AI Agents at SETI, 51% Weekly AI Use, and Elegoo’s 0.02 mm Scanner
A 2026 Reuters Institute survey finds weekly generative-AI use at 51% across six core countries, while AI agents at the Allen Telescope Array process as much as 2.4 terabits of data per second. Elsewhere, Elegoo’s M1 Ultra claims blue-laser accuracy up to 0.02 mm, and research into animal communication reminds us that not every signal needs a screen.
Generative AI Becomes a Weekly Habit, but Not Yet a News Habit
The Reuters Institute for the Study of Journalism’s 2026 report, based on YouGov fieldwork from August 14 to September 28, finds that 51% of respondents in six countries surveyed across all three report waves used dedicated generative-AI systems at least weekly. That share was 34% the year before. Daily use in those six countries climbed from 7% in 2024 to 16% in 2025 and 28% in 2026. The broader survey covered nine countries: Argentina, Australia, Denmark, Finland, France, Japan, Norway, the United Kingdom, and the United States. This is a useful distinction: the report measures both deliberate use of systems such as ChatGPT and incidental encounters with AI embedded in existing services. About seven in ten respondents said they encountered AI through platforms including search engines, social media, device operating systems, and news media. AI is no longer only an app someone chooses to open; increasingly, it is a feature encountered along the way.
The report’s comparisons matter because adoption is happening through two different routes. Intentional use means a person actively turns to a dedicated system; incidental exposure means AI appears inside a familiar product, often without requiring a separate decision. Weekly use rose in each of the six countries included in all three waves, with reported increases ranging from 13 percentage points in the United States to 20 points in Japan and 21 points in France. Among the three countries included for the first time in 2026, weekly use stood at 51% in Australia, 48% in Finland, and 57% in Norway. The United Kingdom had the lowest weekly use in the survey at 42%. These figures are estimates from online samples, not probability samples with a conventional margin of error. The report cautions that differences of two percentage points or less should generally be treated carefully and that online surveys may under-represent people who are not online.
The news-specific numbers tell a more restrained story than the general-use figures. Across the six countries with comparable data, the share who said they used generative AI to get news rose from 3% in 2024 to 6% in 2025 and 9% in 2026. In the United States, 12% said they had used generative AI for news in the past week. The Reuters Institute notes that this US figure now rivals the online reach of CNN and The New York Times in the comparison it cites, while remaining behind Yahoo! and Fox News among the brands included in the 2026 Digital News Report. The report also finds a trust divide: people who use systems such as ChatGPT, Gemini, Copilot, and Claude tend to be more willing to trust them and more optimistic about AI’s effects than non-users. But public caution about AI in news has changed little over the three years of this research, even as everyday AI use has grown quickly.
For anyone building information products, the practical tension is not simply whether audiences encounter AI; it is whether they deliberately rely on it, and for which tasks. The report says AI use is more common in professional occupations than elementary ones, and that people with paid access are more likely to use advanced tools such as agents. It also stresses that its survey findings depend partly on respondents’ recollection and understanding of terms, and that online opt-in samples have limitations. I find the distinction between routine use and trust especially useful: a person can encounter AI daily without wanting it to select or explain the news. The report itself used ChatGPT 6, Claude Opus 5.5, and Claude Fable 5.1 to assist with data analysis, copy-editing, and text suggestions; the authors say they reviewed the manuscript line by line and that substantial writing remained theirs. That disclosure is a small but concrete example of the questions newsrooms must answer: what work was delegated, what was checked, and who remains accountable for the final account.
