How to make Planck's messiest CMB data play nicely with standard analysis tools?

Planck's low-frequency CMB data gives the tightest measurement of how the universe reionized, but its non-Gaussian shape breaks standard statistical tools. Lee found that log-normal distributions mimic the real likelihood well enough to pass MCMC tests against ACT and Planck combined data, then released a public Python package so others can use these compressed likelihoods without rewriting their entire analysis pipeline.