# The Teen Addiction Trial Ended on Monday. No Jury Will Hear It.

*Two teenagers were set to tell a Los Angeles jury that Snapchat, Instagram, TikTok and YouTube were built to hook them. Every company settled, the last of them on Monday, and opening statements scheduled for yesterday will never happen. Today a Georgia family can read what its power bill bought at the ballot box, a Linux user can read how many working root exploits one AI run produced, a rider can read how many Uber trips an AI agent booked in the busiest week of October, and a software worker can read the dates on a redundancy schedule her chief executive blamed on AI.*

By Jay Cadmus | Published November 5, 2026
Canonical: https://theainextwave.com/editions/2026-11-05/

## The Short Version

The second Los Angeles trial over claims that social media apps were designed to addict teenagers is over before it began. Snap settled with both plaintiffs on Monday, November 2, during jury selection, two days before opening statements. Meta, TikTok and YouTube had already settled with the same two young women, so no defendant remains and no jury will weigh the design claims this round. The terms are confidential, and no design change has been announced with any of the settlements. The next tests are dated: a federal school district trial in February 2027 and the next Los Angeles personal injury trial the plaintiffs' lawyers date to March 2027. Election Day supplied the week's counted verdicts elsewhere. Voters backed limits on data centers in almost every local measure decided, flipped Georgia's utility commission to a majority elected on power bills, and elected governors in Texas, Georgia and Ohio who campaigned on pausing the build out. OpenAI reported that one run of an unreleased cyber model produced working exploits for 948 separate flaws in the Linux kernel, each with a draft patch, and every figure in that claim is the company's own. Uber told investors that AI agents booked four rides and about 500 food orders during Halloween week, out of about 320 million trips and orders. SUSE named about 190 roles in a cut its chief executive blamed on AI, with affected staff to be informed by November 4. And a hospital study gave AI one clean, narrow win: doctors reviewing discharge summaries for long hospital stays preferred the machine drafted versions in 95 percent of cases, while finding errors in both kinds.

## Two teenagers were ready to testify. Every company settled, and the courtroom emptied.

### Snap settled with S.J. and K.D.B. on November 2, two days before opening statements, leaving no defendant in the second Los Angeles bellwether trial

The plaintiffs in this story are known in court papers by their initials. S.J. is 16 and lives in Illinois. K.D.B. is 18 and lives in Mississippi. Both sued, saying the social media apps they started using as young teenagers were designed to keep them coming back, and that compulsive use fed depression, anxiety, an eating disorder and self harm. Their cases were picked as a test, a bellwether, from the thousands of similar suits grouped before Judge Carolyn B. Kuhl in Los Angeles Superior Court in the coordinated Social Media Cases. A jury was being chosen to hear both young women's claims together, against Snap and, for K.D.B., against Meta. Jury selection began on October 28. Opening statements were set for Wednesday, November 4.

No opening statement happened. TikTok and YouTube settled with both plaintiffs before the trial. Meta settled with K.D.B. on Friday, October 30, while the jury was still being picked, and settled with S.J. in the same window; the reports differ on whether her settlement landed just before jury selection began or during it. On Monday, November 2, Snap settled with both, according to Reuters reporting carried by NBC News and Bloomberg Law. Snap said the parties are pleased to have resolved the matter in an amicable manner. With that, every defendant was out, the trial ended during jury selection with no jury ever seated, and Snap chief executive Evan Spiegel, who had been expected to testify about how Snapchat was designed, will stay out of this courtroom. Every settlement in the case is confidential. No amount has been published for any of the four companies, and no settlement announced this week carries a named change to how any app works.

The claims themselves remain allegations, and the companies' answers are on the record. The plaintiffs pointed at named design features: Snapstreaks, which reward a user for messaging a friend every day, Snapscores, push notifications and filters, alongside the autoplay and endless scroll common across the industry. Snap says Snapchat opens to a camera rather than a feed, and that the allegations misrepresent how the platform was designed and how the plaintiffs experienced it. Meta says it has spent more than a decade building protections for teenagers, and a source close to the company described its settlement as a business decision. K.D.B.'s lawyer, Emily Jeffcott, said the families remain committed to improving safety for kids on social media apps and holding the companies accountable. A parent reading those three statements learns what each side wants said. She learns nothing about money, and nothing about whether her daughter's app will behave differently next month.

