I like that this turns Oura API data into charts while dropping proprietary composite scores in favor of metrics the author can inspect. The caveats make the analysis more grounded: the author questions step counts and calorie estimates, and notes that waking heart-rate data only goes back about a month, so I’d read this as personal exploration rather than health evidence.

I trust this lecture’s warning that tuning can change the conclusion: its DP-SGD example says naive settings were almost 100x off, and its repeated-data example says well-tuned regularization changes the qualitative result.

A higher benchmark score does not automatically mean an agent is better, especially when tasks are subjective and runs vary. I’d read this for its three-part approach, pairing human evaluation with offline evaluation and production monitoring, and for the attention to uncertainty and task selection.

I’d start with the threshold-tuning section: a model can separate failures from passes perfectly and still catch none at the default 0.50 threshold. I like that the example chooses thresholds on training folds and measures them on held-out folds.

I like the startup-equity analysis because it follows the “lottery ticket” past the headline valuation to dilution, exercise costs, taxes, and the trouble of selling private shares. The comparison is assumption-heavy, and the compensation figures are dated, so I’d read it for the mechanics of the tradeoff rather than as current salary advice.

Taking the log of the trajectory probability turns the product into a sum, and the environment-transition terms vanish when differentiated with respect to policy parameters. I like this clear account of how REINFORCE learns from sampled trajectories without differentiating through the environment; the connection to variance and trajectory length keeps the cost of that choice in view.