1
Global timekeepers agree to abolish leap seconds from May 2027
Global timekeepers have agreed to stop adding leap seconds to official time, ending a practice that has kept Coordinated Universal Time aligned with Earth's uneven rotation. Unboxholics says the extra seconds have caused malfunctions and outages on online platforms, and that the change simplifies internet infrastructure. Official time may now drift up to an hour from Earth-rotation time before any correction. EarthSky says that gap would likely take centuries to appear, while Unboxholics says centuries or even millennia. Unboxholics also reports the decision accelerated a planned 2035 phase-out, backed by Google, Microsoft, Meta and Amazon.
Key facts
- Leap seconds will stop being added to official time starting May 20, 2027.
- The General Conference on Weights and Measures adopted the resolution on October 15, 2026.
- Unboxholics reports 27 leap seconds were added between 1972 and 2016, the last on December 31, 2016.
- The UK's National Physical Laboratory estimates about a 30% chance of a negative leap second within a decade.
Commentary
Sources: EarthSky, Unboxholics
2
Artemis II results show scope for human science at the Moon
Science News has published a story headlined around the claim that Artemis II results show how much science people can do at the Moon. The headline frames the Moon as a setting for substantial human research, but the specific experiments, results or conclusions behind that framing are unconfirmed in the material available. The story is credited to Lisa Grossman, whose author note says she lives in Minneapolis and whose background spans astronomy and science writing. Nothing in the material confirms the Artemis II findings, figures or mission details that the headline relies on, so the claim remains unverified here.
Key facts
- Science News published the story under the headline 'Artemis II results show how much science people can do at the Moon.'
- Lisa Grossman is the outlet's astronomy writer and is credited on the story.
- Grossman has an astronomy degree from Cornell University.
- Grossman holds a graduate certificate in science writing from University of California, Santa Cruz.
Commentary
Sources: Science News
3
USC researchers find way to create renewable supply of cancer-fighting immune cells
USC researchers report a way to produce a potentially renewable supply of genetically engineered cancer-fighting immune cells, described in a study published in Cell. The team worked with immature granulocyte-monocyte progenitors (GMPs), which normally develop into macrophages, and found they could be expanded extensively in the laboratory while retaining the ability to become functional immune cells. In mice, engineered GMPs established themselves in bone marrow and slowed both blood cancers and solid tumors, with stronger effects when they also carried an immune-activating signal. Qi-Long Ying said the work challenges the view that long-term self-renewal in blood formation belongs mainly to hematopoietic stem cells. The researchers also reported that the cells restored bacterial infection-fighting ability in mice with chronic granulomatous disease. The cancer results remain preclinical.
Key facts
- GMPs can be grown extensively in the laboratory while keeping their ability to become functional immune cells.
- Engineered GMPs carrying a chimeric antigen receptor slowed disease progression in mice with blood cancers and solid tumors.
- Adding an immune-activating signal produced stronger antitumor effects in mice than the receptor alone.
- Stanford University collaborators independently showed GMPs could be maintained for extended periods and genetically engineered.
- Mature macrophage therapies have been limited by rapid clearance, which the progenitor approach may help overcome.
Commentary
Sources: Science Daily
4
Google demonstrates magic state cultivation on its Willow quantum processor
Google Quantum AI says it has turned a proposed error-reduction technique into a working hardware experiment, demonstrating magic state cultivation on its superconducting processor. Magic states are special quantum resources needed for certain operations in fault-tolerant machines, and producing them accurately has traditionally consumed a large share of a computer's resources. Cultivation works like a filter, rejecting flawed attempts until a usable state grows. The experiment ran at small scale and discarded most attempts, and Google acknowledges the approach must still be shown at larger scale. The result is an engineering step, not evidence of a large, useful quantum computer or immediate commercial advantage.
Key facts
- Google Quantum AI demonstrated magic state cultivation on its Willow superconducting processor.
- The peer-reviewed study was published in Science and reports a roughly 40-fold error reduction versus uncultivated states.
- Logical T-states reached a reported fidelity of 99.99%, with about 8% of attempts retained.
- The prepared state was transferred from a distance-3 colour code into a distance-5 surface code.
- A separate October 2026 preprint argues non-destructive postselection can raise resource costs in some regimes.
Commentary
Sources: News9live
5
Brain organoids decode Braille letter signals in step toward biohybrid computing
Bristol University researchers reported that small brain organoids, linked to a tactile sensor, produced neural signals that distinguished between Braille letters, according to a Frontiers in Artificial Intelligence article. Each organoid, containing about 10,000 neurons, was fed electrical patterns converted from sensor data for the 26 Braille letters, and a support vector machine decoded the responses. A single organoid yielded a decodable signal about 61% of the time, while a team of three reached 83%. The researchers do not yet know why groups performed better, though variation between organoids may make their responses complementary. The team said the work tests only signal encoding, decoding and response, not learning or long-term change. The project aims to explore whether energy-efficient biohybrid computing is feasible.
Key facts
- Brain organoids produced letter-specific signals decodable by a support vector machine.
- A single organoid gave decodable signals 61% of the time.
- A team of three organoids gave decodable signals 83% of the time.
- Each organoid is under a millimeter across and contains about 10,000 neurons.
- The team of three outperformed direct decoding of noise-obscured signals.
Commentary
Sources: Frontiers