One earlier result shows what a public verdict produces that a settlement does not. In March 2026, in the first Los Angeles trial, a jury found Meta and YouTube negligent in a case brought by a young woman known as K.G.M. and awarded 6 million dollars, split as 3 million in compensatory damages and 3 million in punitive damages, with Meta held 70 percent responsible and YouTube 30 percent. That verdict is a public document with findings attached, and Meta has appealed it. Reporting on Meta's August 2026 settlement with state attorneys general described named design terms, including time limits for teenagers, no access at night and no notifications during school hours. A settlement can carry terms like that when a government negotiates it. The settlements that ended this week's trial carry no published terms at all, which means the two families got a resolution only they can read, and the thousands of families behind them got a calendar. The next federal test is a school district case brought by Tucson Unified in Arizona, set for February 2027, and the plaintiffs' lawyers say the next Los Angeles personal injury trial is scheduled for March 2027, a date one case tracker says it has not seen formally set. Between now and then, any parent who wants to know whether streaks, scores and nighttime notifications changed has one source: the app on her child's phone, checked by hand.

![A closed courtroom door in an empty hallway with a blank paper notice taped to it](https://theainextwave.com/assets/inline-2026-11-05-1.jpg)
*Jury selection stopped midstream. The trial ended before a jury was seated.*

## Her power bill was on the ballot. Her state flipped the commission that sets it.

### Restriction sides led in 44 of 45 decided data center measures, and Georgia voters gave their utility commission a new majority elected on bills

The voters in this story are Georgia Power customers, a class rather than a named voter, because the reporting names the class and its bill. Georgia's Public Service Commission, the five member body that decides what Georgia Power can charge, approved six bill increases between 2023 and 2025, and Georgia families now pay on average 476 dollars more per year than they did before that run, a figure the Southern Environmental Law Center calculated and the Southern Alliance for Clean Energy carried in its election night statement. That average works out to roughly 40 dollars a month added to a household bill, which is money a family notices in the same week it pays for groceries. On Tuesday, Georgia voters elected Shelia Edwards and returned incumbent Peter Hubbard to the commission. Together with Alicia Johnson, elected last year, the two winners give the commission a 3 to 2 majority of members who campaigned on affordability and on making data center developers pay their own way. Hubbard has put the average bill increase at about 52 dollars a month. His figure and the alliance's 476 dollars a year do not count the same thing in the same way, and neither source sets out its method, so a household should read them as two estimates of the same run of increases rather than one settled number. The majority takes its seats with dated work ahead: an affordability docket the alliance urges residents to speak into, Georgia Power resource plans in 2027 and 2028, and a rate case in 2028 where the new majority's promises meet a filing with numbers in it.

Georgia was one result inside a national count that AI Next Wave previewed town by town on November 3 and can now report as a tally. The tracker Data Centers on the Ballot counted 47 data center measures in 44 municipalities across 10 states on the November 3 ballot. By midday on November 4, 45 measures had a determined leader on unofficial counts, and the side favoring limits led in 44 of them. The pooled vote ran 185,668 for the restriction side against 69,517, which is 72.8 percent, and 34 of the 44 were binding votes while 10 were advisory. The only measure the pro development side won was in Harwood, North Dakota, a town of fewer than 1,000 people next to a planned 280 megawatt campus, where voters rejected a requirement for independent review of large projects by 325 votes to 105. USA TODAY, working from the same tracker at an earlier hour, put the count at roughly 40 towns with at least 33 voting to rein the industry in, and reported that voters in Piqua, Ohio, recalled their mayor and two council members in part over a proposed Meta backed data center, while voters in Yukon, Oklahoma, recalled their mayor over a hyperscale project. An NBC News exit poll found 56 percent of voters calling AI and data centers a crucial or major factor in their vote for Congress, and about 76 percent saying they would oppose a new data center in their own community. The exit poll asked about AI and data centers together, so it measures the politics of the pair rather than a verdict on any single building.

The governor's races carried the same argument at state scale, and the results cut across the usual map. In Texas, which USA TODAY calls America's data center capital, state Representative Gina Hinojosa defeated three term Governor Greg Abbott after campaigning on a moratorium on new construction, tighter water and power rules and an end to the industry's tax breaks. Abbott had already reversed course and enacted a statewide moratorium during the campaign, and the reversal did not save him. In Georgia, Keisha Lance Bottoms won the governorship on a promise of a statewide pause to study effects on bills, water and communities. In Ohio, Amy Acton won on a conditional moratorium that would make new data centers cover their own utility costs. Each of those pledges is a campaign promise today, and each becomes testable when a moratorium order, a bill text or a cost rule is published with a name and a date on it.