At the Allen Telescope Array, AI Agents Help Search the Radio Sky
Scientists at California’s Allen Telescope Array have integrated NVIDIA Holoscan, CyberEther, and Nexus into a system intended to speed the search for possible signals from distant civilizations. According to the source report, AI agents can respond to natural-language instructions to control the telescope, improve incoming data, and search for candidate signals. The observatory’s data challenge is substantial: the array can generate terabits of data every second. The integrated system, running on NVIDIA RTX PRO 6000 Blackwell GPUs, is reported to handle up to 2.4 terabits per second, with processing performed at the observatory. The team has already recorded what the article describes as the first real-time AI detection of a pulsar in the Crab Nebula. A pulsar is the rapidly spinning remnant of a star that emits regular pulses of radiation. That is a scientific milestone in signal detection, not evidence of extraterrestrial life: the reported object is a pulsar, and the SETI search for possible technological signals is the broader application.
The technical design is a chain of data handling and control rather than a single magic model. NVIDIA Holoscan supports processing sensor data directly on GPUs, so the system does not first need to store every incoming stream on hard drives. CyberEther, developed by the SETI Institute, provides signal-processing capabilities and uses NVIDIA CUDA-X tools including cuFFT and cuBLAS. Those libraries support computations used in processing signals, while Nexus is described as a web-based platform connecting AI agents with the telescope and its computing resources. Through natural-language instructions, researchers can monitor data, inspect live spectrograms, and change signal-processing tasks. The system can also help point the telescope, adjust settings, correct data errors, and present results for further examination. The source says these agents can take on tasks that might otherwise occupy scientists for weeks or years; it does not claim that an agent independently establishes that a signal is alien.
This architecture addresses a specific bottleneck: radio observatories collect data at a scale that makes it difficult to examine every stream manually and in real time. GPU processing at the observatory reduces the need to move all data to storage before analysis, while combining telescope control, signal processing, and visualization in a connected workflow. The named institutions and tools are important here: the Allen Telescope Array supplies the observations, the SETI Institute developed CyberEther, and NVIDIA provides Holoscan, CUDA-X, and the GPUs identified in the report. The stated throughput—up to 2.4 terabits per second—describes the system’s processing capacity, not a count of confirmed signals or discoveries. The source also reports testing open models including NVIDIA Nemotron for observatories that may need local processing or may operate without internet access. It presents remote sites, including a possible observatory on the far side of the Moon, as a future use case rather than an existing deployment.
The editorial lesson is that the AI agent’s value here is operational: it brings telescope commands, high-rate data processing, and human inspection closer together. That design can help researchers spend more time reviewing potentially interesting results, while preserving the distinction between a candidate worth investigating and a verified interpretation. Natural-language control also creates a practical need for precise commands and visible system state: when an instruction changes instrument settings or processing tasks, operators need to know what changed and what data was affected. That is analysis of the architecture, not a claim that the report documents a failure or a specific safety incident. The article does not link the pulsar detection to alien technology; it describes a real-time AI-assisted detection of a known class of astronomical object. In SETI, where an unusual signal is a reason to investigate rather than a conclusion, that distinction is as important as the throughput number.
Elegoo M1 Ultra Brings Blue-Laser and Infrared Scanning to a Sub-$2,000 Price
Elegoo, previously known for hobbyist 3D printers, has entered the handheld scanning market with the M1 Ultra, a scanner reviewed by TechRadar as a notably refined first product. It combines blue-laser and infrared structured-light scanning across five main configurations, aimed at objects ranging from small mechanical parts to larger structures and people. The highest stated accuracy is 0.02 mm in the finest blue-laser mode. The scanner weighs just over half a kilogram, includes a three-inch touchscreen, supports wired USB and wireless operation, and uses a swappable battery. Its recommended retail price is $1,699 in the United States, £1,399 in the United Kingdom, and €1,699. TechRadar’s reviewer describes the hardware as solid and the scan quality as exceptional for the price, while noting that the software interface still has traces of hobbyist styling. That mixture makes the product interesting: Elegoo is competing in a market with established makers such as Creality, but is doing so with a feature set that includes both portable operation and high stated precision.
