Phones Explode on Wall Street as the AI Trade Meets Its Steamroller
Something broke on Thursday, and everyone on Wall Street heard it happen in real time. The phones at banks and hedge funds rang nonstop as investors watched the AI trade, the single most crowded bet in modern markets, come apart at the seams. The Nasdaq 100 fell 6.4 percent, erasing roughly $2.5 trillion in market value, while the Cboe Volatility Index, better known as the VIX, staged its biggest intraday jump in more than 18 months, spiking 11.75 points to 28.36, an eight-month high. "The steamroller showed up today," as one wealth adviser put it, and nobody saw it coming until it was too late.
The trigger was political. Democrats delivered a sweeping midterm victory, and by Thursday morning investors had convinced themselves that a changing of the guard in Washington could stall the pace of AI infrastructure expansion. What made the whiplash so violent was the contrast with the previous session, when markets had rallied on the same election results. aside the perceived threat, piled back into hyperscalers and semiconductors, and assumed the AI trade was bulletproof. They were wrong. "There was massive de-risking into midterms and move and got sucked back in," said John Flood, a partner at Goldman Sachs, whose phones, in his own words, were "exploding with asks for color."
The Hedging Panic
The options market told the story perhaps better than the cash market. Demand for protection exploded. Put premiums across index products ran at roughly two times normal levels as traders rushed to buy downside insurance. The activity was a mix of two camps: those scrambling for fresh protection in semiconductors and AI-linked names, and those cashing out hedges they had already put on, locking in gains before the situation got worse. As Steve Sosnick, chief strategist at Interactive Brokers, put it: "Yesterday there was the perception that no one needed protection and now it's 'Oh my god, we need it.'" That emotional whipsaw, from complacency to terror in 24 hours, is precisely the pattern that tends to mark turning points in crowded trades.
The Machine Amplification
But it wasn't just humans panicking. A large part of the selling was mechanical. Leveraged and inverse ETFs, along with volatility-targeting systematic strategies, rebalance near the market close and tend to react as volatility rises, which creates a feedback loop: markets fall, these funds are forced to sell, which makes markets fall further. According to an intraday note from Nomura, leveraged ETFs were on track to dump roughly $47 billion of stocks into the close. That would rank as the second-largest rebalancing flow in Nomura's data going, and most of it was concentrated in tech and AI-focused funds. In other words, the very structure of the modern market, built around leveraged bets on the same handful of names, turned a political headline into a two-and-a-half-trillion-dollar rout.
Hedge Funds and Retail Both Break
The pain was broad. Goldman Sachs prime brokerage data showed that large multi-strategy hedge funds were down "significantly" thanks to outsized exposure to large AI stocks, with technology, media and telecom-focused funds sitting 2.6 percent lower than a year ago. The real anomaly, though, was retail. On a typical selloff, individual investors buy the dip. Not this time. Vanda Research reported that retail investors were on track for their largest day of selling across single stocks since the COVID crash, with the dip-buying bid effectively vanishing from individual names. When the marginal buyer stops buying and starts selling alongside the forced sellers, that is when drawdowns stop being orderly.
The Question That Matters Now
The only question that matters going forward is whether this was a trade or a regime change. If the AI unwind is merely a trading-type rebalance, an overextended positioning getting flushed out after weeks of euphoria, markets can stabilize and the theme can resume. But if investors are undergoing a lasting re-evaluation of the AI investment theme itself, if the market has decided the capex supercycle no longer justifies its valuation, then this is secular, not cyclical, and that, as Sosnick warned, is problematic. Thursday's steamroller didn't answer that question. It just made sure everyone on Wall Street, from hedge fund desks to retail accounts, is now asking it.
THE $2.5 TRILLION AI REALITY CHECK
Wall Street just experienced its worst day since last year’s tariff turmoil, wiping out an estimated $2.5 trillion in total market value and pushing the Nasdaq composite into correction territory. This historic rout marks a violent reality check for the artificial intelligence trade, exposing the fragility of a top-heavy, tech-dependent market that had been riding a yearslong bull run on the promise of endless AI infrastructure spending.
The selloff was triggered by a perfect storm of political shifts and cracking fundamentals. Just two days after midterm elections swept a bloc of AI-skeptics into power, investors began pricing in a new regulatory reality. This political anxiety was compounded by underwhelming earnings reports from hyperscalers like Amazon and Alphabet, alongside a partial rebound in crude oil prices that reignited inflation fears. The illusion of uninterrupted AI growth has officially met the friction of real-world policy and profitability constraints.
The damage was devastating for the AI supply chain. The so-called picks and shovels plays, including semiconductor and memory giants like Nvidia and Micron, bore the brunt of the sell-off. Data-center developers, cloud-computing providers, and power producers tumbled in unison. Interestingly, the mega-cap hyperscalers like Microsoft and Alphabet sidestepped the worst of the pain. However, Tesla and SpaceX both fell more than nine percent, likely reflecting trader anxiety over Elon Musk’s close ties to the incoming administration and their massive new Texas chip factory ambitions.