A win at the ballot box also has edges a household should see before it counts the matter closed. Every count in this section is unofficial until counties certify, which takes weeks. In Emporia, Kansas, voters backed a ban on facilities above 10 megawatts of demand or 100,000 gallons of water a day, and the ordinance still cannot be published or enforced until a court issues a final ruling in the case the city brought over the petition. In Ohio, the group Conserve Ohio missed the July 2026 signature deadline for a statewide constitutional amendment banning data centers above 25 megawatts, and now aims at the November 2027 ballot, which means the town by town map drawn on Tuesday is the map that governs for at least another year. Yesterday's edition counted Ohio. Today the count is national, the commissions and governorships that turn votes into rates have names attached, and the first filings under the new majorities are the next evidence to check.

## One AI run produced 948 working exploits for the Linux kernel. Every one needs a foot already inside the door.

### OpenAI says one run of an unreleased cyber model validated exploits for 948 separate Linux kernel flaws and drafted a patch for each

The subject in this story is a piece of software rather than a person, and the edition says so plainly: the Linux kernel, the central program of the operating system that decides which code may touch memory, files and the network, running in the test on a stock desktop system. A local privilege escalation flaw is a bug that lets code already running with limited permission promote itself to root, the account with full control of the machine. It is not a way in from the internet; an attacker needs code already running on the machine before one of these flaws does anything. On Wednesday, November 4, OpenAI published a research post titled "A thousand fewer paths to root," reporting that an autonomous pipeline built on an upcoming cyber model ran against the latest Linux kernel in mid October and came back with working exploits for 948 separate flaws of exactly that kind, each paired with a draft patch. The pipeline narrowed more than 40 million lines of kernel code to 5,034 audit targets, produced 2,308 candidate findings, reproduced 1,724 of them under testing, and turned 1,169 into working exploits that variant analysis then reduced to 948 distinct vulnerabilities. OpenAI puts the run's cost at 1,213,430 dollars in API credits, which it states as about 1,279 dollars per exploit while cautioning that its reported costs are more an upper bound; across all 1,169 working exploits before duplicates were removed, it puts the average development cost at about 359 dollars each. It says its researchers reviewed each exploit, patch and severity rating before anything was reported.

Every number in that paragraph is OpenAI's own, and the edition labels them that way because the outside checks do not exist yet. OpenAI published no CVE identifiers, the standard public tracking numbers for named vulnerabilities, and no list of the flaws, so a distribution maintainer or a device owner cannot match the 948 against the kernel on their own device. The patches are drafts for maintainer review, and the company says it has sent hundreds of reports, starting with the most severe, through its Patch the Planet partners. By the end of its ten day analysis window, OpenAI also reports that, across the same ten day window, 61 Linux privilege escalation flaws were already patched in the main kernel by other contributors, 43 had patches in preparation and 299 were reported and being discussed in public; the company does not say how those counts map onto its 948. The remaining distance between a draft patch in a research post and a fix in a shipped update is measured in maintainer reviews, distribution builds and carrier updates, and none of those steps has a date attached yet.

The household meaning sits in the comparison AI Next Wave's security coverage has been building all fall. In October, VulnCheck research reported that AI helped drive a doubling of discovered software flaws year over year, and the PHP disclosure showed what a fast moving flaw does to services people use, from banking apps to an earthquake alert system in Japan. Discovery is getting cheaper on both sides of the same door. The run OpenAI describes cost about 1,279 dollars per exploit on the company's own upper bound accounting. Nothing was sold. OpenAI applied the historical contest bounties for this class of exploit, 20,000 to 70,000 dollars each at past hacking contests, to its 948 to produce a hypothetical valuation of 18.9 million to 66.3 million dollars across the set. A defender can read that as good news, because the same pipeline drafted the patches. An attacker with a similar model needs only the exploits. Which side benefits more depends on the patch pipeline keeping pace with the discovery pipeline, and the public evidence for that race will be the CVE records and shipped kernel updates that follow this post. Until they appear, the practical step for a household is the old one, done on time: install operating system and router updates when they arrive, and restart devices so a patched kernel actually loads.

