Century Report claim / AC-02
AI exposure across the labor market
AI is stress-testing the entire architecture of employment at once. Two-thirds of US and European jobs face automation exposure. The technology is revealing fragility, not creating it.
Original claim from After Capitalism: Why the Impending System Collapse Is Natural, and How to Help the Commons Build Something Better, Your Job Has No Place in the Age of AI. That's a Good Thing..
- Editorial assessment
- Leaning supported
- Claim horizon
- 2025-2032
- Supporting entries
- 164
- Complicating entries
- 16
Published assessment
The claim under examination
The tracker stores Leaning supported as this claim's editorial assessment. Evidence evaluation is complete through .
Counts describe qualifying entries. Each entry carries its own relevance, context and limitations. The published status remains an editorial judgment.
Complicating evidence
SemiAnalysis launched GDPx with 206 workplace tasks lasting up to eight hours across 11 domains; Opus 5.5 led at 12%, Carbon scored 10.9% and Astra 9.6%.
Open this editionSupporting evidence
GitHub cites Faros AI findings that pull requests spend 441% longer in review as teams move from low to high AI adoption.
Open this editionEditorial review record
- Status review date
- Not supplied in this release
- Reviewer
- Not supplied in this release
Status rationale
A written rationale for the current status is not supplied in this release.
Claim framework
How the claim is expected to unfold
AI doesn't create the fragility of wage dependence; it exposes what was always there: that survival is conditional on market demand for specific human tasks, and that most workers had no structural cushion. When AI can perform those tasks, the system's design flaw becomes visible.
Assumptions the claim depends on
Technical
- AI capability continues expanding across white-collar task categories
Economic
- Employer incentives favor AI adoption over maintaining headcount
What could falsify the claim
These are the potential falsifiers recorded with the claim.
- Net new job creation from AI exceeds displacement through 2030
- AI-exposed workers successfully transition to new roles at scale
Complete record - Complicating
Complicating evidence 16 entries
Newest entries appear first. Every entry links to the edition that covered it.
Read the other evidence recordNovember 2026
SemiAnalysis launched GDPx with 206 workplace tasks lasting up to eight hours across 11 domains; Opus 5.5 led at 12%, Carbon scored 10.9% and Astra 9.6%.
October 2026
Proton cancelled November AI re-interviews and promised no AI-related job losses in its next reorganization; an employee said the measures would enter performance reviews.
Apollo chief economist Torsten Slok found flat profit margins and revenue per worker across the S&P 493 outside the seven largest tech firms.
September 2026
CESifo economists found no significant displacement or hiring decline among US bachelor's recipients aged 22 to 25 through August, while a separate tracker counted 94,046 tech layoffs through August.
September data showed no AI-driven surge in US job losses.
An analysis found 2026's largest tech layoffs at Intel, Amazon, Oracle, Verizon, and Meta were mostly net headcount declines and reallocation rather than AI-driven cuts.
August 2026
Meta shelved planned team cuts of up to 60% after agents generated 220% more code but only 36% more shipped features, while major incidents rose 40% and firefighting rose 70%.
Unemployment rose 0.77 points among the most AI-exposed workers versus 0.85 for the least exposed, while recent-graduate unemployment reached 5.6%.
At least 13 of 21 documented security-robot deployments since 2015 ended in canceled contracts as Knightscope and other vendors pivoted back toward human guards.
BBC documented 70-90 hour weeks and multi-day "sprints" at OpenAI, Anthropic, Meta, and Google despite years of AI four-day-workweek promises, with a UC Berkeley study finding AI expanded total work demands rather than reducing them.
June 2026
High-intensity AI adopters grew both overall headcount and entry-level roles.
Ford rehired more than 350 veteran engineers to correct errors its automated design and production systems introduced, after becoming No. 1 in JD Power initial quality for the first time in 16 years.
A survey found that only 1% of laid-off workers attributed their job loss to AI, while a Science study reported a one-third figure.
Research showing unemployment attributed to remote work rather than AI.
April 2026
A Guardian investigation documented workers describing AI-generated output as "workslop" that increases rather than reduces workload, with 40% of non-managers reporting AI saves them no time.
March 2026
OpenAI plans to nearly double workforce from 4,500 to 8,000 - AI companies themselves still hiring aggressively.