This was not just a fundamental repricing; it was a mechanical unwind. As losses mounted from the opening bell, options traders scrambled to cover their positions, sparking a downward spiral. Systematic hedge funds piled in, amplifying the momentum. The most brutal damage was inflicted on highly leveraged bets. Traders who rode triple-leveraged semiconductor ETFs to magnified gains earlier this year saw those positions obliterated, with the Direxion Daily Semiconductor Bull 3X ETF plummeting 39 percent in a single session. By late afternoon, individual investors were on pace for their biggest day of net stock sales since the pandemic.
The immediate focus is now on policy friction, particularly in Texas. Wall Street initially dismissed Governor Greg Abbott’s data-center moratorium as election-year theater. However, with Democratic Governor-elect Gina Hinojosa calling for a continued pause on new approvals until the state legislature can pass real protections, investors are rushing to reassess project timelines. As one portfolio manager noted, if a project thought to be six months from revenue is actually two years away, the financial models break. Earnings forecasts for next year are now highly suspect, and the market must learn to price AI growth with political and regulatory risk firmly in the equation.
OpenAI Drops GPT-6 Cyber: Higher Prices, Unmatched Exploit Capabilities, and the Next Frontier in AI Security
OpenAI has officially launched GPT-6 Cyber, a high-powered, specialized security model designed to uncover software vulnerabilities, construct working exploits, and respect authorized boundaries better than ever before. Engineered as a domain-specific variant of GPT-6 Astra, this new release brings dramatic performance upgrades for enterprise security teams—alongside standard API pricing that significantly outpaces competing models from Anthropic and Google.
The Performance Leap: Kernel Bugs, Exploit Benchmarks, and Real-World Flaws
Under OpenAI’s Preparedness Framework, GPT-6 Cyber earns a Critical rating for cybersecurity capability, up from the High rating of its predecessor, GPT-5.6 Cyber. The model demonstrated a 48.5% score on ExploitGym—a 51% relative improvement over GPT-5.6 Cyber. Beyond controlled benchmarks, the real-world results are staggering: researchers using GPT-6 Cyber reproduced 1,724 Linux kernel vulnerabilities and created working exploits for 948. Furthermore, OpenAI reported discovering 81 previously unknown, high-severity vulnerabilities in Mozilla Firefox, supplying reproducible tests and proposed fixes.
Premium Pricing: How GPT-6 Cyber Compares to Anthropic and Google
Accessing this level of cyber intelligence comes at a steep premium. OpenAI has priced GPT-6 Cyber at $20 per million input tokens, $100 per million output tokens, and $2 per million cached input tokens. This represents a 60% increase in input cost over GPT-5.6 Cyber. Compared to rivals, GPT-6 Cyber’s base rates are double Anthropic’s Claude Mythos 5.1 ($10 input/$50 output), five times Google’s Gemini 4 Argon Cyber Permissive ($4 input/$20 output), and over thirteen times Google’s budget option, Gemini 3.8 Flash Cyber ($1.50 input/$7.50 output).
Gated Access and Mandatory Guardrails: Daybreak Red
To prevent misuse, OpenAI is restricting access via Daybreak Red, its gated authorization program for verified enterprise security testing. Applicants must submit strict documentation outlining specific defensive workflows. The model also features reinforced scope-following capabilities to prevent unauthorized honeypot attacks, achieving a 100% target compliance rate when paired with deployment safety prompts. OpenAI has also expanded its infrastructure with programs like Daybreak Red for Critical Infrastructure and a $1 billion subsidy program for resource-constrained defenders.
The Bottom Line for Security Leaders
As cyber agents evolve from simple code reviewers to autonomous bug hunters, security leaders must balance efficiency with economic trade-offs. While raw token prices put GPT-6 Cyber at the top of the price scale, its capacity to output validated findings, reproduce obscure bugs, and save developer hours could easily justify the spend for high-stakes enterprise defense.
The AI Bubble’s Moment of Truth
Anthropic was supposed to deliver the largest initial public offering in financial history, but a market panic has thrown those plans into chaos. Tech giant Anthropic targeted a towering 1.5 trillion dollar valuation for its debut scheduled right before Thanksgiving. Now, major institutional investors are signaling that the math no longer adds up after a massive selloff wiped 2.5 trillion dollars off the Nasdaq composite in a single day.
