Blog

Astronomers report the first convincing direct detection of radio signals from an exoplanet — and no, it is not aliens, Live Science reported Wednesday, covering a new arXiv preprint.

A team led by Kevin Ortiz Ceballos at the Center for Astrophysics | Harvard & Smithsonian used South Africa’s MeerKAT array to observe Beta Pictoris b, a gas giant about 64 light-years away with roughly 10 times Jupiter’s mass. They detected rapid, recurring, highly circularly polarized radio bursts. By comparing radio images with background quasars as fixed reference points, the team localized the emission to the planet rather than its host star — something no prior radio detection had unambiguously done.

The signal matches electron-cyclotron maser emission from auroras. Emission up to about 3.5 GHz implies a magnetic field of at least roughly 1,250 gauss at the source — the first direct exoplanet magnetic-field strength measurement. For comparison, Jupiter’s field is about 4.3 gauss and Earth’s about 0.5 gauss. Oxford astrophysicist Suzanne Aigrain, who was not involved, called it the first truly convincing direct detection and stressed the signals are natural, not technosignatures. The preprint has not yet been peer-reviewed.

Sources: Live Science (Pandora Dewan); arXiv preprint 2609.16720

Chinese President Xi Jinping is in Washington for a three-day state visit — his first to the U.S. since 2015 — with President Donald Trump set to greet him personally at Andrews Air Force Base ahead of a White House summit and state dinner, the BBC reported.

On Wednesday, U.S. Treasury Secretary Scott Bessent said after a second meeting this week with Chinese Vice Premier He Lifeng that Washington is open to extending the U.S.–China trade truce that expires November 10 or examining a larger deal China floated Sunday. “We’re fine with both, and we’re going to do what’s best for the American people,” Bessent told reporters, according to Reuters via The Korea Times. Earlier talks also covered an AI-safety incident “hotline,” tariff cuts on non-strategic goods, and rare-earth compliance — where U.S. officials say Chinese deliveries have “not been up to par.”

Agenda items for Thursday’s summit include trade, AI, and Taiwan. The U.S. trade deficit with China stands at more than $91 billion per latest Census data cited by the BBC. Congress has approved a multibillion-dollar Taiwan arms sale awaiting Trump’s signature; he has held off, treating it as leverage. Experts told the BBC major breakthroughs remain unlikely even as Trump rolls out unusual pomp for an adversary — airport greeting, military flyover, and a guest list that may include Sam Altman, Jensen Huang, and Elon Musk.

Sources: BBC (Brandon Drenon); The Korea Times / Reuters; Reuters

Anthropic says its AI Claude has “autonomously discovered” a new enzyme system similar to the machinery behind the gene-editing tool Crispr — the first result from the company’s newly launched wet lab, The Verge reported Wednesday.

The company said Claude searched a massive DNA-sequence database, with scientists’ involvement limited to the initial prompt and later lab work. Over 21 hours, nearly 950 Claude agents worked through about 210 million tokens before one spotted an unusual repeating pattern and flagged it for human review. Further analysis and lab testing revealed “a previously uncharacterized enzyme system found in bacteriophages,” viruses that infect bacteria.

Anthropic said it is still working to understand what the enzyme system does, but felt it was important to share early to demonstrate Claude’s capabilities and give the broader community insight. Despite comparisons to Crispr, it remains unclear whether the find will have practical applications. The premature announcement comes as Anthropic seeks to attract more scientists and expand into areas like drug discovery ahead of a possible public listing.

Sources: The Verge (Robert Hart)

An FDA medical-device advisory committee meets today to vote on whether the agency should approve the first multi-cancer blood test, NPR reported Wednesday.

Grail’s Galleri claims it can detect up to 50 cancers — many not routinely screened — as an add-on to tools like colonoscopy and mammogram, not a replacement. The test costs $949. No multi-cancer screen has yet won FDA approval for safety and accuracy; FDA has put Galleri on an expedited path, and a Medicare coverage law for FDA-greenlit multi-cancer tests raises the stakes.

Supporting studies enrolled more than 140,000 UK NHS patients and more than 35,000 North American patients age 50 and up. Controversies remain: the test did not meet its primary endpoint of reducing mortality overall, nor of cutting combined Stage III and IV cancers; later NHS updates said Stage IV cancers alone decreased over three years.

City of Hope’s Cristian Tomasetti called the moment a new era while noting study-endpoint technicalities. Abbott Cancer Diagnostics CMO Tom Beer pointed to the roughly 70% of cancer diagnoses and deaths that come from cancers without routine screens, such as pancreatic cancer. FDA may require marketing caveats on false negatives and false positives.

Sources: NPR (Yuki Noguchi)

Comma.ai is facing a federal investigation after five reported crashes involving its aftermarket hands-off driver assistance technology — including two crashes that resulted in three deaths — TechCrunch reported Wednesday.

NHTSA’s Office of Defects Investigation says the crashes involved devices failing to detect or respond to slow or stopped vehicles in the same lane. Across four of the crashes, as many as 11 people were injured, some seriously. Some crashes, including at least one fatal, may have involved forked or modified Comma software.

The company, founded in 2015 by George Hotz, sells devices plus openpilot software that control adaptive cruise control and automated lane centering in a mode similar to Tesla Autopilot or GM Super Cruise, with camera-based driver monitoring and a requirement that drivers be able to retake control.

One fatal case: in February 2026 in Ascension Parish, Louisiana, a 2022 Toyota RAV4 running FrogPilot — a third-party openpilot fork — struck a stopped, fully marked Gonzales Police Department unit with emergency lights on; two rear-seat passengers died, according to Louisiana State Police. Comma’s site warns that ACC performance can be affected by non-moving vehicles in the lane and that abrupt braking is limited. Tesla and Ford have faced similar stopped-object probes.

