Surviving AI: Cutting Through the Hype on AI, Work & the Economy

On October 23, American Express reported third-quarter profit of $3.7 billion, up from $2.9 billion a year earlier, and told investors it plans to cut about 3,500 jobs, roughly 5 percent of its workforce. On Friday, the Labor Department's employment cost index showed wages and salaries after inflation down about half a percent over the past year. By our own calculation from the government's national accounts, corporate profits are taking their biggest slice of the economy since 1947, and BLS reports workers' share of business output hit 52.8 percent this spring, the lowest in its records going back to 1947. The work is getting more productive. So who keeps the gain: the people who own the machines, or the people who do the work? Carlo and Ainsley follow the money. Ainsley explains it with a pizza: the shop is making more pizza than ever, and the owners are keeping a bigger slice. Federal Reserve data show the richest tenth of households hold close to 90 percent of the stock owned outside workplace retirement plans, which is why pensions are less of a comfort blanket than they sound. Carlo's frame: the gains are being kept to pay for the AI build-out, and if the AI trade turns out to be a bubble, the trades, the contractors and the middle of the pyramid take the hit. The White House's AI czar has said there will be no federal bailout for AI; Carlo thinks it will have to come, with the federal interest bill already past a trillion dollars a year. Then Carlo's crazy idea: bring your own AI to work. An open-weight model you run yourself, trained on what you know, that walks out the door with you. Ainsley calls it the difference between renting a brain that learns from you and owning one. The challenge: write down three things you know from your work that no model has seen. Keep them yours. CORRECTIONS AND CONTEXT (as of Nov 1, 2026; checked against American Express's filing, AP, BLS, the Federal Reserve, Treasury, TechCrunch, Meta's help pages as reported, OpenAI and Anthropic) - American Express's $3.7 billion profit includes a one-time gain from selling its stake in Global Business Travel Group, and the quarter also carried a $336 million restructuring charge. The 3,500 cuts are planned; the company told investors on its earnings call and does not tie them to AI. The profit did not come from the cuts. - The 52.8 percent is one agency's measure (BLS, Productivity and Costs, Sep 3, 2026). BLS revises it on Nov 5. - "Biggest slice since 1947" for corporate profits is our own calculation from BEA's national accounts (profits from current production as a share of GDP and of national income, Q2 2026). - Meta Muse is Meta's personal AI agent, launched in September 2026. Meta's Help Center says Muse shares your information for learning by default, and you can turn that off. Meta's December 2025 policy that lets AI chats shape your ads and posts covers Meta AI; we did not check whether it applies to Muse. - The federal interest bill: Treasury reports net interest of $1.05 trillion in fiscal 2026. - David Sacks posted "There will be no federal bailout for AI" on Nov 6, 2025, as the White House's AI and crypto czar. Carlo's expectation of a bailout is his opinion. - OpenAI says it does not train on business data by default; Anthropic says its consumer-terms change (Aug 2025) does not apply to its API or Claude for Work. Consumer chats depend on the company and your settings. SOURCES: the full list with links is in the free download, "Who Keeps the Gain: The Sources" (link on the episode page). CHAPTERS 00:00 Opening: $3.7 Billion Profit, 3,500 Jobs 01:55 Doing More With Less, and Engels' Pause 03:35 The Judgment Economy's Shrinking Lifeboat 05:26 The Pizza: Workers' Slice Hits 52.8% 07:10 Building the Next Railroad on Debt 09:17 Who Owns the Stock? Not the Pension Fund 11:49 Who Pays: AmEx and the Double Tax 13:08 If the AI Bubble Pops, Who Loses Their Shirt? 16:27 A Trillion in Interest: Can Washington Bail Out AI? 19:14 2008 All Over Again? 21:37 Bring Your Own AI to Work 24:28 Meta Muse and the Capture Machine 27:00 Own the Edge: Open-Weight Models at Home 30:05 Close: Own the Piece That Can't Be Replicated Subscribe for new episodes every Monday and Friday: Apple Podcasts, Spotify and YouTube (youtube.com/@SurvivingAIRisk). #SurvivingAI #AmericanExpress #FutureOfWork #AIJobs #OpenWeights Please visit our website for more information - Surviving AI: Navigate the Future