Political Shocks and Investor Panic
The sudden cooling of AI enthusiasm was triggered by a double punch of political and market headwinds. Following recent midterm election results that brought a wave of AI-skeptical lawmakers into office, fresh regulatory moratoriums have threatened to halt massive data center infrastructure builds across the United States. Coupled with private market fatigue—where rival OpenAI recently struggled to reach its funding target by securing only 11 billion dollars at a discounted 1.25 trillion dollar valuation—investors who once viewed AI infrastructure as a must-own trade are suddenly demanding lower entry prices.
Inside the Numbers
Despite the high-altitude turbulence, Anthropic's underlying top line continues to show explosive momentum. The firm reached an annualized revenue run rate topping 80 billion dollars at the end of October, up from the previous month. Backed by 146.6 billion dollars in cash on hand as of late September, some prospective buyers hope a successful public offering could leave the enterprise with a fortress balance sheet holding over 200 billion dollars in cash reserves.
What Comes Next for Silicon Valley
Company leadership and financial advisers are holding high-stakes meetings to rethink the IPO strategy. To navigate the souring market sentiment, Anthropic may have to lower its target valuation closer to 1 trillion dollars or reduce its capital target from 100 billion dollars down to around 80 billion dollars. Alternative options include delaying the listing altogether to wait out the public market storm. Because Anthropic’s ecosystem is deeply intertwined with mega-cap tech leaders like Google, Amazon, and Nvidia, how executive leadership handles this public market debut will ripple across the entire tech ecosystem for years to come.
The artificial intelligence build has seemed due for a reset as investors have increasingly questioned the enormous amount of capital being committed to AI infrastructure, especially as interest rates have risen. Today, we saw those worries translate into a sharp selloff to the Nasdaq, the Philadelphia Semiconductor Index, and companies closely associated with AI infrastructure.
I’m not overly surprised. More importantly, I’m not overly concerned. Building data centers requires tremendous amounts of capital, and higher borrowing costs earlier this year raised the hurdle for new projects. And while demand has remained significant, some parts of the semiconductor ecosystem may have extended too far in preparing for long-term demand with near-term purchases of expensive equipment. On top of that, Democratic victories in the midterm elections sparked concerns about potential legislation that could slow data center construction. These are legitimate concerns. But they’re not the end of the AI story. None of this changes our longer-term view that demand for compute will continue to grow and that AI represents a powerful structural investment theme. What may change is the speed at which capacity gets built, and that wouldn’t necessarily be a bad outcome. A more measured pace could ultimately make the AI build healthier and more sustainable. I do not view today’s decline as the bursting of an AI bubble. Instead, I see a reset in one of the market’s strongest and most crowded trades. The reset may continue, but in my view, lower prices and more reasonable valuations would help improve the longer-term opportunity set.
Trump threatens 300% tariffs after Democrats win Congress
President Donald Trump said Thursday that foreign companies that do not build factories in the United States could face tariffs of up to 300%. He made the remarks two days after Democrats won both chambers of Congress. Speaking to reporters at the White House, Trump said that if he scrapped tariffs now, "all of these factories would stop being built." He also said 99% of companies building in the US are doing so to avoid tariffs of "50, 100, 150, 250 percent," The Korea Herald reported. He gave no targets or timeline for the 300% figure, according to Maeil Business. He added that "a lot is being prepared" for the coming weeks.
His comments back up what trade experts have said since the Nov. 3 midterms. Democrats will scrutinize his tariffs, but the president's veto and his remaining trade powers mean they are unlikely to roll back much.
Oversight, not repeal
"The swing in Congress, the results of the election, are very unlikely to have a significant impact on the president's trade agenda," Blake Harden of Ernst & Young's Washington Council told Reuters. He pointed to the broad trade authority Congress has handed to the executive branch. Overriding a veto would take 67 senators and 290 House members, which is well beyond Democrats' projected majorities. Experts expect the new majorities to focus on hearings and investigations instead.
Rep. Richard Neal, expected to lead the House Ways and Means Committee again, said Americans are "tired of paying for Trump's needless war with Iran and price-hiking tariffs." Sen. Ron Wyden introduced a bill in July that would require Congress to approve major tariff measures before they take effect, according to The Dong-A Ilbo. Mireya Solis of the Brookings Institution told the paper that if Congress tries to curb his authority, Trump "will be quick to wield his veto power."
Durable legal footing
In February, the Supreme Court ruled that the International Emergency Economic Powers Act does not authorize tariffs. Since then, the administration has turned to other laws, including Section 301 of the Trade Act and Section 232 of the Trade Expansion Act. Neither requires annual approval from Congress, and both "remain the most legally durable tariff authorities," according to law firm Baker McKenzie. Section 232 duties now reach beyond steel, aluminum and autos to cover pharmaceuticals and drones, South Korea's trade agency KOTRA said, as reported by The Korea Herald.
"If there is one set of congressional bills you would expect Donald Trump to oppose, it would be legislation that would reduce the tariff authorities," James Lindsay of the Council on Foreign Relations told Barron's.