The scanner’s workflow helps explain what the specifications mean in practice. The optical system combines emitters, cameras, and illumination; the user selects among scanning modes suited to different objects, then moves the handheld unit around the target while the software tracks and captures its geometry. TechRadar reports scanning speeds of up to 100 frames per second over USB and 60 frames per second over Wi-Fi, with blue-laser scanning resolution listed from 0.05 to 2 mm. The live preview appears on the touchscreen, which can also adjust exposure and zoom. After capture, processing typically takes a couple of minutes depending on the size and complexity of the scan. The reviewer found that the first preview could look rough, with detail becoming visible after the initial processing stage. Completed models could be exported for 3D printing or imported into Blender for editing. The package includes a calibration board, tripod, batteries, USB cable, and reflective markers, all stored in a hard case.
The review also makes clear that a scanner’s numerical accuracy does not remove the need for preparation. Glossy black plastic and highly reflective surfaces may still require scanning spray, and marker placement affects how easily a target can be captured. In wireless mode, the scanner connects over Wi-Fi and its screen can show a live preview, which the reviewer found helpful when moving around larger objects. There is a trade-off: the scanner takes over the computer’s Wi-Fi connection, preventing access to a normal workshop wireless network during that connection. TechRadar found the early Elegoo Scan software stable during its testing, without freezes or glitches, and described editing and removing unwanted geometry as straightforward. The reviewer also noted dated gradient styling in some controls, including start and pause, exposure, and zoom. This is a specific market proposition rather than a claim that every user will obtain 0.02 mm results on every surface: performance depends on mode, target, and setup.
For makers and small businesses, the combination of five modes, wireless preview, a screen on the device, and sub-millimetre stated resolution may reduce the friction between measurement and a usable digital model. The price places the M1 Ultra below $2,000 in the US list price cited by TechRadar, while its stated 0.02 mm maximum accuracy gives professionals a concrete specification to compare with alternatives. The reviewer’s comparison is primarily with established 3D-scanning manufacturers such as Creality, and the piece frames Elegoo’s improved finish as a shift toward more professional hardware. My caution is the familiar one from measurement science: a very fine specification is not the same as a guarantee of a finished model’s accuracy in every workflow. The review itself emphasizes surface preparation, tracking, and marker placement. Elegoo’s challenge, then, is not only to produce an impressive scanner but to make the software’s decisions and preparation requirements clear enough that users can reliably move from captured geometry to printing or editing.
Chimpanzee Head Bobs: A Gesture Whose Meaning Depends on Context
The Lifesciencesworld article asks why chimpanzees bob their heads and describes the gesture as a non-verbal signal used to communicate intent and shape social interactions, particularly in dominance displays and attempts to initiate interaction. Its central point is that head bobbing is rarely meaningful in isolation. Researchers must read it alongside vocalizations, posture, facial expressions, and the situation in which it occurs. The source says males bob their heads more frequently, especially in contexts involving dominance and competition, while females may use the gesture more subtly during social grooming or when seeking support. Young chimpanzees may use it more often as they learn a troop’s social rules, experiment with communication, or attract attention from mothers and caregivers. These are contextual descriptions, not a single fixed translation such as “yes” or “come here.” The source does not provide a numerical frequency or a count of observed gestures, so there is no responsible way to attach a percentage to the behavior from this material.
The proposed evolutionary explanation is functional: a visible gesture can transmit information quickly, including at a distance, without requiring a vocalization that might attract a predator. The source suggests that such a signal could help animals navigate risky social situations, including challenges to dominant individuals, with less chance of physical harm. The gesture is also described as one component of a wider communication repertoire that includes facial expressions such as the play face and grimace; postures such as bipedal stance or crouching; gestures such as reaching and grooming; and vocalizations including pant-hoots and screams. The article stresses that head movement’s meaning depends on details of the movement as well as context, though the scraped material does not specify a measured taxonomy of movement shapes. It says that evidence supports a combination of innate predispositions and learning: young chimps may have an inclination to bob, then refine what the gesture does through observation and interaction with troop members.
