Planning a trip up Mount Shasta? Please don’t ask your favourite chatbot to map out your route as it may be hallucinating. CC-licensed photo by wikiphotographer on Flickr.
You can sign up to receive each day’s Start Up post by email. You’ll need to click a confirmation link, so no spam.
A selection of 9 links for you. Because it seemed to be there. I’m @charlesarthur on Twitter. On Threads: charles_arthur. On Mastodon: https://newsie.social/@charlesarthur. On Bluesky: @charlesarthur.bsky.social. Observations and links welcome.
UN charts retreat in fight to limit climate change, but not giving up • AP News
Seth Borenstein:
»
In a strategic retreat in the fight against climate change, the United Nations has acknowledged that Earth has lost its chance to prevent some of the worst of global warming’s harms. So now the world must figure out how to get temperatures below the danger zone and back to safety, officials say.
Earth is soon going to exceed the safe temperature threshold agreed upon by the world in the 2015 Paris climate accord because of warming from the burning of coal, oil and gas, according to a bleak first-of-its-kind report issued by the United Nations Environment Programme Wednesday.
Instead of giving up, scientists and UN officials recommend trying to limit excessive warming and then bring the temperature back down to the agreed-upon threshold. But it still will mean decades of increased weather-related disasters, rising seas, extinctions and melting ice and glaciers.
“Humanity is speeding past the limit to avoid climate catastrophe,” United Nations Secretary-General António Guterres said.
After a period of over-the-limit warming, the new strategy envisions getting back below the limit by ending fossil fuel use and developing some kind of way to suck industries’ carbon dioxide out of the air.
“It’s a reality check. We have to live with this warmer world,” said report co-author Richard Betts, a climate scientist at the University of Exeter. “But there’s still a lot we can do to limit warming.”
…instead of preventing warming of 1.5 degrees [Centigrade], now the UN said it wants policymakers to “prepare for and navigate a future’’ in which the globe crosses that limit and then corrects course and cools temperatures. The report envisions that global temperatures will pass 1.5 “within the next few years,” follow a hump-shaped path of more warming, peak, then gradually cool to below the 1.5 mark, ideally by the end of the century. The report and experts call it “overshoot.”
“Let’s be very clear, 1.5 target is still the goal. But now we need to approach it differently from above,” said UNEP Executive Director Inger Andersen.
«
“Approach it from above” is not the most encouraging phrase, but it is realistic.
unique link to this extract
Analysis: China’s CO2 emissions fall in Q2 2026 due to plummeting oil use • Carbon Brief
Lauri Myllyvirta:
»
China’s carbon dioxide (CO2) emissions fell by 1% in the second quarter of 2026, as oil consumption plummeted amid the strait of Hormuz crisis.
The country’s use of oil fell by 9% overall and by 16% for transport, after the disruptions to supply from the Gulf through the strait.
China’s total CO2 emissions fell despite a continued rebound in coal-fired power generation.
This is the first time that reductions in oil consumption have been responsible for a fall in CO2 emissions overall. In all previous cases, coal consumption has been the main driver.
Other key findings for the second quarter of 2026 include:
• Electric vehicles (EVs) and public transport have become key factors in China’s oil demand, enabling transportation levels to increase even as fuel use fell sharply
• The effect of EVs on oil consumption was almost twice as large as would be expected based on the increase in the number of EVs on the road alone, as the usage of existing EVs surged
• Oil consumption displaced by EVs in China in the first half of 2026 exceeded the UK’s total oil consumption over a six-month period
• These structural factors are not sufficient to account for the size of the fall in oil consumption, leaving behaviour changes as the other explanation
• “Curtailment” of solar and wind output caused coal power to rise, despite strong hydro output, solar and wind capacity growth, as well as slower demand growth
• Major increases in coal-power capacity and a power market that continues to favour coal limited the amount of coal generation displaced by new wind and solar capacity
• Defying expectations of a boom, annual growth in coal use for chemicals production slowed down to 8%, from 15% in 2025 and 19% in the first quarter.The second quarter of 2026 was a busy time for China’s government planners, with numerous energy-related five-year plan documents being released.
These plans list new measures to address solar and wind curtailment, as well as signalling a higher bar for the approval of new coal-power plants, but add few new quantitative targets.