Where relief may come
Analysts say Canada is the most likely place for any easing. Henrietta Treyz of Veda Partners expects the 50% tariffs on about $20 billion of Canadian goods imposed this summer to be lifted, according to Barron's. Meanwhile, an open Section 301 investigation into industrial overcapacity could still bring new duties on China, India, Vietnam and other countries. "I don't think he's going to let his foot off the gas. He's in legacy-building mode right now," former Trump trade adviser Kelly Ann Shaw told Reuters.
THE AI POWER GRAB IS OVER. HERE'S OPENAI'S WHITE FLAG.
Something remarkable happened this week in American politics, and almost nobody in tech saw it coming. On Tuesday, voters across the country delivered a blunt verdict on the AI data center boom, electing candidates at every level of government who promised curbs on these facilities and more oversight of the industry. And just two days later, OpenAI — the most aggressive builder of them all — capitulated with a public pledge it almost certainly hoped it would never have to make.
WHAT OPENAI JUST PROMISED
The company released something called a National Community Compact, and the commitments in it are striking precisely because they address every complaint communities have been raising. OpenAI says it will stop using nondisclosure agreements when negotiating data center deals — the same NDAs that kept residents in the dark about what was coming to their towns. It pledges that local households will not pay for the electricity powering these facilities. It promises clear data on noise, water consumption, and energy use. It will reduce power draw during peak grid hours, send electricity back to the grid, and set local hiring goals. And it will publicly report on whether it actually keeps any of this.
"Households shouldn't bear the costs of powering data centers," the company's policy document states, in language that reads like it was written by someone who just watched an election night map turn against them.
THE ELECTION THAT CHANGED EVERYTHING
Here's the context that makes this so dramatic. Gina Hinojosa just won the Texas governor's office — the first Democrat there in more than three decades — on a platform that included an immediate moratorium on data center development. Democratic gubernatorial victories from Georgia to Ohio to Pennsylvania now threaten stricter scrutiny over roughly half of all proposed AI data center additions in the country. Per Data Center Watch, at least 45 projects worth $68 billion were disrupted by local opposition in just the second quarter of 2026. In dozens of towns that held referendums, exactly one voted in favor of a data center.
OpenAI insists the timing is a coincidence, that the compact had been in the works for months. Maybe. But Chris Lehane, the company's chief global affairs officer, wasn't shy about reading the room. "People are certainly going to be mindful of how voters express themselves," he said. He even acknowledged that some OpenAI projects may now be delayed.
WHY THIS MATTERS
Lehane made a telling observation about the communities OpenAI is building in: many are places "particularly impacted by the globalization of the last 20, 30 years," where residents have, in his words, been "really burned by companies and others in the past." That's an admission that Big Tech walked into these towns carrying the exact baggage of every company that promised jobs, delivered pollution, and left. The compact is a direct attempt to escape that history.
There's a bigger strategic play here, too. OpenAI explicitly says it wants this framework adopted into actual laws governing AI expansion. Translation: if the industry is going to be regulated, better to write the rules yourself — voluntarily, generously, on your own terms — than have hostile state legislatures write them for you after a wave of anti-data-center governors take office.
The AI industry has spent two years treating community opposition as a PR problem to be managed with charm offensives and town halls. This week proved it's a political problem to be won or lost at the ballot box — and the industry is losing. OpenAI's compact is either a genuine turning point or a sophisticated preemption play, and honestly, it might be both. What's not debatable is the trigger: voters spoke, and within 48 hours, the most powerful AI company on Earth changed how it does business.
Democracy, it turns out, still has leverage. Even over compute.
If you found this useful, share it with someone tracking the AI buildout — and subscribe for more on the politics of the compute race.
Perplexity launches Teammates, always-on AI agents for ongoing work
Perplexity launched Teammates on November 5, 2026. Teammates are persistent AI agents that take ownership of ongoing work instead of waiting for prompts one session at a time. Access requires the company's $200-per-month Max plan, so the launch puts Perplexity directly up against OpenAI and Anthropic in the race to build AI coworkers.
According to Perplexity's help center, Teammates is a research preview open to individual Max subscribers on the web. Enterprise Pro and Enterprise Max accounts are not included by default. The company says the feature is still under development and its behavior may change.
Built on Computer
Teammates run inside Perplexity Computer and use all of a user's tools and files there. Each teammate coordinates Computer sessions and works with other teammates to keep projects moving, and it remembers and learns across workstreams such as marketing, sales, support, operations and finance, Perplexity said in its launch post. The agents draw on context users already have in Perplexity, including Projects, Memory, Skills, connectors, files and past sessions. Unlike Computer's temporary sandbox sessions, they persist from one job to the next.