Complete record - Supporting
Supporting evidence 164 entries
Newest entries appear first. Every entry links to the edition that covered it.
Read the other evidence recordNovember 2026
GitHub cites Faros AI findings that pull requests spend 441% longer in review as teams move from low to high AI adoption.
A Northeastern University survey of 8,174 hiring decision-makers found 26% of those whose decisions AI shapes now hire fewer entry-level candidates; September JOLTS data showed hiring at its slowest pace.
PwC analysis of 2.4 million US entry-level postings found AI-influenced roles seven times more likely to demand senior-level skills, as colleges began training data-center technicians for Microsoft.
Canva cut or relocated 1,271 jobs, 22% of staff, offering six months' salary, a year of extended healthcare, and $10,000 startup grants to departing employees.
Anthropic says internal Claude consumption grew over 250-fold from January to September; its hypothetical list-price bill first exceeded monthly personnel expense excluding stock compensation in September.
PostgreSQL's previous-stable-release fixes rose from 248 to 433 as LLM-assisted reports uncovered longstanding defects; human patch review became the bottleneck delaying version 19.
October 2026
Cisco gave engineer Padmanabhan Krishnamurthy unlimited AI tokens for a proposed two-person rewrite of 98% of its legacy code within 24 months, starting with a four-week trial.
Microsoft reported $36.6 billion quarterly profit on $92.3 billion revenue two days after offering voluntary buyouts to about 60,000 employees.
AFL-CIO President Liz Shuler named labor-drafted AI legislation covering worker protections, privacy, data rights and economic impacts its highest priority for the next Congress.
The Netherlands plans to eliminate about 25,000 full-time civil-service posts by 2035, over 15% of its workforce, under savings plans including AI adoption.
Microsoft offered buyouts to up to 60,000 employees across Copilot, CoreAI, Windows, Xbox, finance and HR. Proton told staff it will re-interview them on AI use between November 4 and 11.
China's State Council ordered AI-displacement monitoring, contingency plans, re-employment support and safeguards against discrimination and privacy violations in AI-managed hiring.
Infosys cut fiscal 2027 growth guidance to 1% to 2% as clients demanded AI productivity savings through lower prices and outcome-based fees.
Barclays announced about 5,000 job cuts; CFO Anna Cross cited fewer developers as AI raises productivity, while CEO CS Venkatakrishnan emphasized wider technology investment.
Disney will pay about $60 million a year to employ 66 of Character.AI's 86 staff, with 56 already resigned from the bankrupt startup, while its lawsuits and regulatory investigations stay with a bankrupt estate.
Philadelphia police arrested 78-year-old retired professor Carolyn Marvin for disorderly conduct after months of standing in front of Uber Eats delivery robots to protest courier displacement.
Anthropic's own employment data showed junior technical hires fell from 27% to 9% of new hires as AI agents logged six times staff hours and meeting time rose 20%.
Fed Chair Kevin Warsh characterized AI investment as a demand shock arriving before supply, as central bankers considered higher interest rates.
Accenture centrally funds $29 monthly per employee for many Claude users, down from some users' $500 to $1,000 spending; projects or teams must cover overages.
Epic released a preview launcher rebuilt by a small team using AI, extending its production shift toward model-written software.
A financial-sector survey found a quarter of firms tie pay or promotions to AI proficiency and weigh AI skills nearly four times above academic pedigree.
A Goodwin partner left to lead AI-native law firm OmniLex; McKinsey began testing applicants through simulated client interactions.
Citigroup CEO Jane Fraser said AI coding assistance cut the bank's technology vendor expenses by 30%.
ECB chief economist Philip Lane warned AI replacement threats may weaken worker bargaining, moderate wage growth and reduce labor's share of income.
Epic reportedly mandated model-written code for its delayed Disney shooter, with developers directing systems such as Claude.
Singapore President and World Bank Jobs Council co-chair Tharman Shanmugaratnam urged governments to make crisis-era hiring supports permanent as AI displaces tasks.
Staff at HarperCollins, Simon & Schuster and Hachette told WIRED that AI now drafts publicity copy and agent emails, while entry-level publishing pay averages $47,583 and savings flow upward.
China announced an employment initiative adapting worker skills to AI development, reporting 10.52 million urban jobs created over nine months, 87.7% of the full-year target.