Researchers generally study the behavior by observing chimpanzees in natural habitats or captive settings and recording the frequency, duration, context, and accompanying behaviors of head-bobbing events. The source says computer vision and video analysis are becoming increasingly useful, and identifies future work with more sophisticated technologies and methods as a research direction. It also notes gaps: limited research compares the behavior across chimpanzee subspecies, and variations in social structure or environment might affect its frequency or nuances. It does not report a particular computer-vision model, dataset size, or experimental result, so those details should not be invented. The article mentions the prefrontal cortex, amygdala, and motor cortex as likely contributors to regulation of the behavior: planning and decision-making, emotional processing, and movement control, respectively. It explicitly frames the specific neural mechanisms as still under investigation rather than settled.
The human comparison is suggestive but not a license to equate gestures across species. Human nodding commonly signals acknowledgement or agreement; chimpanzee head bobbing has similarities as a visible signal but may carry different meanings in different social contexts. The source says there is no direct evidence that chimpanzees fully understand the nuances of human head bobbing, although they are skilled at reading body language and may perceive a nod as acknowledgement when it is paired with other positive cues. It also notes that some chimpanzees have been taught human sign language, which has offered insight into symbolic communication. For language researchers, this is a useful reminder that communication is not only the production of a signal but the interpretation of signals in context. A gesture can have a social function without being a word, and the same movement may not mean the same thing across settings, individuals, or species. In that respect, the careful observer’s job resembles good annotation: record the surrounding evidence before assigning a label.
From Neanderthal Cordage to Ocean Voyaging: Rope as Stored Knowledge
Alice Friedemann’s notes on Tim Queeney’s book Rope: How a Bundle of Twisted Fibers Became the Backbone of Civilization make a persuasive case for an invention that is easy to overlook. Friedemann writes, “Rope was far more important than I’d realized to human civilization.” The oldest example described in the material is a cordage specimen from Abri du Maras, dated to 50,000 years ago and identified as more sophisticated than a simple two-strand cord: it is a three-strand rope. Bruce Hardy is named as the researcher who discovered the specimen. The article says researchers argued in a paper in Nature that making cordage required knowledge of plants, seasonality, planning, and retting, as well as an understanding of number and structure. Rope could bind a spearpoint to a shaft, secure a boat, help haul stone, or measure land. Its history is not just a list of useful objects; it is a record of people combining materials, procedures, and shared knowledge to solve practical problems.
The construction process shows why twist matters. Fibers are first twisted into yarn, and yarns are twisted together into strands. For the three-strand rope described, the direction of twist is alternated at each stage: if yarns have an S twist, they are combined into a strand with a Z twist, and the final strands are twisted together in the opposite direction. The source explains that friction and the fibers’ surface irregularities resist sliding. Each strand also wraps around the others in a helix; when pulling force is applied, that helix tightens and helps lock the rope together. The material distinguishes S twist, in which elements are twisted left, from Z twist, which goes right. The ends can unravel unless secured with a knot or with whipping, a lighter line wrapped around the rope. Hand-twisting is laborious, and a mammoth tusk with four holes is discussed as a possible early tool for systematizing production: grass fed through the holes and spun could form a four-strand rope.
The examples span very different forms of engineering. The source describes Egyptian rope-making from halfa, papyrus, date-palm fibers, flax, grasses, animal hide, and animal hair. A Fifth Dynasty tomb at Thebes, dated 2520–2400 BCE, provides the earliest written mention of the process cited in the article; another Thebes tomb from 1500 BCE depicts three men making two-strand rope, with workers twisting strands, closing them together, and controlling the lay with a marlinspike. Rope also served as a measuring tool: Egyptian surveyors were called “rope stretchers,” and the article describes ceremonial rope-stretching to mark building foundations. It connects the same practical principle to Babylon, referring to the 1754 BCE Code of Hammurabi image of a measuring rope. For Stonehenge, the source describes ropes and teams hauling stones on sledges and relays Professor Mike Parker Pearson’s suggestion that as many as 4,000 people may have helped move the large sarsen stones. That estimate is attributed, not presented as a settled headcount.