Users give a teammate a responsibility, such as triaging bug reports, tracking metrics or researching a market. The agent then plans the tasks, watches connected tools, acts when something changes and asks the user when it needs input. A teammate can be added to a Slack channel, where coworkers can give it work. Each one runs in its own environment, and users can review its work, take over its desktop or redirect it. Users can also follow each teammate's progress in its own thread in the Perplexity app.
Price and competition
Free and Pro users have to upgrade to Max to use Teammates. Native mobile and desktop apps are coming soon, and Perplexity specifically mentions iOS.
The Verge compared Teammates with OpenAI's Dots and Claude Duet, which also run around the clock and handle a range of tasks. Digital Trends listed a wider field that includes OpenAI's workspace agents, Microsoft's Copilot Cowork and Anthropic's Claude Cowork. It noted that OpenAI and Microsoft aim their products at organizations, while Perplexity is starting with individual subscribers.
Ambitions
CEO Aravind Srinivas called Teammates "persistent coworkers that take over responsibilities and get tasks done end to end". He set out a larger goal in a post on X: "Over time, we envision every org having hundreds of teammates working for each person".
Outside study finds Anthropic AI improving twice as fast
Anthropic is improving its AI models about twice as fast as it did a year ago, according to an independent assessment released Thursday by the nonprofit METR. The finding comes weeks after CEO Dario Amodei urged the industry to slow its pace. The report is one of the most detailed outside looks yet at whether AI is speeding up its own development. It arrives as Anthropic prepares for an initial public offering expected later this month, the Associated Press reported.
What METR found
METR staff ran a "high-access evaluation" at Anthropic from Aug. 12 to Oct. 29. METR's report concludes that as of October, the company is "likely improving the capabilities of their models on well-specified tasks at around twice the rate" it did between March 2024 and October 2025. The group measured progress with Anthropic's internal benchmark index and found scores have grown about 1.9 times faster since October 2025, with a 95% confidence interval of 1.5 to 2.8 times. In other words, Anthropic has made roughly two years of its earlier progress in the past year.
METR said it was "confident" that faster growth in training compute did not cause the speedup. It said most of the current acceleration likely comes from Anthropic's heavy use of AI in its own research. Reuters reported that METR put AI's share of the acceleration at perhaps 80%, while calling that figure uncertain. Researchers also found no convincing evidence of major bottlenecks that could slow progress over the next year.
The report came with caveats. METR said the true rate could be closer to 1.3 or 3 times the historical pace. The speedup might apply only to tasks that can be graded automatically. Anthropic's redactions also limited how much detail the group could publish, AP reported.
Disagreement over Anthropic's own risk report
In a separate review, METR said Anthropic's August risk report "understates the level of acceleration that was occurring." It said Anthropic's internal "2x Acceleration Threshold" had plausibly been crossed by July 15. METR still agreed that the direct catastrophic risks from those models were low. It added that it "cannot rule out a substantial further acceleration" over the next six months.
Amodei's call to slow down
In a Sept. 12 essay, Amodei urged the AI world to "slow the pace at which we improve the capabilities of AI models." He warned that recursive self-improvement "could outrun our ability to understand and control these systems." In the same essay he named METR as a suitable outside evaluator, according to AP. The company has also published its own research showing that AI is already accelerating its development work.
Anthropic did not immediately respond to requests for comment. METR said AI cannot yet make "highly significant and fully autonomous contributions" to model training, but noted that Anthropic researchers described "signs of life" suggesting that could change soon.
Anthropic launches Personal Finance in Claude
Personal Finance lets users connect bank and credit card accounts through Plaid with read-only access to analyze spending trends.
Image: @MacRumors on X
Claude users can ask questions about their finances after connecting bank and credit card accounts through Plaid. The access is read-only, and the tool can analyze spending trends.
Airbnb’s AI Travel Agent Is Coming in 2027 — And It’s More Than a Chatbot
Airbnb is bringing AI into travel planning, promising a personalized agent that can build itineraries, compare stays, and manage bookings.
Airbnb is doubling down on AI. The company says its new AI travel agent will launch in the first half of 2027, with ambitions that go well beyond answering questions in a chat window.
Unveiled in its Q3 shareholder letter, the agent will use a custom-built interface that adapts to what travelers need in real time. Think interactive maps, photos, side-by-side comparisons, calendars, and itineraries — generated on demand rather than buried in a text-heavy conversation.
CEO Brian Chesky says travel planning is different from everyday AI tasks. People want to explore, browse photos, compare properties, and make decisions together. In fact, Airbnb says 80% of travelers enjoy researching their trips. The goal, then, isn’t to automate the fun out of travel planning, but to make it smarter and more personal.
The agent will also work beyond the booking stage. Travelers will be able to use it to change reservations, resolve payment issues, request refunds, rebook stays, and access 24/7 safety support — capabilities that generic AI agents cannot easily replicate.