Sources: TechCrunch (Sean O’Kane)

Top Iranian and U.S. officials held talks at the United Nations in New York on a day President Donald Trump said he could “annihilate” the Islamic Republic if the conflict went unresolved, the BBC reported.

The meeting was mediated by Qatar and brought together Iran’s foreign minister Abbas Araghchi and Trump envoy Steve Witkoff — their first sit-down since the June ceasefire unravelled. No breakthrough was reported.

Witkoff said the sides completed discussions “that we hope will prove constructive and promising,” with mediators who “shuttled between the two sides throughout the day.” He gave a thumbs-up leaving the U.N. and later confirmed the talks on X. Trump told reporters a “very productive” three-hour meeting with the Iranian delegation had taken place on the UNGA sidelines.

Iran’s Baghaei, via IRNA, said Tehran’s conditions include an end to war on all fronts and an end to U.S. “acts of aggression.” Context for the shuttle: U.S. and Israeli air strikes on Iran began Feb. 28; Iran attacked Israel, U.S. bases, and Gulf allies and blocked the Strait of Hormuz — previously carrying about 20% of world oil and LNG — with intermittent fire since the ceasefire collapse and a U.S. blockade of Iranian ports.

Sources: BBC (Bernd Debusmann Jr / Paulin Kola)

A newly described amoeba from California’s Lassen Volcanic National Park can eat and reproduce in water hot enough to kill every other known complex life form, scientists report in Cell, via phys.org on Tuesday.

Named Incendiamoeba cascadensis — roughly “fire amoeba from the Cascades” — the single-celled organism was observed undergoing mitosis at 63°C (145°F). In the lab it grew robustly at 57°C, previously the high-water mark for amoeba growth. When temperatures climb further, it can protect itself up to 70°C (158°F) by changing shape and forming a protective outer layer, then recover after a return to cooler water.

Researchers had long treated about 60°C (140°F) as the upper limit for eukaryotes. The overall heat record still belongs to the archaeon Methanopyrus kandleri at 122°C. The paper’s DOI is 10.1016/j.cell.2026.08.043.

Sources: phys.org (Cell Press); Cell (DOI 10.1016/j.cell.2026.08.043)

OpenAI is expanding its GPT-6 generation with updated smaller Sol and Luna models — landing about 90 minutes after Anthropic’s Opus 5.5 drop — and cutting API pricing for the 6-series pair to half the cost of the 5.6 series, TechCrunch reported Tuesday.

Sol targets complex work like coding; Luna is aimed at high-volume clerical jobs such as summarizing, extracting, and quick questions. OpenAI credits caching and inference improvements for the price cut. On an internal factuality eval based on de-identified conversations where users flagged mistakes, GPT-6 Sol makes about half as many mistakes as its predecessor and reaches Astra-level reliability at lower cost, with a lower coding error rate claimed. The company says the models outshine Anthropic’s Fable and Opus on many tasks.

Sol and Luna are available in ChatGPT Work and Codex for most paid accounts and in the ChatGPT API; Luna is also rolling to the desktop Free and Go tiers, with a gradual ChatGPT app and website rollout through the day.

Sources: TechCrunch (Lucas Ropek)

President Donald Trump used his United Nations General Assembly address Tuesday to call on every nation to enforce the “complete economic isolation of Iran,” pairing oil-price politics with an open threat to “annihilate” Tehran if no deal is reached, The Guardian reported from New York.

“If we stand united, we will soon see a world free of the last 51-year menace of Iranian terror … and oil prices will come plummeting down even lower than they were at the start of the conflict,” Trump said. He thanked the U.N. Security Council for condemning Iranian attacks on Hormuz shipping, claimed he was close to settling the war in Ukraine, and defended the Maduro arrest and Venezuela oil deal with the line “To the victor belong the spoils.”

On technology, Trump said the United States would not rein in AI development, rebranded the race as “superintelligence,” called out-of-control AI warnings a “hoax,” and declared: “I’m not going to stifle growth of something that will be bigger than the Industrial Revolution.” He also announced the U.S. will build two new major military bases in Greenland under an agreement with Denmark and Greenland — with no adversary military presence or sensitive investments allowed without U.S. written approval. A meeting with Ukraine’s Volodymyr Zelenskyy on an energy truce was still expected later on the New York itinerary.

Sources: The Guardian (Andrew Roth)

Cisco Talos researchers say they have identified Windows malware that plots its next moves by polling up to four large language models and taking orders from that hive mind — with no human command channel — WIRED reported Tuesday.

Dubbed CLOSEDQUORUM, the tool checks with DeepSeek, Qwen, Mistral, and Google Gemini to build a consensus on what to do inside a target system. If one AI service is down, it still polls the others, creating redundancy so the system stays closed. Talos linked the sample to cybercriminal forum chatter about credit-card fraud going back to 2025; it is designed to steal login credentials and cryptocurrency. Researchers could not confirm who built it or whether it has been used in real-world attacks.

The find came via CAIRN — Cognitive Artifact Intelligence Research Network — an open-source framework Talos shared Monday to fingerprint and classify AI-integrated malware. Lead researcher Ryan Fetterman said publicly named AI malware families were still scarce a year after Ukraine’s CERT-UA warned about LAMEHUG, but CAIRN has since surfaced about 20 additional examples. “Now what we’re seeing is that it’s becoming operationalized,” said Matt Olney, senior director of threat intelligence at Cisco Talos.

Sources: WIRED