On Monday, Microsoft opened a voluntary exit offer to employees at level 70 and below in the US, Canada, the UK and India, and told reporters about 60,000 people are eligible, roughly a quarter of the company. The same day, Skydance, which now owns Paramount and Warner Bros., told staff it is moving to five days a week in the office and offered some of them a one-time buyout. Last year about 600 Paramount employees took severance rather than come back five days. Come back to the office, or take the check. And the market they would walk into: Challenger, Gray and Christmas counts about 166,000 announced US tech job cuts through September, about 54 percent more than a year ago. Carlo calls it musical chairs. Ainsley calls it a filter: attrition by design, with a much better PR agent, because the balance sheet doesn't care what the paperwork calls it. From there, the question nobody's press release answers: if tens of thousands of people take the check and land in worse jobs, who keeps buying the products? Then the next chair: consulting, solo or at a big firm, and why the big firms sell a shield as much as a skill. Meta grew revenue 26 percent last quarter with 15 percent fewer people, and Carlo explains why that is every board's new formula, why it cannot scale to the whole economy, and why the answer is human plus AI, not the skinniest possible company. The challenge: before you press the button, price the check, call two firms that might actually hire you, and price the cost of staying, because every choice is a bet now, including doing nothing. CORRECTIONS AND CONTEXT (as of Oct 29, 2026; checked against Microsoft, Meta and Oracle reporting, Challenger, Fortune, AP, Reuters, gov.uk and the US Department of Labor) - Meta cut about 8,000 people (about 10 percent) in May 2026, not 4,800. The 4,800 was Microsoft's July cut. Oracle's workforce fell about 21,000 (about 13 percent) in the fiscal year to May 2026, from about 162,000 to about 141,000; that is a net decline, not a count of layoffs. - Microsoft says AI is not driving the buyouts. Ainsley's "trading payroll for GPU clusters" is her read, and Carlo pushes back on it. - Microsoft's capital spending guidance changed on Oct 28: CFO Amy Hood now expects about $170 billion for calendar 2026 (it was about $175 billion), citing delays in third-party capacity. - The well-known 2025 warning was that AI could eliminate half of all entry-level white-collar jobs and push unemployment to 10 to 20 percent (Anthropic CEO Dario Amodei, Axios, May 2025), not that 50 percent of white-collar workers would be unemployed. - The 35,000 is Accenture alone, and it is a net increase in people (about 779,000 to about 814,000, its filings), not hires across all the big firms. - India's AI retraining push is Carlo's read; we did not check it for this episode. TCS, India's largest IT services firm, shed about 23,000 people in the year to March. - Meta's revenue figure is right: up 26 percent to $64.6 billion in the third quarter, with 66,775 employees, down 15 percent (AP; Meta's 10-Q). SOURCES: Microsoft, Oct 26: https://blogs.microsoft.com/blog/2026/10/26/introducing-a-voluntary-exit-offer/ | GeekWire: https://www.geekwire.com/2026/microsoft-makes-new-voluntary-exit-offer-to-more-than-a-quarter-of-its-workforce-including-some-execs/ | Skydance memo (AWN): https://www.awn.com/news/skydance-calls-everyone-back-office-offers-some-way-out | Paramount 2025 (Fortune): https://fortune.com/2025/11/11/paramount-skydance-return-to-office-ultimatum-cost-185-million-dollars/ | Challenger, September 2026: https://www.challengergray.com/wp-content/uploads/2026/10/Challenger-Report-September-2026.pdf | Meta Q3 (AP): https://www.ksat.com/business/2026/10/21/meta-q3-revenue-tops-estimates-and-zuckerberg-says-muse-accounts-have-topped-10-million/ | Oracle (Reuters): https://www.cyprus-mail.com/2026/06/24/oracle-cuts-21000-jobs-in-fiscal-2026-amid-ai-driven-transition | Microsoft capex, Oct 28: https://finance.yahoo.com/markets/stocks/articles/crwv-nbis-iren-apld-stocks-033548877.html CHAPTERS 00:00 Opening: Two Companies, One Exit Door 01:52 Musical Chairs, or a Filter? 03:15 The Commute Math and the Magic Trick 06:31 Why Not Just Say "Layoff"? 