Samaya reported GPT-6 Astra, Claude Fable 5.1 and Opus 5.5 beat bias-corrected analyst consensus on earnings surprises across 456 companies worth more than $5B each.
Flock Safety planned to cut about 270 of roughly 1,500 employees after voluntary buyouts amid community and legislative opposition to its AI license-plate network.
The International Trade Centre estimates transcription, data-entry, translation and web-research jobs fell about 50% since 2022.
Co-op Legal Services deployed an OpenAI model scoring more than fifty aspects of probate calls, giving managers pass-and-fail results partly to identify sales improvements.
Across 243 AI-linked layoff events at 208 employers, the AI Layoffs register found employer-stated AI attribution and job counts for 59,454 of 358,974 jobs, or 17%.
Gallup found 28% of US employees use AI weekly and 54% never use it at work; seven in ten earning at least $10,000 monthly report faster work.
An NBER working paper estimated AI writing appeared in 29% of 2026 US STEM dissertations; users were less likely to enter academic research, a correlation rather than established causation.
HackerRank launched its Chakra AI interviewer after 500,000+ trial interviews, testing candidates in real repositories and questioning their decisions; candidates cannot inspect scoring rubrics.
Challenger data showed AI drove 120,136 of 2026's 573,195 announced US job cuts through September, 21% and the top-cited reason, as bank postings for "agent orchestration" skills jumped 1,721%.
California barred employers from using AI to decide firings or infer workers' emotions from biometric data and required written notice when AI drives mass layoffs.
September 2026
McKinsey estimated about 11 million US workers, 7% of the workforce, may need to change occupations by 2035 because of AI, as a UK minister called for "pandemic preparedness"-style contingency planning.
UK's IPPR drafted a "pandemic preparedness"-style AI jobs report proposing to rework the tax code so retaining workers costs less than automating them, as government minister Kanishka Narayan named universal basic income as a policy option.
Microsoft released Copilot Code, letting non-developers build and host purpose-built workplace apps from plain-language descriptions.
Tesla ramped Optimus humanoid production to several hundred units weekly at Fremont, with some workers moved from self-driving data annotation to generate training demonstrations for the robots.
Goldman Sachs estimated call-center employment has fallen 39% below trend in the US, 33% in Canada, and 27% in Germany since 2022 as AI slows white-collar hiring.
Alibaba's DAMO Academy published RADAR, an open-source model reading a single abdominal CT for 146 diseases across 18 organs, beating 23 of 26 specialist radiologists and lifting junior readers to expert-level performance.
A US tracker counted 210,741 layoffs through mid-September, with 49% citing AI or automation, while former Uber staff said AI tools reduced some workloads to one-fifth despite management-layer pushback.
Zurich Airport put two WeRide-built Level 4 electric shuttles into service with no driver or onboard safety monitor after more than 15,000 km of staged testing, relying on a remote operator watching multiple vehicles.
The share of UK computer-science graduates landing coding jobs fell from about 40% to 28% in a year, with demand also declining for economics graduates in well-paid finance roles.
OpenAI launched GPT-6 Astra-powered ChatGPT for Financial Services, producing acquisition research, checked models, charts, and a house-style pitchbook in about ten minutes for work traditionally done by analysts.
ByteDance's Xpert recruited more than 50,000 professionals to teach AI their trades for 100 to 500 yuan per task as Chinese youth unemployment reached 17.9%.
Blizzard's 1,900 CWA-represented workers ratified a first contract requiring the studio to bargain over workplace AI use and granting laid-off staff 14-month recall rights into any bargaining unit.
A hedge fund manager rebuilt his firm to run entirely on AI agents at under 1% of its former roughly $5 million payroll.
Edinburgh food-delivery riders documented earning about half what the same hours paid four years ago and began reconstructing Deliveroo's Frank pay-setting algorithm from their own logs.
OpenAI reported its coding agents now log 3.1 workdays of effort for every human workday and said it has hit its "automated research intern" milestone, while chief scientist Jakub Pachocki cautioned the gains remain uneven across task types.
About 2,000 University of Sydney staff walked off the job seeking protections on AI-driven job security, workload, and workplace fairness.
August 2026
Bank of England Governor Andrew Bailey warned G20 ministers that AI-fueled investor optimism, high leverage, and concentrated markets could amplify a financial correction.