«
It’s not Net Zero – some way from that – but if emissions really have peaked, that’s encouraging. Thhat one of the world’s most repressive governments is the one doing the most about the climate crisis is ironic. But it’s also the government which is the most technocratic.
unique link to this extract
Inside a conservative operation to hijack your social media feed • The Atlantic
Ali Breland:
»
If you’re short on cash, there is someone in Riga, Latvia, who can help. “You can literally make 100k a month with this campaign,” a young man with blond curls and a thick accent says into a camera. All you have to do is post on social media video clips of “anything that supports Republican.” Unfamiliar with US politics? No problem: “If you’re not from the USA, like, What the fuck is even Republicans? like me,” then simply go to Gemini or ChatGPT, he suggests, “and be like ‘Yo, Gemini, I want some videos promoting Republican values. Give me ideas.’ Go to Gemini. Get ideas. Boom.”
The man, who goes by Serviuos, is promoting a campaign called Elephant Clipping, run by his company, ClipIt. The company pays subcontractors, many from poorer countries in Eastern Europe and the global South, to chop up videos the company provides into bite-size pieces in the hope that the clips will go viral online.
Such “clipping” operations have become a popular means of manipulating people’s attention. Brands, influencers, musical artists, and others spend six-figure sums on such promotional campaigns. Perhaps inevitably, the practice is now moving into politics, often without the transparency required to know exactly what is being served up and by whom. Five months before the midterms, though, Serviuos is up-front about the goal of the Elephant Clipping campaign: “So that republican win in USA,” he writes in one message. I sent Serviuos a list of questions about his work in American right-wing politics. He wrote back, “Seems very out of context,” but didn’t comment further.
…After an anonymous tipster flagged Elephant Clipping to me, I was able to access its Discord server and other related political-clipping campaign chats just by clicking “Join,” noting on a form that I was based in the United States, and answering some questions about my availability and desire to clip. Once I was in, I could see the mechanics of the campaign, which pays huge sums of money to people—many of whom appeared to be outside the US—to push video clips from accounts designed to look like those of ordinary Americans spontaneously touting reasons to favour the GOP.
…Google Drive folders in each of the Serviuos and Wedøe-run campaigns are filled with files uploaded by what appear to be three employees of Enclave & Key, a Las Vegas–based marketing company: Jules Katz, the chief of staff; Benjamin Goodman, a vice president of operations; and Jack O’Hara, the head of media production at an Enclave & Key subsidiary. O’Hara and Katz seem to have uploaded videos to the Google Drive with their company email address. In at least one instance, O’Hara also used what appears to be his personal email address. It’s not specified what role, if any, Enclave & Key plays in the campaigns.
Enclave & Key’s CEO and founder is Blake Wynn, the 26-year-old nephew of the billionaire and conservative mega-donor Steve Wynn.
«
So there’s your answer. (Gift link.)
unique link to this extract
I went to the loneliest baseball game on Apple Vision Pro • The Verge
Mia Sato likes watching baseball, and tried out Apple’s new live coverage using immersive video on the Vision Pro:
»
On Vision Pro, you watch the game primarily from two angles, next to each team’s dugout when that team is batting (you’re stuck there, you don’t get to move around as you please). I liked this angle for certain batters, because you can see the vertical movement of pitches come in without anything blocking them. But for say, a righty batter whose team is watching from the third base dugout (or a lefty whose team is watching from the first base side), their body partially obstructs the view. The immersive view also doesn’t track the action in the outfield — your perspective is frozen next to the dugout. As a result, I felt both in the thick of it and at a strange remove; even ground balls and infield plays feel a bit far away, especially if you’re watching from the third base side and there’s traffic on first and second base.
The immersive broadcast includes a floating jumbotron-style screen that shows a more traditional TV broadcast, along with player stats. I often found my attention drifting up to the standard broadcast, because once the ball leaves the vicinity of the batter, it’s quite hard to track what’s actually going on. The immersive replays from new vantage points like the outfield or closer to first base helped to bridge the feeling of being at a distance. But the Vision Pro’s view isn’t very reactive to what’s happening in real time. When Red Sox second baseman Nick Sogard hit a fly ball early in the game, it almost was a home run — but the immersive view was not very compelling. Baseball is full of weird moments, and it made me wonder how well those would be captured on the Vision Pro.
…On Saturday, I had a game on the TV while I made dinner for a friend, and on Sunday, a group of us caught the last Red Sox-Yankees game, live at Yankee Stadium, where the Yankees crushed the Sox 16-1. It all made me wonder who or what watching a game on Vision Pro is for. It’s remarkable technology that felt weirdly sterile and isolating as I sat on my couch, my friends watching me watch something alone.
All of the most memorable baseball-related moments I’ve experienced happened in the company of others, and none of them depended on how good my view was or how close I was to the players. Immersive viewing might be a fun way for a casual fan or a techy person to engage with the game in a low stakes, convenient way. But if baseball is part of the rhythm of your day, or your summer, or your year, it’s hard to imagine regularly choosing the headset over just about any other way to follow the sport.