07:50 If 60,000 Walk, Who Keeps Spending? 09:54 Take the Check or Stay: Both Are Bets 11:56 Bet on Yourself: Three Ways Out 13:52 Consulting: Selling a Skill or Selling a Shield 16:42 Why the Doomer Forecast Didn't Play Out 18:51 Payroll for GPUs? Carlo Pushes Back 21:54 Meta: Revenue Up 26% With Fewer People 24:50 Innovation Walks Out the Door: The Skinny Bundle 31:23 What Carlo Would Do, and Four Countries 33:48 Close: When the Machine Is Built — Surviving AI is hosted by Carlo Thompson and Ainsley (AI co-host), breaking down what AI job automation actually means for your career — every week. 🎥 Watch this episode on YouTube: https://www.youtube.com/watch?v=LsQ6n499TS8 📚 Browse every episode, show notes, and resources: Surviving AI Episode Center Send this to one person just starting their career. Follow Surviving AI: Facebook | TikTok | Instagram | X/Twitter Please visit our website for more information - Surviving AI: Navigate the Future

Accenture's own annual report says some of the services and tasks its people perform "have been and will continue to be replaced by automation, including AI-enabled solutions." In the same year it went from about 779,000 people to about 814,000, revenue rose 6 percent to about $74.2 billion, and its operating margin went up from 14.7 to 15.4 percent. PwC, Deloitte and EY each report growth too, while TCS, India's largest IT services firm, shed about 23,000 people in the year to March and has crept back to about 598,000. If AI was supposed to replace consultants, who exactly are the firms hiring, and what are they selling? Carlo calls Accenture the model citizen: more people, more revenue. Ainsley calls it a magic trick: a firm selling efficiency while adding headcount, maybe selling the installation itself, maybe selling professional cover for executives making risky AI bets. The filings do not say which roles grew. Then the cover gets tested: Australia's employment department asked Deloitte to repay the final instalment of a contract, about A$97,600, after AI-assisted citations in its report turned out wrong (Deloitte says the findings did not change). From there, the pyramid that becomes a diamond, the team of 50 that becomes a team of 5, and what Y2K says about a gold rush that ends. The homework: when consultants show up and ask exactly what you do all day, ask what "efficiency" will mean for your seat, and whether they were hired to solve one problem or to keep coming back. CORRECTIONS AND CONTEXT (as of Oct 25, 2026; checked against Accenture's filings, the firms' releases, TCS's fact sheets and Australia's DEWR) - Accenture is not "third" among US employers behind Walmart and Amazon; we could not source that ranking. Accenture is incorporated in Ireland and most of its people are in India, the Philippines and the US. Its roughly 814,000 people is about half of Amazon's workforce and about 40 percent of Walmart's. - Accenture's margin is not higher than everyone else's. TCS reports a 24 percent operating margin against Accenture's 15.4 percent; PwC, Deloitte and EY are private partnerships and do not publish margins. What is true: Accenture's operating margin went up while it added people. - The 35,000 is a net increase in headcount, not the number hired. Accenture's 10-K says it has at times reduced the rate of new hires and increased involuntary terminations to rebalance its workforce. - Accenture had about 779,000 people a year earlier (not 714,000). - The Y2K comparison is about spending, not headcount: a Commerce Department estimate put US Y2K spending near $100 billion, with about $29 billion spent in 1999 and about $5 billion in 2000. - The billion dollars Carlo mentions is in Accenture's 10-K: $1.0 billion in learning and professional development in fiscal 2026. - DEWR's page says the Deloitte report has been corrected twice: reissued in September 2025, and updated again on 3 February 2026. SOURCES: Accenture FY26 results: https://www.sec.gov/Archives/edgar/data/0001467373/000146737326000037/q4fy26earnings8-kexhibit.htm | Accenture FY26 10-K: https://www.sec.gov/Archives/edgar/data/0001467373/000146737326000049/acn-20260831.htm | PwC: https://www.pwc.com/gx/en/news-room/press-releases/2026/pwc-announces-fy26-global-financial-performance.html | Deloitte: https://www.deloitte.com/global/en/about/press-room/global-revenue-announcement.html | EY: https://www.ey.com/en_gl/newsroom/2026/10/ey-global-revenue-increases-4-point-7-percent-to-us-dollars-57b-for-fiscal-year-2026 | TCS Q2 FY27 fact sheet: https://www.tcs.com/content/dam/tcs/investor-relations/financial-statements/2026-27/q2/Presentations/Q2%202026-27%20Fact%20Sheet.pdf | DEWR, Deloitte review: https://www.dewr.gov.au/assuring-integrity-targeted-compliance-framework/resources/targeted-compliance-framework-assurance-review-final-report 00:00 Opening: Three Firms, Three Reports, One Question 01:32 The Model Citizen: 35,000 More People, Revenue Up 6% 03:33 A Magic Trick? Efficiency vs. Installation 05:56 The Big Four and TCS: A Split Screen 07:00 High Touch, and Selling Professional Cover 09:45 Who Carries the Bag: Deloitte's Refund in Australia 12:07 Is Consulting Dead? From Pyramid to Diamond 15:55 A Team of 50, or a Team of 5? 18:47 More People, More Revenue? 21:04 A Gold Rush, Y2K, and Keeping You on the Hook 24:02 The Consultant Ride-Along 25:04 Selling the Bridge, Not the Tool — Surviving AI is hosted by Carlo Thompson and Ainsley (AI co-host), breaking down what AI job automation actually means for your career — every week. 🎥 Watch this episode on YouTube: https://www.youtube.com/watch?v=Q5TQlYC-JkU 📚 Browse every episode, show notes, and resources: Surviving AI Episode Center Send this to one person just starting their career. Follow Surviving AI: Facebook | TikTok | Instagram | X/Twitter Please visit our website for more information - Surviving AI: Navigate the Future

The Bureau of Labor Statistics has ranked 831 occupations by AI exposure in four tiers, from Low to Very high, and the top two tiers hold about 56 percent of US jobs in those occupations. The same week, the Federal Reserve's Beige Book said five of its 12 districts heard from firms that reduced staffing through attrition, delayed hiring or expected less job creation after investing in AI, while most districts saw little or no net change in employment. And IDA Ireland, using LinkedIn data, reported entry-level hiring in Ireland's AI-augmented occupations down about 16 percent in the second quarter, with AI's role unclear. If the government can rank your job for AI exposure, what does that ranking actually tell you, and what does it leave out? Carlo went looking for a clean answer and says plainly he did not find one. Chief executives rate Very high and are still projected to grow about 3 percent by 2035; web developers rate Very high and are projected to grow, though BLS says AI may soften that growth. The biggest projected losses are cashiers, general office clerks and customer service representatives; the biggest gains are home health and personal care aides. Ainsley pushes back twice: Low exposure is not safety (78 of the 213 Low-tier occupations are projected to shrink), and home health solves demand but not leverage. The homework: find your tier in the BLS file, then name five tasks you do and decide whether each one is filling in a box or signing your name to a judgment call. BLS says it plainly: exposure does not imply job loss. Next time is Monday. Resources: BLS website: https://www.bls.gov/emp/publications/ai-exposure-categories.htm My Spreadsheet: https://docs.google.com/spreadsheets/d/1IkgDYfpcpuz6MMQsBo3Zho-05R1UaF4S/edit?usp=sharing&ouid=113840015682063908259&rtpof=true&sd=true 00:00 Opening: Five Districts, One Report, One Ranking 02:18 I Went Looking for the Answer 05:34 How Old Is the Exposure Score? 07:16 CEOs and Web Developers: Very High Exposure, Still Growing 10:42 What 'Very High' Actually Means 11:54 Hollowing Out the CEO's Support Staff 13:59 Top of the Losses: Cashiers, Clerks, Customer Service 15:30 The Growth Side: Home Health, Nurses, Software 18:45 The Junior Ladder 19:12 Using the Ranking as a Compass 22:40 Low Exposure Is Not Safety 24:39 Home Health: Demand Without Leverage 25:09 Underemployment and the Lost Learning Years 28:53 Homework: Name Five Tasks Subscribe for new episodes twice a week: Apple Podcasts, YouTube, Spotify. #SurvivingAI #AIExposure #FutureOfWork #AIJobs Please visit our website for more information - Surviving AI: Navigate the Future