Meta is testing Watney, ABB, and Kinova robots to swap cables and reset data-center servers, targeting up to 80% of some workloads, as more than half of tracked 2026 layoff events cite AI or automation.
Bill Gates reversed his 2023 "manageable transition" stance, calling AI job disruption "utterly, utterly a unique moment" affecting every occupation and expecting "far fewer" jobs overall.
A Richmond Fed brief found job-finding rates fell approximately 13 points for strongly-attached workers in AI-exposed occupations since late 2022 versus approximately 2 points for less-attached workers.
SpaceXAI released Grok 4.6 alongside Grok Bot, an always-on agent that signs into workplace apps and independently completes multi-step tasks.
Major Indian IT-services firms cut headcount by as much as 6% as AI absorbed standardized services work, with the sector's stock index down 18% in 2026.
BBC reporting found staff at OpenAI, Anthropic, Meta, and Google logging 70-90 hour weeks and multi-day sprints four years after AI four-day-workweek predictions, with a UC Berkeley study finding AI expanded task scope.
Rippling released a console tracking per-employee AI token spending after its own AI bill reached 40% of R&D payroll.
Airbnb reported cutting concept-to-launch time 60% using AI.
Waymo opened driverless rides to everyone in Dallas, Zoox began paid Las Vegas service, and Wayve/Uber received London licenses for supervised robotaxis.
In a blinded study of 1,682 readers, AI-generated stories rated more absorbing and higher-quality than human work, while Spotify brought 30,000 labels into consent-based AI covers.
NHTSA cleared Zoox's steering-wheel- and pedal-free robotaxi under a two-year exemption covering up to 2,500 vehicles annually.
July 2026
An Apollo study across 321 occupations found the most AI-exposed jobs lost 6.7% in real-wage growth after 2023 without detectable employment decline, as call-center cuts spread.
SK Hynix offered memory-chip engineers signing bonuses near $476,000 as workers defected from Samsung amid the AI-driven shortage.
Visa cut roughly 2,600 jobs and explicitly cited AI as it restructures staffing across the payments network.
Recursive Superintelligence signed a $410M multi-year AWS compute deal to scale self-improving systems that automate its own R&D, prioritizing agent count over headcount.
Amazon closed its AGI-focused AI agent research lab, laying off researchers and redirecting roughly $1 billion toward deployment.
British Gas, Uber, and Patreon announced layoffs citing AI or changing behavior, while Google's ATLAS study of 15 million interactions found AI now touches 88% of the workforce.
Monday.com cut ~630 employees, or 20% of staff, to refocus on its AI Work Platform, and Nine cut ~30 newsroom jobs citing "extreme" AI disruption.
Claude Tag produces 65% of Claude Code team pull requests as idea-to-shipment time falls from 6-12 months to roughly one week, shifting the human bottleneck from implementation to specification.
A 24-year-old Istanbul developer runs a two-person studio shipping games with Claude Code that once needed a dozen developers, as Turkish game-studio cohorts fell from 200 to 84 to 30 companies.
Unionized Hyundai workers in Ulsan cut shifts short and scheduled four-hour strikes after 15 negotiation rounds failed over the company's plan to deploy 25,000-plus Atlas humanoid robots.
Stanford's Canaries Dashboard showed employment in the most AI-exposed occupations slipped 0.5 percent while least-exposed roles grew 0.2 percent. Early-career workers in exposed fields were down 2.7 percent, and senior-level roles accounted for 71 percent of the increase in software developer postings.
More than 200 economists, Nobel laureates, and AI lab leaders signed a statement urging policymakers to act now on AI job displacement. New York became the first state to freeze permits for hyperscale data centers over 50 megawatts.
Google reported that 75% of its code was AI-generated, and software engineers shifted from writing code to reviewing and orchestrating AI-generated code.
China dropped its numerical urban jobs target for 2026-2030, the first time in decades, citing rising employment uncertainty as AI spreads through the economy.
Microsoft eliminated roughly 4,800 jobs in an Xbox reset that gutted roughly half of id Software and a quarter of Obsidian, while committing $2.5 billion to embed 6,000 AI engineers inside customer operations.
Tech-sector cuts surpassed 120,000 roles in 2026, with Microsoft cutting about 4,800, Oracle 21,000, Meta 8,000, and Cisco 4,000, many at companies posting record revenue.
