Sorted by date · America/Guayaquil timezone

Latest signals

Page 25 / 38 · 447 articles

Sep 14, 2026Apple

iOS 27 shows Model Delegation: Siri could route to Claude or GPT-5.6

MacRumors on September 14, 2026 reports that private frameworks in iOS 27 and macOS Golden Gate show Model Delegation: Claude as a Siri extension and an Inference provider in Model Manager Services able to replace Apple’s server-side Siri model with e.g. GPT-5.6 receiving the Siri planner prompt and tool definitions. Finding by sleuth “pdfu.” DMA EU context. Not yet user-facing; ChatGPT extension exists on Golden Gate RC; Claude entitlement not opened to third parties. Caveat: reverse-engineering of private frameworks—Apple has not formally announced this productization.

Sep 14, 2026Microsoft

Microsoft opens public consultation on a humanist Code of Conduct for MAI models

On September 13–14, 2026, Satya Nadella announced a Code of Conduct for first-party MAI models, opened to a six-week public consultation. Driven by Mustafa Suleyman and Humanist Superintelligence, the 37-page draft holds that “people matter more than AI,” rejects model consciousness and legal personhood, requires interruptibility, transparency, and non-manipulation, and contemplates SI only under human control. The MAI family includes Thinking-1, Code-1-Flash, Image-2.5/2.6, Transcribe, and Voice.

Sep 14, 2026MIT

MIT unveils HardFlow: plug-and-play hard constraints for generative models

MIT News on September 14, 2026 describes HardFlow, an algorithm from MIT researchers that plugs in at deployment for pretrained generative models and enforces hard constraints on the final output only—not every intermediate step. They report perfect constraint satisfaction in robotics manipulation, maze navigation, and text-guided image editing, with better solution quality than baselines and faster collision-free robot paths.

Sep 14, 2026Gobernanza

Anthropic, OpenAI and Google advance toward an industry AI safety standards body

On September 14, 2026, CNN and The Information reported that Anthropic, OpenAI and Google have held a working group since July on an industry AI safety standards body modeled on FINRA. Catalyst: Demis Hassabis’s July 14 essay proposing a U.S.-led Frontier AI Standards Body; it predates Amodei’s “pace the frontier” essay. Altman supports a testing/auditing body; Anthropic leans toward government partnership; OpenAI emphasizes voluntary standards plus some state legislation.

Sep 13, 2026Robótica

NEURA and SECO co-design Dragonwing modules for the 4NE1 humanoid in Europe

Around September 7, 2026, NEURA Robotics and SECO announced co-design of Qualcomm Dragonwing compute modules for the 4NE1 humanoid and cognitive cobots; NEURA Gym training hub in Italy; Made-in-Europe Physical AI narrative. Speakers include SECO CEO Max Mauri and NEURA CEO David Reger. NEURA earlier closed a Series C up to ~$1.4B with Tether, Qualcomm, Amazon, NVIDIA, Bosch, and others.

Sep 13, 2026Qualcomm

Qualcomm Dragonwing IQ10: global commercial availability slated for September 2026

Qualcomm’s Dragonwing IQ10 Robotics Reference Design targets up to 700 TOPS sparse / 350 dense, 18 Oryon cores, 64GB LPDDR5x ECC, up to 12 GMSL2 cameras, ROS2, EtherCAT/CAN-FD, and -40–70°C. Unveiled at Computex (June 2026); global commercial availability is scheduled for September 2026 versus NVIDIA Jetson Thor. Pricing was undisclosed as of early September—shipping confirmation verification in progress.

Sep 13, 2026Política

U.S. Senate: bipartisan draft duty of care for frontier AI

Around September 12, 2026, Reuters aides described a bipartisan Senate draft (Thune, Cruz, Klobuchar) with duty of care for frontier developers, mandatory safety testing, incident reporting, and Commerce/DHS power to block catastrophic-risk releases (bio/nuclear called out); likely federal preemption of state AI laws. Context includes Anthropic’s threat report on bioweapons-assist thresholds.

Sep 13, 2026Química

ARCHE: autonomous agent for chemical mechanistic discovery (arXiv:2609.11147)

Preprint arXiv:2609.11147 (Sep 10, 2026) presents ARCHE, an autonomous agentic system for chemical mechanistic discovery combining a reasoning model, a domain chemistry model, and a tool registry. It is validated on asymmetric-catalysis TS, a radical C–I cleavage pathway, and a selectivity descriptor in Ni migratory cross-coupling; code is public.

Sep 13, 2026Biología

ProteinTalks models a virtual cell from more than 38 million measurements

Nature (Sep 9, 2026) presents ProteinTalks, a virtual cell built from more than 38 million temporal protein-abundance measurements on perturbed breast-cancer cell lines. It predicts drug efficacy/synergy, resistance proteins, and organoid/clinical transfer. DOI 10.1038/s41586-026-11001-9.

Sep 13, 2026Ciencia

MatBrain generates 30,000 structures and prioritizes 38 materials in 48 hours

Nature Machine Intelligence (Sep 10, 2026) describes MatBrain, a dual-model agent—analytical Mat-R1 (30B) and tool-executive Mat-T1 (14B)—that generated 30,000 candidate structures and identified 38 promising materials in 48 hours for catalyst design. DOI 10.1038/s42256-026-01298-6.

Sep 13, 2026NVIDIA

NVIDIA’s Nemotron 3 Ultra hits IMO 2026 gold threshold with natural-language proofs

NVIDIA reported Nemotron 3 Ultra post-trained specialists scored 30/42 at the IMO 2026 gold threshold using natural-language proofs only—no formal prover, tools, or internet. Paper arXiv:2609.10712 (~Sep 9) lists Ivan Moshkov, Stephen Ge, George Armstrong, Wei Du, Sadegh Mahdavi, and Igor Gitman; they release checkpoints, data, code, solutions, and Nemotron-IMO-Bench (200 problems).

Sep 13, 2026Google

Google launches Gemini 3.8 Flash and Flash Cyber for trusted defenders

On September 2, 2026, Google introduced Gemini 3.8 Flash and Gemini 3.8 Flash Cyber. Flash holds $0.75/M input and $3.75/M output, same as 3.7. Flash Cyber arrives via the Fairwind Program for trusted defenders (vulnerability detection and automated patching). Authors Tulsee Doshi and Raluca Ada Popa; third Flash release in about six weeks.

Full archive Timeline