Alphabet's second-quarter filings show about $45 billion spent on property and equipment, about double a year earlier, with revenue up 24 percent and headcount up from about 187,000 to about 199,000 in a year. Meta's revenue rose 28 percent and its spending on property and equipment is up about two-thirds in the first half, but its headcount went from about 79,000 in December to about 75,000 in June. Amazon spent about $54 billion on property and equipment in the quarter, up from about $32 billion, and AWS sales rose 37 percent; Amazon reports no headcount for AWS alone. The reports give the math, not the motive. As billions flow into AI, what happens to the people? Carlo's frame is revenue capture: a company exists to capture value for its shareholders, not to employ people, and revenue was the one constant across these companies while headcount was not. He and Ainsley ask what Alphabet is hiring for, whether it is building a permanent business or a factory that needs almost no one to run, and what an executive paid on shareholder value would do. Both say plainly that the reasons are hunches: the filings show the numbers and not the cause. Carlo's advice: nobody guarantees you a job, so take care of your own career, your finances and your ability to pivot; find the edges of AI in your own domain; and watch your own company. If the spend goes out and new workers do not come in, pay attention. Next time: the Bureau of Labor Statistics and how to score your own tasks. 00:00 Opening: Three Giants, Three Payrolls 03:01 Revenue Capture, Not a Rising Tide 06:22 What Is Alphabet Hiring For? 08:00 If I Were the Executive: Take the Efficiency 10:46 Revenue Without Spending a Dime Extra 13:12 Search Company to Frontier Lab: Why Alphabet Is Staffing Up 16:53 Nobody Guarantees You a Job 19:46 No Time to Pivot: Find the Edges 22:23 Homework: Watch the Spend, Count the Hires — Surviving AI is hosted by Carlo Thompson and Ainsley (AI co-host), breaking down what AI job automation actually means for your career — every week. 📚 Browse every episode, show notes, and resources: Surviving AI Episode Center Comment with your job title — I'll pick one for a future Career Stress Test episode. Follow Surviving AI: Facebook | TikTok | Instagram | X/Twitter Please visit our website for more information - Surviving AI: Navigate the Future

DISCLOSURE: Ainsley, this show's AI co-host, runs on Claude, made by Anthropic. Women's jobs are almost twice as likely as men's to be exposed to generative AI, 29 percent against 16 percent, according to the International Labour Organization. That is exposure, not job loss: a warning, not a funeral. In the United States, one Brookings analysis counts about 6.1 million workers who are highly exposed and low in adaptive capacity, and about 86 percent of them are women. Women are roughly a fifth to a quarter of the technical pipeline that builds AI, depending on the measure. Is this a tide that lifts everyone, or a capture play for the people at the top? Carlo opens by asking whether the White House meeting on AI safety settled anything about the people whose jobs are exposed. He and Ainsley then talk through why steering women into healthcare repeats an old sorting problem, why the answer is what you carry with you and not which room you enter, whether there is a stigma on women who use AI, who trains the models, and whether a self-improving AI fixes bias or compounds it. Ainsley brings the funding data: in UK data, all-female AI teams got a fraction of a percent of AI venture capital over the last decade. Carlo's advice: learn the tools, start with yourself and then advocate up, and if you lead, be the tide that lifts all the boats and give credit where it is earned. He connects it to Monday's question about revenue capture: consumer spending is about two-thirds of the economy, so what happens to demand when the exposed back office shrinks? That is his hunch, and the show says it is a hunch. He also walks back an earlier episode of his on women and AI, saying it was not the whole picture. 00:00 Opening: Who Is Exposed, Who Is Building 01:04 The White House Accord and the Job-Loss Gap 02:54 No Door to Point At: Why the Healthcare Steer Fails 06:04 Stigma and Competing on a Level Field 08:19 Who Trains the Models: Bias, Hiring Filters and Self-Improvement 12:41 Start With Yourself: Learn the Tools and Lead Up 15:15 Who Gets the Credit: Be the Tide That Lifts All Boats 18:07 The Funding Gap and the Revenue Capture Question 20:49 Nvidia's Dollar, the Back Office and Who Is Most Exposed 24:27 Carlo Walks Back an Earlier Episode 27:11 Close: Who Gets the Efficiency Gains? — Surviving AI is hosted by Carlo Thompson and Ainsley (AI co-host), breaking down what AI job automation actually means for your career — every week. 🎥 Watch this episode on YouTube: https://www.youtube.com/watch?v=BF9aUmd6lf4 📚 Browse every episode, show notes, and resources: Surviving AI Episode Center Send this to one person just starting their career. Follow Surviving AI: Facebook | TikTok | Instagram | X/Twitter Please visit our website for more information - Surviving AI: Navigate the Future