Amazon's Mechanical Turk platform was hollowed out by bots and fraud, with the micro-task economy it pioneered thinning as the layer of labor it coordinated moved up the stack.
AI was the top-cited layoff reason as finance and information sectors shed approximately 150,000 jobs year-to-date.
A July 2 analysis found that conventional labor statistics struggle to measure AI's workforce impact, as surveys were built to count jobs gained and lost rather than roles reorganized around AI collaboration.
A major investment bank projected 15 million US worker displacements from AI.
Challenger, Gray & Christmas attributed roughly 102,000 job cuts so far this year directly to AI, with finance and information sectors shedding around 28,000 positions a month.
The federal Workforce Pell program opened this month to fund short-term skills training for displaced and transitioning workers.
The BLS, Stanford's Digital Economy Lab, and the California Policy Lab are building the first instruments precise enough to separate automation from augmentation in payroll and unemployment-insurance data.
June 2026
A Unitree humanoid running Flexion's software composed simulation-trained skills to autonomously fetch a parcel by stairs and elevator, unpack it, and shelve the items from a single command.
Cursor's data showed a growing share of AI-generated code merging into production with little or no human review. Cloudflare grew its engineering ranks 45 percent while valuing builders over those who verify others' work.
Bloomberg reported on Silicon Valley burnout through engineer Matt Van Horn, who described engineers growing afraid to log off because agents keep working when they do not. The cognitive load migrated from writing code to holding context across a swarm of processes that never sleep.
Oracle cut 21,000 roles it attributed to AI, disclosed in a regulatory filing alongside borrowing to expand AI capacity.
Anthropic's June Economic Index documented Claude usage shifting from chat to long-running agentic work.
California launched a real-time AI job-loss tracker linking occupational AI-exposure measures to monthly unemployment-insurance claims, built as a partnership between the state's Employment Development Department and the California Policy Lab at UC.
Major AI firms and employers launched a $500M retraining fund.
Oracle disclosed in an SEC filing that AI adoption helped drive 21,000 job cuts, 12.9% of its staff, over the past year, while announcing plans to borrow up to $50 billion to expand AI cloud capacity.
Claude Code creator Boris Cherny described 'loops' of agents endlessly prompting other agents to write code, with agents submitting pull requests continuously in his own work.
AI performed physical robotics tasks 20x faster than the fastest human team.
Klarna replaced full-time customer service roles with AI and contingent gig labor.
A North Carolina engineer won a religious exemption from mandated AI use.
Massive AI-driven layoffs with explicit AI causation demonstrates employment architecture under stress from automation.
AI stress-testing employment architecture as even AI companies impose extreme monitoring and restructuring on their own workforce.
A trillionaire predicting AI/robotics will make money irrelevant suggests imminent automation pressure on employment architecture.
Major corporate investment in AI systems specifically designed to automate engineering jobs.
Massive investment in highly flexible factory automation directly targets manufacturing jobs with adaptable robotic systems.
Company closed offices and eliminated jobs, citing AI-native structures.
Systematic targeting of back-office roles for AI replacement demonstrates employment architecture being stress-tested across multiple sectors simultaneously.
JPMorgan readying AI agents to operate unsupervised for 1 to 2 hours this year; chief analytics officer Derek Waldron cited an "intellectual coherence" threshold and reported 20% gross private-banking efficiency gains.
AI handling routine enterprise tasks at scale demonstrates automation pressure across knowledge work employment sectors.
AI becoming the primary driver of job cuts with specific automation deployment confirms widespread employment disruption.
AI tools disrupting traditional software development workflows and cost structures at major technology companies.
Massive productivity gains in engineering demonstrate AI's potential to automate significant portions of technical jobs.
Profitable companies cutting significant workforce due to AI capabilities demonstrates automation pressure across employment architecture.
May 2026
Direct workforce displacement to fund AI capabilities demonstrates automation pressure on employment across enterprise software.
AI costs reaching parity with human labor creates direct economic pressure for employment displacement decisions.
Direct training of robots to replace domestic workers using human-provided data shows automation extending to service employment.
Major tech companies cutting significant portions of their workforce specifically due to AI automation.
Editorial workers facing AI monitoring systems shows AI stress-testing employment architecture across knowledge work sectors.
Complete automation of debt collection represents AI stress-testing employment architecture with entire occupations facing elimination.