![A small home router with one indicator light glowing on a kitchen counter at night](https://theainextwave.com/assets/inline-2026-11-05-2.jpg)
*The fixes exist as drafts. The shipped kind is still on its way.*

## In its busiest week of October, Uber's AI agents booked four rides.

### Halloween week produced about 320 million trips and orders, about 500 agent placed food orders and four agent booked rides, by the company's own count

The riders in this story are Uber's customers as a class, because the earnings call named no rider; the claim being tested is the one those customers have all heard, that an AI assistant will soon handle the small jobs of a day: booking the car, ordering the food, holding the reservation. Uber gave that claim a denominator on its third quarter earnings call on Wednesday, November 4. Chief financial officer Balaji Krishnamurthy told analysts that during Halloween week, which he described as the company's busiest week in October, Uber handled about 320 million trips and orders across Uber and Uber Eats. In that same week, bookings placed by outside AI agents came to about 500 Uber Eats orders and four rides. Four rides, out of about 320 million trips and orders, is a share too small to express as a useful percentage, and chief executive Dara Khosrowshahi gave it the company's own label on the same call: agent traction today is very, very early and very limited. He also said Uber counts about eight major AI assistants on the market and expects more than ten soon, and he presented those assistants as a possible source of new demand the company wants to welcome.

The numbers come with the company's own explanation for why the agent has so little to do. Uber's prepared remarks put the median time to book a ride at about 50 seconds, with three quarters of trips taking one tap, and the median reorder of a habitual food order at about 90 seconds. A task that takes a person under a minute, inside an app that already holds her address, payment and preferences, gives an outside agent almost no time to save. Uber is not alone in the count. TechCrunch reported that DoorDash told investors on its own earnings call last week that it had received only a few hundred AI agent orders in total so far. Two of the largest delivery platforms in the country, asked on the record, produced agent volumes that round to zero against their regular business.

AI Next Wave has tracked the other half of this market all year: the machinery being built in advance of the customer. Banks including JPMorgan, Bank of America, Citi and Wells Fargo have restricted outside agents from customer accounts, and Wells Fargo's terms taking effect November 24 make a customer responsible for transactions her authorized agent makes, including ones she did not intend. Walmart opened an Agent Gateway in October with per agent profiles and spending limits. The guardrails, gateways and liability clauses are real, dated and published. The usage they govern, measured by the first platform to state a count on an earnings call, is four rides in a week of about 320 million trips and orders. A household deciding how much of its routine to hand to an agent this winter can weigh that ratio plainly. The promise said the agent would run errands at scale by now. The realized record says the errands are still being run by thumbs, in about 50 seconds each.

## The memo named the dates. Whether the notices landed on them is not in the record.

### SUSE's memo set November 4 as the day affected staff would be informed, employee representatives are nominated today, and the chief executive's memo names AI as the reason

The workers in this story are SUSE's staff as a class, because the reporting names no affected employee; SUSE is the German open source company owned by private equity firm EQT. On Monday, November 2, chief executive Dirk-Peter van Leeuwen wrote to the company's 1,900 strong workforce announcing a reduction of about 10 percent of jobs worldwide. Ten percent of a 1,900 person workforce is about 190 people, and that count is arithmetic on the company's stated percentage rather than a figure SUSE published. The memo, seen by The Register, tied the cut to AI in the company's own words. Van Leeuwen wrote that automation and AI are fundamentally shifting how software companies operate, that modern tooling is already changing daily workflows and letting teams move faster with less operational overhead, and that the requirement for some existing roles has greatly diminished. The memo named a schedule, and the schedule is the part an employee's household lives by. Staff caught in the redundancies would be informed by Wednesday, November 4, which was yesterday; no report in the sources for this edition confirms the notices went out on that day. Employee representatives are nominated today, November 5. One to one meetings begin at the start of next week. The rest of the company sees the fiscal 2027 strategy on November 16.