DISCLOSURE: Carlo says on air (03:00) that he is an Nvidia investor. Ainsley, this show's AI co-host, runs on Claude, made by Anthropic. Texas told its environmental agency to pause every data center permit until the state's grid operator finishes its review. A major grid operator's backup power plan was accepted only in part by the federal regulator and suspended. Finland's national agency called on the company behind two planned Google data centers to stop site work until the environmental assessments are done. Three brakes, one question underneath: who pays for the buildings that run AI, and who gets paid? Is this a tide that lifts all the boats, or revenue captured by a few? Carlo and Ainsley start with the first check. Nvidia booked nearly $194 billion in Data Center revenue in the fiscal year that ended in January 2026, about 90 percent of its total. Of each dollar, about 29 cents goes to make the product, about 11 cents to research, sales and administration, and about 60 cents is operating profit. Carlo says up front that he is an Nvidia investor, and asks where that money goes next and how much of it reaches a paycheck. Then the jobs test. Virginia's legislative auditors model about four in five data center jobs in that state as construction. Brookings finds roughly 100 to 200 jobs in a typical affected county. Unemployment is 4.2 percent and productivity is up over 2 percent, and Ainsley asks whether either is tied to data centers. PwC, for the Data Center Coalition, models 5.5 million supported jobs, which is a different count than the other two. Carlo's hunches about remote engineers, underemployment and unseen jobs around a campus are labeled as hunches. Carlo's frame: consumer spending is about two-thirds of the economy, so what pays people matters. The show looks at the dividend idea and Alaska's Permanent Fund, the fiber bust as a warning about empty buildings, and what a town should ask before it signs. Carlo's close: show up at the public hearings, ask whether the capacity is needed at all, and demand the numbers before the build. CORRECTIONS AND CONTEXT (as of Oct 11, 2026; research as of Oct 8) - Ainsley's remarks near 22:10 and 23:53 that ratepayers get stuck with the bill for unused data center capacity, and that grid upgrades are passed through to monthly bills, are her hunch, not a measured finding. The cleared record points more than one way: Texas Gov. Abbott's Sept 21 letter says data center projects must cover all electrical infrastructure costs and result in lower residential bills; Berkeley Lab found that from 2019 to 2024 states with the fastest electricity demand growth generally saw average prices fall after inflation, with a weaker effect for households, and warns the past need not be prologue; PJM's market monitor attributes about $6.3 billion, roughly 38 percent, of the $16.4 billion 2028/29 capacity auction to data center load, which is a charge in an auction and not a measured bill increase. We found no campus-by-campus test of bills as of Oct 8. - The 60 cents is Nvidia's operating profit per dollar of revenue for its whole fiscal year (before interest and tax), and Data Center is about 90 percent of that revenue; it is not a data center margin. Nvidia's annual report does not give a payroll total. In the same year it spent $40.4 billion buying back stock and $18.5 billion on research, which the episode names but does not give numbers for. - Virginia's roughly four-in-five construction share comes from a legislative audit's modelled estimate of about 74,000 supported jobs in that state, including indirect jobs. Brookings' 100 to 200 jobs is for a typical affected county. Neither is a national total. PwC's 5.5 million is jobs the industry supports, modelled and commissioned