drop the derived spinning-up notes, full-text the nanda draft

The notes file listed 11 links; 9 are already in the full page cache and the
80k Hours one has its own cache, so it only carried Rocktaschel, now in the
research_taste reading list. The Nanda shared draft was a 902-word excerpt of
a local download; wassname supplied the Google Doc, whose text export gives
the full 11318 words.

refs/research_taste.md

@@ -196,7 +196,7 @@ Most relevant sources cached for this reference:
 - Neel Nanda, research-process sequence: [explore/understand/distill](../docs/evidence/nanda_research_process_explore_understand_distill.md), [key mindsets](../docs/evidence/nanda_research_process_key_mindsets.md), [research taste](../docs/evidence/nanda_research_process_research_taste.md), [shared draft](../docs/evidence/nanda_research_process_shared_draft.md), [paper writing](../docs/evidence/nanda_highly_opinionated_ml_paper_writing.md).
 - Chris Olah, [Research Taste Exercises](../docs/evidence/olah_research_taste_exercises.md): proxy feedback, mentor ratings, research intimacy.
 - Jacob Steinhardt, [Research as a Stochastic Decision Process](../docs/evidence/steinhardt_research_stochastic_decision_process.md): information rate, de-risking, ceilings, baselines.
-- Joshua Achiam / OpenAI Spinning Up, [research source graph](../docs/evidence/spinningup_research_source_graph.md) and [original cache](../docs/evidence/spinningup_researcher.md): RL apprenticeship, fair comparisons, seeds, preregistration, ablations.
+- Joshua Achiam / OpenAI Spinning Up, [cache](../docs/evidence/spinningup_researcher.md): RL apprenticeship, fair comparisons, seeds, preregistration, ablations. The page's own reading list is in that cache; the one item it sends you to that we do not cache is Rocktaschel et al., [Advice for Short-term Machine Learning Research Projects](https://rockt.github.io/2018/08/29/msc-advice.html).
 - Matthew Rahtz, [Lessons Learned Reproducing a Deep RL Paper](../docs/evidence/amid_fish_reproducing_deep_rl.md): confusion, long iteration times, think more before expensive runs.
 - Henderson et al., [Deep Reinforcement Learning that Matters](../docs/evidence/henderson_2018_deep_rl_matters.md): seed variance, implementation differences, reproducibility reporting.
 - John Schulman, [Nuts and Bolts of Deep RL Research](../docs/evidence/joschu_nuts_and_bolts.md): small test problems, health indicators, multiple seeds, ablations.