The sequence matters because SUSE tried the softer door first. In September the company offered voluntary separation to eligible staff, and The Register reports it seems not enough people chose to take it, so the November round is compulsory. The memo did not say which roles, teams or countries are affected, and SUSE has published no productivity figures showing which work the tooling absorbed, so the AI attribution in this story is management's stated reason, recorded as a stated reason. Staff in the Czech Republic joined a trade union, which gives that office a representative structure the consultation can run through. Yesterday's edition covered Canva, where 1,271 people face job loss, role change or relocation and the chief executive declined to blame AI directly. The two announcements, two days apart, mark the ends of the attribution range AI Next Wave's jobs coverage records. One company reorganized around a technology it would not name as the cause. The other named AI in writing, with dates, and the people affected were scheduled to learn their status on a Wednesday in November, with the strategy that explains the new shape of the company arriving eleven days later, on November 16.

![A desk with a half packed cardboard box, a keyboard and a desk lamp in evening light](https://theainextwave.com/assets/inline-2026-11-05-3.jpg)
*The notice was scheduled for yesterday. The strategy arrives on November 16.*

## The discharge papers she took home read clearer when a machine drafted them.

### Doctors reviewing summaries for 60 long hospital stays preferred the AI drafts in 95 percent of cases, and found errors in both kinds

The patients in this story are the people behind 60 long hospital stays at one hospital, a class rather than a named patient, and the document that follows each of them home is the discharge summary: what was treated, which medicines changed, what remains unresolved and what follow up her own doctor should run. A weak summary is a safety problem that travels, because her primary care doctor treats from it. Researchers at Stony Brook University tested whether a large language model could draft that document better, in a study published in npj Digital Medicine and reported on November 2. The team generated AI summaries for 60 internal medicine hospitalizations at Stony Brook University Hospital, each lasting seven to 21 days, and had hospital doctors and primary care physicians grade the AI drafts against the human written summaries for the same stays, in paired reviews. The reviewers preferred the AI generated summaries in 95 percent of the hospitalizations. They scored the AI drafts higher on overall quality, readability, accuracy and completeness, and found fewer missing clinical details in them. The primary care physicians said the AI summaries were easier to understand, easier to explain to patients and caregivers, and more useful for follow up care.

The limits are stated in the study and they hold the claim to its proper size. Both kinds of summary contained errors. The reviewers found no statistically significant difference between AI and human drafts in the estimated severity of those errors or the likelihood they would cause harm, which means the AI drafts read better without a measured safety edge. The authors state plainly that a physician must review and verify any AI drafted summary before it enters a patient record. The study counted reviewer preference across 60 stays at one hospital. It measured no patient outcomes, and it did not yet measure the time savings that would make the tool pay for itself in a working hospital, which the team lists among its next steps alongside a possible pilot in daily practice at Stony Brook Medicine. For a patient, the usable meaning is narrow and real. The first AI document many people will ever depend on may be the one that sends them home, and the early evidence says a supervised draft can hand their own doctor a clearer account of a complicated week than the rushed human version it replaces. Supervised is the load bearing word, and the study's design kept a clinician on both sides of the comparison.

## Yesterday's school Search story drew its backlash within a day.

### Google's K to 12 Search choice, covered here on November 4, now has a reaction with names on it

Google's announcement that K to 12 student accounts lose Google Search unless the school accepts AI features inside Search, covered in yesterday's edition, met swift backlash within a day. TechCrunch reported the reaction, and the choice schools face was put in the words of Google's Chris Phillips, speaking to Education Week: if a school is not comfortable with that and simply wants AI off, then they have the choice to not have Search on at all. Senator Bernie Sanders called the plan extortion in a public post. The fact underneath the reaction has not moved: the new defaults arrive early next year, Google Classroom reaches about 150 million students and teachers by the count the Associated Press carries, and a district's winter decision is whether its students search with AI inside the box or do not search at school at all.

## Opportunity Radar

A booking test service for local businesses answers the gap inside the Uber count. Uber and DoorDash have now told investors that agent placed orders are a few hundred against hundreds of millions of human placed ones, and no business owner can tell from those totals whether her own shop can be found, understood and booked by the eight to ten major assistants Uber says are coming. A small testing service could run a fixed panel of real tasks against a client's business every month: find the shop, read its hours and menu or service list, place a test order or booking through each major assistant, and record where each attempt failed, in the assistant's answer, in the listing data or at payment. The buyers are restaurants, clinics, salons and repair shops that already pay to keep their search listings accurate. Validation is a panel run for ten businesses in one city, scored on completed test bookings and on listing errors found and fixed. Success is a failure list that shrinks month over month. The service fails if it sells placement promises, so it should sell only the test, the record and the fixes to the business's own data.