Eventually, Airbnb wants the experience to cover more than accommodation, extending into hotels, local services, and travel experiences.
The announcement comes as Airbnb reports a strong Q3, with revenue climbing 19% year over year to $4.87 billion. The company also raised its 2026 revenue growth outlook to the high teens.
The bigger bet: Airbnb isn’t just adding an AI chatbot to its app. It wants to build a dedicated AI-powered travel experience — one that helps people discover, plan, book, and manage an entire trip.
No soup – or in this case,chili— for you: Wendy's on Thursdayrebuffeda $1.4 billiontakeover bidspearheaded by Nelson Peltz's Trian Fund Management. The burger chain's board reaffirmed its confidence in CEOBob Wright, whotook the helmin May, and announced its adoption of a "poison pill" that would keep any hostile takeovers at bay for the next year. Also Thursday, Wendy's reported third-quarter results that beat expectations on earnings and revenue. However, the company's same-store sales fell for the seventh quarter in a row.
The IRShas updatedthe federal income tax brackets for the 2027 tax year, lifting the thresholds to account for inflation. Financial advisorDan Sheehansays the changes "helpprotect taxpayersfrom 'bracket creep,' where inflation driven wage increases can push more income into higher tax brackets without a corresponding increase in purchasing power." The standard deductions will rise to $16,600 for single filers and $33,200 for married couples filing jointly. Related provisions like capital gains thresholds and gift and estate tax exemptions also get a boost.
The AI Data Center Backlash Is a Myth Meta Wants You to Buy
Meta’s executive team is out doing damage control, claiming that recent election results and community pushback will make building AI data centers nearly impossible. They want you to believe the public is rising up in an anti-tech crusade, rejecting the physical infrastructure of the artificial intelligence boom. It makes for a dramatic headline, but it completely misdiagnoses the situation. The public is not afraid of the future. They are simply tired of subsidizing Big Tech’s private infrastructure while getting nothing in return.
The real conflict has zero to do with an irrational fear of artificial intelligence and everything to do with basic economics. When a tech giant rolls into a region, it demands massive local power grid capacity, consumes millions of gallons of water for cooling, and negotiates massive tax breaks from local governments. In return, a fully operational data center generates a handful of security and maintenance jobs while driving up local utility bills for everyday residents. Framing community opposition as an "anti-AI backlash" is a convenient narrative shift that lets tech leadership blame public ignorance rather than their own lopsided dealmaking.
If Meta and its peers actually want to build gigawatt-scale facilities without public uproar, the solution is obvious: stop asking host communities to carry the burden. When tech companies pay for their own dedicated clean energy generation, fully fund local municipal services, and pay their fair share of regional taxes, local resistance vanishes. The public does not hate data centers. They hate bad deals. Until Silicon Valley changes its playbook from resource extraction to genuine partnership, every single data center proposal deserves to get voted down.
Why Meta’s AI Bet Is Headed for a Compute Wall
Meta is projecting supreme confidence in its massive infrastructure spending, reassuring investors that even if its consumer AI tools stall, the underlying hardware can easily be leased out to third parties. This assumes that raw compute power will hold its value indefinitely like digital real estate. It is a massive miscalculation. We are fast approaching a market dynamic where raw processing power becomes a hyper-commoditized, low-margin asset, rendering these multi-billion-dollar physical investments far riskier than leadership admits.
The assumption that proprietary tech giants will permanently dominate the infrastructure landscape ignores the rapid rise of open-source models and hyper-efficient algorithmic design. Smaller, leaner models are continuously proving that you do not need a gigawatt facility to deliver state-of-the-art results. By locking up tens of billions of dollars in physical plant and equipment today, Meta is tying its capital to current hardware architectures that could easily be rendered inefficient by tomorrow's algorithmic breakthroughs.
Building massive data centers without a clear, high-margin software business to absorb that compute is a classic speculative bubble. Calling compute "monetizable no matter what" sounds suspiciously like commercial real estate pitches right before a market shift. If Meta cannot convert its consumer AI features into hundreds of billions in direct revenue, those colossal facilities will quickly transform from prized strategic assets into crushing corporate overhead.
THE HYPE VERSUS THE REALITY OF AWS AGENTIC AI
Amazon Web Services recently announced Kiro bots, a new feature in its Kiro Crew development workspace designed to handle recurring development jobs with persistent memory. The tech industry is currently drowning in a sea of blind optimism regarding agentic AI, with every vendor promising that autonomous bots will soon write, test, and deploy our code while we sip coffee. But a closer look at AWS’s actual implementation reveals something far more grounded, and frankly, much more interesting. This is not about replacing developers. It is about managing the messy reality of delegating tasks to machines.