«
That last point – about company – feels really important to the success or otherwise of this. The Vision Pro does have ways to do “companionship” on calls, so you’d hope Apple is working on this element of it, after years of missing the point about how to cover sports with it. (Thanks Joe S for the link.)
unique link to this extract
Four major AI models suffer rare overlapping downtime • Ars Technica
Kyle Orland:
»
Cloud-based AI models operated by OpenAI, Anthropic, xAI, and Google suffered a rare and overlapping set of significant service interruptions over a period of hours Thursday morning.
Anthropic first reported a “partial outage” related to “elevated errors on requests to Claude Mythos 5.1, Claude Fable 5.1, and Claude Opus 5” at 9:23 am (all times Eastern). The company reported that it had “identified the cause” of the error roughly 15 minutes later, before reporting that “a fix has been deployed” and the issue was resolved by 12:16 pm. A separate incident report indicated “elevated errors on requests to Claude Sonnet 5” for a brief period just after noon.
OpenAI, meanwhile, reported “elevated errors across ChatGPT and Codex” were resulting in “degraded performance” as of 10:43 am Thursday morning. A mitigation put in place a little more than half an hour later led to the issue being marked as “resolved” by 12:55 pm.
As of this writing, xAI’s Grok currently displays a user-facing error message that the model “is experiencing issues” and that the company is “working on restoring service as quickly as possible.” User-submitted reports from DownDetector regarding Grok shot up from less than 10 just before 9 am to 1,365 by 9:45 am and have fallen to 273 as of this writing.
While Google has not publicly acknowledged any issues with its Gemini model, reports of problems on DownDetector similarly spiked from just 23 around 10:30 am to 412 just after 11 am. Data collected by API checker StatusGator also shows what it terms a “likely outage” for the Gemini API between 10:45 and 11:15 a.m. before the status returned to normal.
«
The joke used to be that whenever Facebook or Twitter had one of their periodic downtimes, productivity around the world would increase dramatically for however long the outage lasted. Would love to have had some flies on the wall at various coding outlets when this happened. Maybe there was a rolling outage as desperate programmers overloaded each model in turn as the previous one went down.
unique link to this extract
Serbians targeted with spyware in country’s “largest documented wave of surveillance” • The Guardian
Aisha Down:
»
At least 14 people from across Serbian civil society were targeted with advanced spyware earlier this year in what the digital rights group Share Foundation said was the largest documented wave of such infection in Serbia to date.
The wave of spyware infections was discovered in August after Apple notified people in 110 countries that they had probably been victims of mercenary spyware.
The Citizen Lab, which carries out forensic analyses of spyware attacks, released an analysis on Wednesday showing that at least one of the Serbians had been targeted with the NSO Group’s Pegasus spyware, which would have given the attacker “total access to the device”.
Share said those targeted included members of a Serbian student movement that has been a thorn in the side of the government of Aleksandar Vučić since 2024. Coalescing out of countrywide protests that followed the collapse of a train station that winter, it is one of the largest student movements in Europe.
Vučić has been in power for more than a decade, during which time critics have said he has overseen a hollowing-out of institutions while steering the country closer to Moscow.
There is no evidence that Vučić’s government was the entity that targeted the student with the Pegasus spyware. However, other spyware used against the students, including a software called NoviSpy, has been previously connected to Serbian authorities.
«
So far so equivocal, you think. Then you come to the next paragraph:
»
In an interview with Euronews Serbia TV on Wednesday, Ana Brnabić, the parliamentary speaker and a ranking member of the ruling Serbian Progressive party, denied that students had been spied on. “Absolutely not,” she said. “I do not believe a single word they’ve [students, Share] said.”
«
Suddenly, the students are eminently believable. The goings-on inside Serbia have largely evaded wider notice, but trouble is brewing there.
unique link to this extract
Climbers rescued from Mount Shasta after relying on AI chatbot to plan trip • SF Chronicle via Yahoo
Gregory Thomas:
»
Three California college students had to be rescued from Mount Shasta on Monday during a poorly planned summit climb in which they overshot their itinerary, meandered off-route and ran out of food and water.
According to authorities with Siskiyou County and the U.S. Forest Service, the novice climbers, from Roseville, said they’d entrusted Google’s AI tool, Gemini, to help them craft key logistics of their trip – from the best route up the mountain to the types of food they should pack and how much water they should carry with them. The county Sheriff’s Office called relying on artificial intelligence tools “a critical misstep.”