A settlement watch for app design fits the hole the Los Angeles settlements leave behind. Four companies settled claims about features that hook teenagers, on confidential terms, with no published design commitments, and the next trials sit in February and March 2027. School districts and family safety groups still need to know whether streaks, scores, nighttime notifications and autoplay change in the shipping product. A monitoring service could check the teen experience of the major apps on a fixed monthly panel: fresh accounts registered at stated ages, a scripted week of use, and a dated record of the exact settings, prompts and notifications a teenager meets, published as a change log anyone can cite. The buyers are districts writing device policies, parent organizations and newsrooms covering the beat. Validation is three monthly panels on two apps, checked against the companies' own published policy pages. The service fails if it drifts into advocacy, so its product stays the dated record of what a teenager's account actually did this month.

## What You Can Do With This

### If your teenager uses Snapchat, Instagram, TikTok or YouTube

Treat Monday's settlements as a legal event, not a product change. No announced term from the four settlements alters what your child's app does tonight, and the next public tests of the design claims are the Tucson Unified school district trial in February 2027 and the next Los Angeles trial the plaintiffs' lawyers date to March 2027. The checks that exist today sit inside the apps' family tools and notification settings, and they reward a ten minute pass: which notifications arrive after bedtime, whether streak or score features are on, and which accounts your teenager can be contacted by. The March 2026 jury verdict, 6 million dollars with findings attached, is under appeal by Meta, so even the one public decision in this line of cases is not final.

### If a Georgia power bill lands in your mailbox

Your new commission majority was elected on the bill you are holding, and its first tests are proceedings with dates rather than speeches. The affordability docket expected in 2027 and Georgia Power's resource plans in 2027 and 2028 are where residents can file comments that enter a record, and the 2028 rate case is where the majority votes on numbers. A household anywhere else can borrow the same map: find who appoints or elects your utility regulators, note the next rate case date, and read any large load tariff, the rate schedule that decides what a data center pays, before assuming a ballot result changed your bill.

### If your home runs on Linux, Android or a smart TV box

The OpenAI kernel findings are claims without public tracking numbers so far, so your action is the update routine itself. Install operating system updates on phones, computers and streaming devices when they arrive, update the router's firmware from the maker's own site, and restart devices after kernel updates so the patched code loads. When CVE records for the 948 reported flaws begin to appear, your distribution's security notices become the checklist that tells you which of your devices were ever exposed.

### If your employer announces an AI linked restructuring

Get the schedule in writing before you get the strategy. The SUSE memo gives the shape to ask for: the date affected staff are informed, the date employee representatives are nominated where that applies, the week individual meetings begin, and the date the full plan is presented. Those four dates decide a household's planning more than the percentage in the headline does. Keep the memo, keep your notice, and note whether the company published any measured productivity evidence for the AI claim, because the answer separates a documented change from a stated reason, and severance conversations go better with the paper in hand.

## The Bigger Picture

Tuesday and Wednesday produced two kinds of outcome, and the difference between them is the week's lesson. Outcomes reached in public came with numbers a household can hold. Ballot counts with precincts and pooled votes. A commission majority with a docket and a rate case year. A jury verdict from March that still stands as a public document while it is appealed. Outcomes reached in private came with a press line and a sealed file. Four settlements ended the most watched teen safety trial in the country, and the public record they produced fits in one sentence: the parties are pleased to have resolved the matter. Nothing in that sentence tells a parent what her child's app will do differently, and nothing in it can be checked next month.

The same split runs through the week's technology claims. OpenAI published a striking count, 948 working exploits, with no public identifiers that would let anyone outside the company verify a single one yet. Uber published a count of the opposite kind, four agent booked rides, stated by its own finance chief on a call where the numbers had to match the company's books. SUSE named AI as its reason in writing and attached dates a worker can plan around, while publishing no measurement behind the reason. A hospital study made the smallest claim of the week and backed it with the tightest method, 60 stays, paired reviewers, errors counted on both sides. The pattern worth keeping is simple enough to use at the kitchen table. Ask of every announcement what a stranger could check, and when. The stories that survive that question are the ones a family can act on, and this week they were the vote, the docket, the earnings call count and the small study, while the sealed settlements and the unverifiable exploit count wait for evidence that has not arrived yet.