Rope also made Polynesian voyaging canoes possible. The Hōkūleʻa, launched in 1975 by the Polynesian Voyaging Society of Hawaii as a modern reconstruction, uses the source’s example of a tradition in which coconut-fiber rope lashed hulls, crossbeams, platforms, and rigging without nails or wooden pegs. Coir comes from the coconut’s outer husk; the husks are soaked, the fibers removed and dried, then twisted into rope. The material says coir’s higher lignin content makes it nearly waterproof in salt water. Navigation, meanwhile, depended on accumulated observation of stars, winds, wave trains, island effects on waves, birds, marine life, clouds, and the reported phenomenon te lapa. Rope and wayfinding are different technologies, but both show how technical capability can be distributed across tools and oral knowledge. Friedemann also cautions that her material consists of “incomplete, disjointed notes,” so its examples offer a window into Queeney’s book rather than a complete, continuous history. That is a useful reminder when turning fragments into a story: the evidence is valuable, but its limits belong in the telling.
Where Robles.AI Fits
The source material does not describe Robles.AI’s products, customers, or current services, so I cannot responsibly claim that the company already offers a particular feature or that it could have prevented a problem discussed here. A grounded connection is possible at the level of capabilities that these cases make relevant. For the Allen Telescope Array architecture, the specific capability to evaluate would be an auditable natural-language control layer: recording an instruction, the system action it triggered, the data-processing configuration used, and the human review that followed. Such a service would relate directly to Nexus’s role connecting agents with telescope and computing resources, and to the source’s description of agents changing settings and signal-processing tasks. This is a proposed fit, not a statement about an existing Robles.AI offering or a claim that the telescope system has a reported safety failure.
For the Reuters Institute findings, a relevant capability would be evaluation of information systems: testing whether an assistant distinguishes sourced facts from inference, signals uncertainty, and identifies when a response is being used for news. That connection is specific to the report’s finding that weekly dedicated AI use reached 51% in the six core countries while news use reached 9% in those countries in 2026, and to its observation that public caution about AI in journalism remains. A practical evaluation could examine whether answers preserve the difference between a candidate radio signal and a confirmed interpretation, or between a scanner’s maximum stated accuracy and a guaranteed result on a difficult surface. These are examples of test questions, not reported Robles.AI deployments and not claims that a model can eliminate error.
The Elegoo review points toward another possible technical service: a data-quality and workflow assessment for measurement pipelines. The scanner’s output depends on scanning mode, surface properties, marker placement, tracking, and post-processing. An assessment could therefore test how faithfully software communicates those conditions and how consistently exported geometry behaves in downstream editing or printing. The source does not say that Robles.AI has such a service, nor does it provide evidence that AI is involved in the M1 Ultra’s scanning workflow. Likewise, the chimpanzee and rope articles are useful reminders that labeling and interpretation depend on context: a head bob is not a universal token, and a rope’s strength comes from a particular arrangement of fibers and twists. If Robles.AI works in model evaluation or data annotation, those would be relevant areas to document explicitly rather than infer from the company name.
The responsible editorial position is therefore conditional: Robles.AI could fit these stories through clearly documented evaluation, traceability, or data-quality capabilities if those are services it actually provides. The useful question for readers and clients is concrete: can a system show what input it received, what operation it performed, which assumptions shaped the result, and what a person checked afterward? For a GPU pipeline processing up to 2.4 terabits per second, for survey findings built from online responses, or for a scanner advertising 0.02 mm accuracy in a particular mode, that evidence trail matters more than a broad promise about AI. Until Robles.AI’s service information is supplied, this section identifies relevant capability areas rather than asserting that the company provides them.