The role of AI in wilderness trip-planning has become a heated topic of conversation in California’s north state following a series of instances in which hikers and campers in the Shasta area became lost or confused after relying on chatbots to help them devise their adventures.
Nick Meyers, lead climbing ranger with the U.S. Forest Service and director of the Mount Shasta Avalanche Center, said it’s a new but worrisome problem.
“This is the first year where we’ve started to have some AI fallout, if you will,” Meyers said. “People are using it for trip-planning – ‘Hey, can I do this up here? ‘ It’s not always correct.”
Earlier this summer, rangers with Shasta-Trinity National Forest found a woman camping in an off-limits area on the south side of Shasta clearly marked with “no camping” signage, Meyers said. She told rangers she had asked ChatGPT where she could camp in the forest.
“AI told me I can camp here,” she said, according to Meyers.
Then in August, while hiking in Shasta’s popular Bunny Flat area, a man stopped rangers to ask for directions. He showed them a print-out map he’d been following and explained that he’d prompted ChatGPT to produce it for him.
“It was terrible – terrible – not even close to accurate,” Meyers said. “It had route lines and place names that were totally off – just ridiculous. We were like, dude, you gotta throw that away.”
«
These are very reminiscent of the stories some years ago of people following GPS and driving into lakes, or following map directions that took them through deserts or snowed-in mountain passes to their deaths. Is this just a question of getting used to being sufficiently distrustful of the machine again?
unique link to this extract
Uber and Big Taxi unite in battle against Waymo • Financial Times
Rafe Rosner-Uddin:
»
Uber is aligning itself with drivers’ unions in a bid to slow the rollout of robotaxis, finding unlikely bedfellows as it faces fierce competition on self-driving technology from rivals such as Alphabet-owned Waymo.
The ride-hailing group that became notorious in its early years for skirting regulation and aggressively lobbying against worker protections has recently joined unions in pushing to limit the rollout of self-driving taxis in locations such as New Jersey and Washington DC.
Across the US, Uber has argued for a slower deployment of autonomous vehicles to ease the transition for drivers — and pushed for rules requiring ride-hailing platforms to have “hybrid networks” that combine self-driving cars with human drivers.
Its position marks a remarkable shift for a company once emblematic of how Silicon Valley overlooked worker protections in the name of innovation, but which now finds itself facing technological disruption of its own core business.
In a May policy paper pushing the idea, Uber noted that drivers’ earnings had dropped in Los Angeles and San Francisco where they face competition from Waymo.
«
Of course it’s all definitely for the best reasons and not to slow down Waymo adoption or let it get too much market share.
unique link to this extract
Homey just made the smart home controller I hoped Apple would • The Verge
Jennifer Pattison Tuohy:
»
Smart homes have gotten really good at making things really complicated. Turning on a light can involve multiple steps, compared to just flipping a switch. Homey, the Dutch-based smart home platform owned by LG, thinks it has a solution.
This week it launched the Homey Portal, a touchscreen smart home controller that looks very promising, if very expensive. Preorders open today at an introductory price of $249/€249, with units shipping in December.
A 2.8-inch round touchscreen device with a stainless steel dial, the Homey Portal can sit on a desk or mount to the wall and is powered by a USB-C cable. You tap the icons to control devices like lights, speakers, thermostats, and shades, and you can twist the dial to adjust brightness, volume, and temperature.
The idea of a dedicated smart home control device that can connect to everything in a smart home is compelling. App, voice, and button control have their place, but they’re not as easy as walking up to a switch and just knowing what to do. The problem so far has been interoperability; not everything works with everything. Homey, with its claimed support for over 70,000 devices from over 2,000 brands, plus Matter, Zigbee, Z-Wave, and Thread, is well-positioned to tackle this.
«
The problem: it costs $249 and doesn’t even play music. The ideal smart home controller is just your voice, rather than a device in a specific place, because you can make yourself heard in most places in most houses. And automation ought to simplify that even further. There’s even a toggle called “Weather” which one has to assume isn’t making the weather happen, just telling you about it – which, again, any voice-driven device from Apple, Google or Amazon could do.
unique link to this extract
| • Why do social networks drive us a little mad? • Why does angry content seem to dominate what we see? • How much of a role do algorithms play in affecting what we see and do online? • What can we do about it? • Did Facebook have any inkling of what was coming in Myanmar in 2016? Read Social Warming, my latest book, and find answers – and more. |
Errata, corrigenda and ai no corrida: none notified
The Overspill: when there's more that I want to say 