AI systems performing cybersecurity work at machine speed shows automation pressure on technical employment.
Military adoption of AI for strategic operations demonstrates AI stress-testing employment across all sectors, including specialized military roles.
Large-scale job elimination due to AI automation at profitable companies.
Large-scale AI-driven layoffs alongside AI handling majority of technical work demonstrates employment architecture being stress-tested by automation.
Direct example of AI displacing skilled workers and management explicitly preferring AI over human capabilities.
Massive AI investment surge (from 17% to >33% of private credit) demonstrates AI is indeed stress-testing employment architecture and economic systems at unprecedented scale.
AI researchers organizing against their own work.
Direct replacement of human manufacturing lines with robots demonstrates AI stress-testing employment architecture at scale.
Mass job elimination and legal disputes over AI-justified dismissals show employment architecture under stress from automation pressures.
AI automation of recruitment processes.
April 2026
AI automation displacing hundreds of data annotation jobs.
Customers Bank's CEO delivered Q1 earnings remarks via an AI clone of his voice and announced a multiyear OpenAI partnership embedding engineers to automate lending and onboarding.
Young people abandoning traditional employment and entertainment industry favoring AI actors shows automation pressure across job categories.
AI agents creating systematic advantages in negotiations.
The rapid automation of coding work at a major tech company.
Simultaneous mass layoffs at major tech companies while increasing AI investment demonstrates employment architecture being stress-tested by automation.
AI systems being trained on employee work patterns to replace human jobs, showing automation pressure across employment architecture.
AI agents automating industrial engineering tasks with superior efficiency demonstrates AI stress-testing employment across sectors.
AI automation targeting knowledge work, with workers forced to participate in their own replacement.
Large-scale tech layoffs at Meta demonstrate AI's stress-testing of employment architecture as automation pressures mount.
AI is already transforming developer work patterns and productivity, demonstrating automation's impact on knowledge work employment.
Concrete example of AI stress-testing employment architecture with significant workforce reduction and job displacement.
Educational sector experiencing systematic job displacement as AI capabilities undermine traditional teaching and assessment roles.
Employment architecture being stress-tested with dramatic workforce composition changes and hiring freezes.
The shift from traditional coding to AI agent management.
Large-scale corporate layoffs in knowledge work roles demonstrates AI stress-testing employment architecture across sectors.
Rising anger about AI among high-usage Gen Z workers suggests growing awareness of automation's employment threat.
AI-driven job displacement affecting both blue-collar and white-collar workers simultaneously.
Skilled workers reduced to AI data-labeling gigs demonstrates automation's impact on traditional employment structures.
Major AI companies promoting reduced work weeks indicates recognition of automation's impact on traditional employment.
GEN-1 demonstrated that a general-purpose AI system can learn multiple physical manipulation tasks, expanding automation deployment from specialized manufacturing into warehousing, logistics, and maintenance.
OpenAI's endorsement of wealth redistribution mechanisms specifically to address AI displacement confirms automation's threat to employment architecture.
AI enabling non-engineers to code demonstrates automation pressure extending beyond traditional job categories.
AI automating complex chip design work demonstrates employment disruption reaching highly skilled technical roles.
Large-scale job eliminations for AI infrastructure investment demonstrates AI stress-testing employment architecture across major corporations.
Global gig workers creating training data for automation technologies that will replace human labor shows AI's comprehensive employment impact.
March 2026
Quinnipiac poll: 70% expect AI to reduce job opportunities; 30% fear their specific job will be made obsolete (up from 21% a year ago).
A Tufts University study predicts 9 million American workers will be displaced by AI/automation within 5 years, indicating significant employment disruption.
By 2026, 43% of U.S. workers were using AI for their jobs compared to 32% average in European countries, showing rapid AI adoption across employment.
Tufts American AI Jobs Risk Index identifies 4.9 million workers across 33 'tipping point' occupations at highest displacement risk within two to five years.
Uber CTO disclosed an AI agent making 1,800 code changes per week at production scale.
Anthropic labor study: AI covers 94% of computer/math tasks theoretically; most-exposed workers earn 47% more than average - displacement targets high-income knowledge workers.
February 2026
Americans lack sufficient savings for extended job searches amid AI-driven displacement - structural vulnerability predates AI.
January 2026 layoffs 108,435 (highest January since 2009). AI cited in 7% of cuts.