1. LLMs are closer to religion than they appear. Watch out for those who like it that way

    The article contrasts two starkly different reactions to AI’s influence on faith: Pope Leo XIV’s sweeping Latin encyclical “Magnifica Humanitas,” which condemns AI for risking human dignity, and a multi‑faith study claiming that AIs never provide religious answers. The papal critique highlights a human‑rights and political dimension—suggesting that AI could erode dignity and even be exempted in the workplace on religious grounds—while the study’s methodology and assumptions (treating “religion” largely as fundamentalist Christianity and ignoring scientific consensus) reveal the risk of politicizing AI to advance particular religious narratives. Both cases underscore how AI’s unregulated evolution can echo the schismatic power of religion: it can become a tool of ideological proselytization or a target for ethical and political opposition, showing that the intersection of technology and faith is already a battleground for shaping society’s future.

  2. AI agents get their own phone directory built atop DNS

    DNS‑AID is an open‑source project that lets AI agents discover one another through the existing Domain Name System, eliminating the need for custom registries or hard‑coded ports. Built on DNS features such as SVCB (and TXT fallback) plus optional DNSSEC, DANE TLSA, and JWS signatures, it enables agents to publish and locate services by name, function, or domain via records like _{agent‑name}._{protocol}._agents.{domain}. The Linux Foundation will govern the project, which was originally developed by Infoblox, and is already supported by major DNS providers (AWS, Cloudflare, Azure, Google, etc.). Developers can get started by installing the dns‑aid tool, initializing a zone, publishing an SVCB record, and optionally signing with DNSSEC. By leveraging DNS’s proven global scalability and ubiquity, DNS‑AID offers a “web‑native” routing layer that could underpin a future $3‑$5 trillion agent‑to‑agent economy.

  3. This is the Microsoft Surface Laptop Ultra with Nvidia RTX Spark

    Microsoft and Nvidia have announced the Surface Laptop Ultra, a 15‑inch Windows 11 laptop that will house Nvidia’s new RTX Spark “superchip.” The chip, originally sold in the DGX Spark mini‑PC, offers up to 20 CPU cores, 6,144 GPU cores, and 128 GB of unified memory, with lower‑spec models starting at 16 GB, and is expected to deliver all‑day battery life, RTX 5070‑level graphics, and up to 1 petaflop of AI compute. The Ultra will feature a 15‑inch mini‑LED panel with 262 ppi, 2,000 nits peak HDR brightness, a large haptic trackpad, and a range of ports (USB‑C, USB‑A, HDMI, SD card slot, headphone jack). Microsoft touts it as the most powerful Surface ever, with a weight under 4.5 lb and a price that has yet to be disclosed. The partnership builds on years of collaboration to bring Windows to Arm‑based devices and to optimise the RTX Spark for Windows 11.

  4. Blue Origin Rocket Explodes in Fiery Setback

    Blue Origin’s 98‑meter New Glenn rocket blew up during a hot‑fire (static‑fire) test on Cape Canaveral’s Launch Complex 36, igniting the launchpad at night; founder Jeff Bezos confirmed all personnel were safe and that the company was already investigating the cause. The test, meant to verify engine performance before an actual flight, was the fourth planned New Glenn mission that would have carried 48 satellites for Amazon’s Leo network. NASA’s administrator, Jared Isaacman, acknowledged the anomaly, pledged a thorough investigation, and noted potential, though still‑unknown, impacts on Artemis‑related lunar missions. The incident follows a prior launch failure in April that prompted an FAA probe, highlighting the challenges of developing heavy‑lift launch capability.

  5. An OpenAI model solved a famous math problem that stumped humans for 80 years

    In May, OpenAI revealed that an internal AI model had disproved the long‑standing Erdős unit distance conjecture, a milestone that earned praise from top mathematicians such as Fields Medalist Tim Gowers and University of Toronto professor Daniel Litt; the result marks the first autonomous AI‑generated proof of a major open conjecture, yet it largely repackaged existing ideas rather than inventing new techniques, reinforcing a near‑term future where AI’s encyclopedic recall and brute‑force reasoning complement human insight and creativity in mathematical research.

  6. Best Sleep Trackers of 2026: Oura, Whoop, and Eight Sleep

    Apple’s own smartwatch remains the most affordable option: the Series 11 ($399) can track four sleep stages using its heart‑rate sensor and accelerometer, and newer models add blood‑oxygen and temperature readings, but it offers no overall sleep score. Garmin’s premium Epix Pro ($1,000) delivers advanced sleep monitoring—stages, SpO₂, respiratory rate and restlessness—but tends to over‑estimate sleep duration and ignores naps in its score. Google’s Nest Hub Max ($229) uses radar and a microphone to monitor sleep without a wearable, yet it consistently overestimates REM and misses wake periods, while the cheaper second‑gen model ($100) offers similar performance. Headband contenders like Muse S Athena ($475) provide EEG‑style brain data plus heart‑rate and respiration, yet can be uncomfortable and prone to failure if removed during the night. Withings’ ScanWatch 2 ($370) offers a full four‑stage profile with respiratory‑rate PPG and additional metrics, though its bulk may disturb sleep and it can misclassify quiet wakefulness. The company’s Sleep Tracking Mat ($200) sits under the mattress and gathers motion, breathing, heart‑rate and snoring data, but its assumption that stillness equals sleep can skew scores and it needs a power outlet. Finally, the low‑cost Sleep Routine app ($60 / yr) gives nightly stage reports with reasonable accuracy but is occasionally buggy. Each device trades off price, comfort and data depth, so the best choice depends on whether you prioritize convenience, comprehensive metrics or affordability.

  7. Universal Audio Volt 876 USB Audio Interface Review: Pro-Level Polish

    In 2006, the author’s attempt to jump into IDM music using a new M‑Audio FastTrack Pro was thwarted by driver headaches and high latency, forcing a switch to a Line 6 DL4 pedal and a return to live band playing; had a plug‑and‑play interface like today’s Universal Audio Volt existed, his entry into computer‑based music would have been smoother. Today, audio interfaces have dramatically improved: prices are lower, quality higher, and latency negligible, especially with class‑compliant designs that work out of the box on Windows, macOS, iOS, and Android. As a result, anyone can buy a $200‑class interface, connect a mic, and produce decent recordings for platforms like TikTok without the technical barriers that once plagued early adopters.

Get it every Monday

One edition a week, seven stories, nothing else. Subscribe with RSS — no email address, no tracking.