THE CORE PROBLEM WITH AI DELEGATION
When developers assign multiple jobs to AI simultaneously, a familiar bottleneck emerges. Someone still has to check which agents stalled, reconcile conflicting results, and re-explain requirements when work resumes. AWS is not trying to solve this with a magic, all-knowing model. Instead, they are proposing a structured hierarchy. You assign one bot to monitor failed builds, another to review proposed code changes, and a third to verify release readiness. Each bot retains its own context and can direct multiple worker sessions. It is a quiet admission that AI, in its current state, requires strict orchestration, not just clever prompting.
MEMORY AND PERMISSIONS DONE RIGHT
The most practical aspect of this launch is how it handles memory and permissions. Each bot’s memory is tied strictly to its specific assignment. A research bot can accumulate context without contaminating the decision-making context of a deployment bot. More importantly, these bots do not have unrestricted authority. A review bot can comment and request revisions, but it cannot approve or merge code. Developers define the job, provide access to relevant systems, and determine what completion actually means. This boundary-setting is crucial for enterprise adoption, yet it is rarely highlighted in the breathless marketing of AI tools.
THE OPAQUE REALITY OF AI COSTS
AWS allows teams to use different models for the coordinating bot versus the worker bots. This is a smart, cost-conscious design choice. You can reserve a highly capable, expensive model for high-level planning while assigning repetitive, low-stakes checks to a cheaper alternative. However, the real cost remains opaque. A subscription fee does not equal a completed task. Consumption depends entirely on complexity, model selection, and how often the bot loops or escalates unnecessarily. Without published benchmarks or measured error rates, teams are left to guess whether a continuously scheduled bot will save time or simply generate a new category of administrative overhead.
NAVIGATING A CROWDED COMPETITIVE LANDSCAPE
AWS is entering a crowded field. Cursor, Anthropic, and OpenAI already offer variations of scheduled tasks, persistent memory, and background execution. Cursor allows agents to create pull requests and demonstrate changes with screenshots. Anthropic focuses on repository connectors and hosted execution. OpenAI separates local desktop tasks from web-based workflows. Kiro’s distinct advantage is not raw power, but product design. It offers a continuing conversation around a specific responsibility, paired with separate memory, distinct permissions, and a dedicated coordinator.
THE VERDICT FOR ENGINEERING LEADERS
The true test of Kiro bots will not be how much code they generate, but whether they actually reduce the management overhead of AI-assisted development. A system that catches more build failures but produces endless, unnecessary escalations is not a productivity tool. It is just a different kind of technical debt. The smartest approach for engineering leaders is to start small. Assign a narrow, recurring responsibility. Constrain the bot’s authority. Review the results rigorously before expanding its scope. A larger crew of AI agents only becomes valuable if it reduces the coordination burden, rather than multiplying it.
Muse Gets a Voice: Meta's AI Agent Can Now Talk — and Animate — While It Chats
Meta is bringing voice conversations to its Muse AI agent, a move announced by CEO Mark Zuckerberg that lets users chat with their agent out loud in real time.
The feature is already live for some users, and it comes with a visual twist: Muse's avatar sits at the top of the conversation and animates in sync with its speech, making the interaction feel more like talking to a character than querying a chatbot. Voice tone is customizable too — you can either describe the vibe you want and let the agent pick, or browse the avatar settings for a list of preset options.
Early hands-on impressions suggest the voice responses are quick, though the agent will occasionally pause mid-conversation when it needs to look something up — the same behavior you'd see in a text chat. The animation is generally smooth, although some avatars land squarely in uncanny-valley territory. One tester who set their avatar to look like Mark Zuckerberg himself (as a deliberate reminder they were talking to a Meta product) found the talking digital Zuck "quite weird" — and noted the voice can't quite be tuned to sound like him either.
For now, voice mode and expressive avatars are rolling out in the US on the Muse app. Support for Meta's AI glasses is coming soon, and the feature will also be available on the Muse Charm, Meta's upcoming wearable gadget expected to ship later this year.
From the "Gen Z stare" to their purportedtelephobia, the nation's youngest workers don't often get a lot of credit for their social skills. But this cohort might be savvier than some of their older peers presume. On LinkedIn, early-career professionals are sharing how they connect with colleagues and build lasting relationships. Their tips include "give first" and "follow the conversation beyond work." See more of their emotionally intelligent advice below — and, as always,share your ownin a post.
Democrats Are Betting American Jobs on Canadian Goodwill
The plan making the rounds in Democratic circles right now is pure political theater dressed up as economic strategy. After years of lecturing the country about the dangers of free trade deals that hollowed out the industrial Midwest, senior Democrats have suddenly rediscovered the magic of open borders for goods. Their first target is Trump’s Canada tariffs. The argument is that these levies are hurting farmers, manufacturers, and consumers in border states. The reality is simpler and uglier: Democrats are preparing to prioritize Canadian exporters over the American workers they claim to represent, and they are counting on enough Republican defections to make the betrayal bipartisan.
