single, optional job: showing the hypothesis loop from a different angle.
Google's reasoning LLM (gemini.google.com). The course uses the free tier, plus an instructor-led “Co-Scientist” research-preview demo.
D2·1 Hypothesis — and nowhere else in the pipeline.
Free tier is enough for this course. Co-Scientist demo access is instructor-led, research-preview — not something every participant needs their own account for.
The generate → critique → refine hypothesis loop: 3 falsifiable hypotheses, each checked against identification-skeptic's confound question, refined, and critiqued once more. Co-Scientist demonstrates this same pattern in a purpose-built research-hypothesis tool.
The loop itself doesn't require Gemini — Claude or ChatGPT run it fine. Gemini is here to show the same reasoning pattern from a different product's lens, not because the course depends on it.
A system specifically designed around iterative hypothesis generation and critique, versus a general chat model doing the same steps because you asked it to.
The falsifiability check (“what result would prove this wrong?”) and the confound check are course rules, not features of any one tool — they apply whether you run the loop in Claude, ChatGPT, or Gemini.
Outside D2·1, Gemini has no role in this course — no Day-1 tool-fluency slot, no data-analysis role, no manuscript or referee role.
The course's own compression notes list dropping the optional Co-Scientist demo as the first thing to cut when squeezing toward 12 hours — it's illustrative, not load-bearing.
Free — no seat-budget line item. The only cost is instructor time to run the demo.
Back to D2·1's decision graph on the home page.