The star that shouldn't be changing, the biology AI can finally make sense of, and the space race nobody warned you about

The Point:

• WOH G64 is changing on a human timescale—which almost never happens with stars: Growing dimmer, seemingly warmer, and ejecting far more dust than before, this red supergiant in the Large Magellanic Cloud is undergoing visible transformations across years rather than millions of years, giving astronomers a rare live window into the final stages of a massive star's life.

• New observations have turned a single-star mystery into a suspected binary system: Simultaneous detection of hot ionized gas and cool molecular gas near WOH G64—conditions that cannot coexist around a single star—points strongly toward a hidden blue companion whose gravity may be stretching the red supergiant's atmosphere, making it more transparent, warmer in appearance, and more prone to dust formation.

• What happens next is genuinely unknown and scientists are watching closely: If the companion star's orbit is elongated and currently near its closest approach, WOH G64 may eventually return to its previous red supergiant state as the blue star recedes—or it may shed its outer layers entirely and reveal something new, making it one of the most closely watched stars in modern astronomy.

The Point:

• AI is giving biology its first tools capable of operating at the scale the subject actually demands: With biological interactions potentially outnumbering stars in the known universe, conventional targeted research can only ever scratch the surface—but AI models like AlphaFold and AlphaGenome are now systematically mapping protein structures and gene variant effects at speeds that compress years of lab work into days.

• The field's defining challenge is moving from correlation to cause and effect: Current AI models excel at identifying statistical patterns but struggle to distinguish genuine causal mechanisms from noise in systems as complex and compensation-prone as biology, and researchers are building hybrid frameworks that combine structured scientific knowledge with massive multi-modal datasets to push past that barrier.

• From burn survivors to cancer to pandemic response, the clinical stakes are concrete and immediate: Labs around the world are already applying causal-aware AI to specific medical problems—mapping cellular identity across 150 million human cells, modeling reindeer antler regeneration as a path to human tissue repair, and integrating omics data with imaging—positioning the field at what researchers describe as the precipice of a genuinely new era in health and medicine.

The Point:

• The satellite industry has quietly become an atmospheric pollution problem: As dead satellites are routinely de-orbited and burned up in the upper atmosphere, they leave behind metallic particles—particularly alumina from aluminium components—that accumulate for years and are already being detected by atmospheric scientists, with the scale about to increase by orders of magnitude if proposed mega-constellations go ahead.

• SpaceX's application to launch one million more satellites would push this into genuinely uncharted territory: Researchers estimate a teragram of alumina could accumulate in the upper atmosphere as a result, with compound effects from rocket launches and re-entries on ozone depletion and atmospheric heating that current science is not yet equipped to fully predict—and satellite builders routinely withhold the material composition data that would allow better modeling.

• The consequences extend from the upper atmosphere to the ground, with no adequate global governance in place: Falling debris already poses a meaningful casualty risk on the ground and to aircraft, orbital collision chains risk triggering the Kessler Syndrome, planned constellations would outshine the stars in naked-eye visibility, and researchers argue that without internationally binding regulation covering the full lifecycle of satellite systems, the atmosphere is being treated as a shared resource with no shared rules.