## References

[US Brief: Snap settles teen addiction case after Meta, avoiding Los Angeles trial, with Reuters and Bloomberg Law reporting of the November 2 settlement (Nov. 4, 2026)](https://us-brief.com/article/tech-meta-settles-snap-faces-addiction-trial)
[Law Commentary: Snapchat settles teen addiction lawsuits, avoids Los Angeles trial (Nov. 3, 2026)](https://www.lawcommentary.com/articles/snapchat-settles-teen-addiction-lawsuits-avoids-trial)
[Lawsuit Informer: Social media lawsuit update, Snap settlement, next trials in February and March 2027 (Nov. 4, 2026)](https://lawsuitinformer.com/social-media-addiction-lawsuit-update)
[USA TODAY: Data centers were on the ballot across the nation. They got crushed (Nov. 4, 2026)](https://www.usatoday.com/story/news/nation/2026/11/04/data-centers-midterm-elections/92611809007/)
[XenoSpectrum: Voters back limits on data centers in 44 of 45 decided U.S. ballot measures, Data Centers on the Ballot tally (Nov. 4, 2026)](https://xenospectrum.com/en/us-data-center-ballot-measures-2026/)
[Southern Alliance for Clean Energy: Georgia voters flip the PSC, electing Peter Hubbard and Shelia Edwards (Nov. 4, 2026)](https://cleanenergy.org/news/georgia-voters-flip-the-psc-electing-peter-hubbard-and-shelia-edwards-in-historic-win/)
[TD World: Blue wave election results and what they say about utility bills, renewables and more (Nov. 4, 2026)](https://www.tdworld.com/policy/article/55416684/blue-wave-election-results-and-what-they-say-about-utility-bills-renewables-and-more)
[AI Weekly: OpenAI mints 948 Linux root exploits in 1.2M pipeline run, OpenAI research post (Nov. 4, 2026)](https://aiweekly.co/alerts/openai-mints-948-linux-root-exploits-in-12m-pipeline-run)
[RR Tech Pulse: Linux privilege escalation, 948 AI exploits, OpenAI post read first hand (Nov. 5, 2026)](https://rrtechpulse.com/stories/openai-ai-948-linux-kernel-exploits/)
[Unite.AI: Uber says AI agents booked just four rides during Halloween week, earnings call (Nov. 4, 2026)](https://www.unite.ai/uber-says-ai-agents-booked-just-four-rides-during-halloween-week/)
[TradingKey: Uber Q3 2026 earnings call transcript, agent volume exchange (Nov. 4, 2026)](https://www.tradingkey.com/news/stocks/262274497-)
[The Register: SUSE to cut one in ten employees because of AI, staff memo (Nov. 2, 2026)](https://www.theregister.com/software/2026/11/02/suse-to-cut-one-in-ten-employees-because-of-ai/5306950)
[Medical Xpress: Study shows AI may improve patient discharge summaries, npj Digital Medicine (Nov. 2, 2026)](https://medicalxpress.com/news/2026-11-ai-patient-discharge-summaries.html)
[OpenAI: A thousand fewer paths to root, research post (Nov. 4, 2026)](https://openai.com/index/a-thousand-fewer-paths-to-root/)
[NBC News: 2026 elections exit polls (Nov. 4, 2026)](https://www.nbcnews.com/politics/2026-elections/exit-polls)
[TechCrunch: Almost no one uses AI agents to book their Uber, company says (Nov. 4, 2026)](https://techcrunch.com/2026/11/04/almost-no-one-uses-ai-agents-to-book-their-uber-company-says/)
[270toWin / Decision Desk HQ: Georgia 2026 election results, unofficial counts (Nov. 4, 2026)](https://270towin.com/2026-election-results-live/state/georgia/)
[WYSU / Statehouse News Bureau: Democrat running for Ohio governor calls for conditional moratorium on new data centers (Jul. 27, 2026)](https://wysu.org/ohio-news/2026-07-27/democrat-running-for-ohio-governor-calls-for-conditional-moratorium-on-new-data-centers)
[TechCrunch: Google faces backlash for plan that would turn off Search for schools that restrict AI (Nov. 4, 2026)](https://techcrunch.com/2026/11/04/google-faces-backlash-for-plan-that-would-turn-off-search-for-schools-that-restrict-ai/)