Ron Wyden is already shopping legislation. Border-state Democrats who just ran against the tariffs are ready to move. Some Republicans who faced voter anger are softening. The media will frame this as a common-sense correction. What it actually is remains a return to the old arrangement that left Michigan, Ohio, Pennsylvania, and Wisconsin paying the price for Ottawa’s protectionism and Beijing’s overcapacity.
Canada has spent decades running a one-way trade relationship with the United States while lecturing Americans about rules-based order. Softwood lumber, dairy, digital services, and energy have all been managed to Canadian advantage. When Trump finally imposed real costs, the same voices that shrugged at Chinese dumping and European industrial policy suddenly discovered the sacred importance of seamless North American supply chains. The sudden concern for Michigan auto workers and Maine lobster fishermen is not newfound empathy. It is electoral arithmetic. The same politicians who spent the last decade celebrating “Buy American” rhetoric now treat tariffs aimed at a wealthy neighbor as an unforgivable sin.
The deeper problem is structural. Congress long ago handed the president enormous unilateral power over trade for a reason. National security, industrial base, and leverage against freeloading partners do not wait for 60-vote Senate majorities. Democrats spent the Biden years using those same authorities when it suited them. Now that the authorities are being used against Canada rather than in the service of climate or labor side deals, the tools are suddenly illegitimate. That is not principle. That is power.
Even if Democrats take the House and pick up seats in the Senate, the math does not work. Overriding a presidential veto requires two-thirds in both chambers. That means peeling off dozens of Republicans who still remember what happened to manufacturing towns under the previous free-trade consensus. Border-state Republicans may grumble about specific tariff rates, but most of them are not prepared to hand the White House a public humiliation that would be read in every foreign capital as American weakness. The Canada fight is being sold as the easy first win. It is more likely to become the demonstration project that proves Congress still cannot reclaim trade authority when it actually matters.
Meanwhile the administration is preparing the next round of Section 301 and Section 232 actions. Industrial robots, medical devices, and Chinese overcapacity are coming. Those measures will hit different constituencies and different supply chains. Democrats will discover that the coalition against “arbitrary” tariffs shrinks the moment the targets are no longer Canadian dairy and lumber. The same lawmakers who want to gut the Canada tariffs will quietly protect the steel and aluminum measures that protect their own districts.
There is a legitimate argument for reforming how tariff authority is used. Congress should not have surrendered so much of its constitutional role. But the current Democratic effort is not about restoring balance. It is about restoring the pre-2016 status quo in which American industrial policy was optional and Canadian, European, and Chinese industrial policy was accepted as the cost of doing business. That status quo produced hollowed-out towns, brittle supply chains, and a political realignment that still confuses coastal Democrats.
If the new majority wants to force votes and create political problems for the White House, they can. They will not reverse the underlying strategy. Tariffs are leverage. Leverage only works if the other side believes you will use it. The moment Congress signals that Canadian exporters have a permanent political constituency inside the United States, the leverage evaporates. The result will not be cheaper goods and happier border communities. It will be another round of quiet concessions, more offshoring pressure, and another generation of politicians explaining why the factories never came back.
The Canada tariffs are not the beginning of a thoughtful rebalancing. They are a test of whether any American administration is still allowed to put American producers first when the producers in question are Canadian. Democrats have chosen their side. The only remaining question is how many Republicans are prepared to follow them.
Google, Meta and TikTok Sue California Over Kids' Social Media Law
Google, Meta and TikTok have each filed lawsuits to block a California law aimed at protecting children from social media harms, arguing it violates their First Amendment rights.
The law—set to take effect Jan. 1, 2027—bars tech companies from offering addictive feeds, endless scrolling and autoplayed videos to kids under 16, and requires "reasonable measures" to verify users' ages. Gov. Gavin Newsom signed it in September.
The companies filed in federal courts across California, also arguing the law violates Section 230 of the 1996 Communications Decency Act, which shields tech companies from liability for third-party content.
The suits come amid mounting scrutiny over platforms' harms to children. In August, Meta agreed to pay up to $18 billion and strengthen child-safety measures on Facebook and Instagram in a landmark settlement with nearly every state over teen social media addiction. In September, TikTok settled with Alabama over addictive features.
California Attorney General Rob Bonta's office, the defendant in all three suits, said it is reviewing the filings and is committed to defending state laws that protect kids online.
TikTok called the law "a blatantly unconstitutional restriction on free expression," saying it limits free speech and takes away parents' rights to decide what's appropriate for their kids. Meta said the law lets the government "dictate how ideas are organized, displayed, or shared," and would force it to either strip age-appropriate features or cut thousands of California teens off from their communities. Google said "overly broad laws like this actually break the systems we've built to help kids find high-quality, age-appropriate videos."