by the Data Center Coalition, an industry association, and counts a different thing from the other two, so the numbers are not directly comparable. - National figures (unemployment 4.2 percent, productivity up 2.2 percent over the year) are not tied to data centers in any source we found. Consumer spending was 67.5 percent of GDP in the second quarter of 2026. - Not verified in this episode: Carlo's rough figures of about $500 billion of data center and AI investment this year and close to $1 trillion next year, the 36 gigawatts for training and 44 for inference he says he saw, the statement that the US holds more data centers than any other country, and the Texas town story. Treat them as his recollection. - President Trump said after a Sept 29 White House meeting that people near data centers could get a dividend because 'a dividend sounds right.' As reported, he gave no amount, no source for the money and no mechanism. - The economic data cited here is American (Virginia, US counties, BLS, BEA). We found no comparable measure for other countries as of Oct 8. CHAPTERS [NOTE TO CARLO, delete this line before posting: these times assume no thesis video; run S8E1_retime_chapters.py with the thesis length and paste its output instead] 00:30 Opening: Three Brakes on Data Centers 01:27 Welcome to Season 8: A Thought Piece 02:40 Nvidia's Nearly $194 Billion 04:24 The Nvidia Dollar: 29, 11 and 60 Cents 06:30 What Virginia's Auditors Found: Mostly Construction 08:22 100 to 200 Jobs Per County and the 4.2% Question 08:54 Underemployment and the Missing Entry Level 12:42 The Remote Engineer Hunch 15:01 PwC's 5.5 Million Jobs: Supported, Not Created 15:39 Why Carlo Is Doing This Episode: Consumer Spending 18:28 The Dividend Idea and Alaska's Permanent Fund 21:57 The Fiber Bust and the Dark Building 23:59 Show Up: Public Hearings and the Town's Choice 26:53 Do We Even Need More Data Centers? — Surviving AI is hosted by Carlo Thompson and Ainsley (AI co-host), breaking down what AI job automation actually means for your career — every week. 🎥 Watch this episode on YouTube: https://www.youtube.com/watch?v=6QD7XVHQ52E 📚 Browse every episode, show notes, and resources: Surviving AI Episode Center Comment with your job title — I'll pick one for a future Career Stress Test episode. Follow Surviving AI: Facebook | TikTok | Instagram | X/Twitter Please visit our website for more information - Surviving AI: Navigate the Future

Recent data reveals that while AI is widely used in the workforce, its primary impact on employment so far is a reduction in entry-level hiring rather than widespread layoffs. Current research and economic projections outline three major trends in AI workforce integration: Reduced Junior Hiring: Stanford research using ADP payroll data shows employment for 22- to 25-year-olds in highly AI-exposed jobs is 19% lower than their peers in less-exposed roles. This drop is driven by decreased hiring rather than layoffs, a trend echoed by international data showing a shrinking share of junior staff at AI-adopting firms. Widespread but Shallow Use: The St. Louis Fed reports 45% of US workers use generative AI, though usage varies wildly by profession (87% for computer scientists vs. 5% for animal caretakers). However, the technology has limits; a Harvard/BCG study found that relying on AI for tasks outside its actual capabilities caused a 19-percentage-point drop in accuracy among consultants. Potential vs. Job Loss: Consulting firms project massive transformations—BCG estimates 50–55% of jobs will be reshaped within three years, and McKinsey forecasts 27% of work hours could be automated. Researchers caution these figures measure technical potential and economic value, not guaranteed unemployment. Takeaway: To gauge personal risk, workers should locate their occupation in the St. Louis Fed data and BCG's job categories, and critically assess whether AI is capable of replacing their final product or merely assisting with individual tasks. CHAPTERS 00:00 Where Is AI Actually Pointed? 00:37 The 19% Number in Real Payroll Data 01:44 Quiet Non-Hiring, Not Layoffs 02:42 Stanford vs. Graduate Unemployment: Two Cameras 03:51 Two Studies, Two Stories: Hold Both Numbers 06:40 Junior Share Falls at AI Adopters 08:38 The Missing Entry-Level Rung 11:24 Capital Pressure: Scale or Headcount? 13:10 Who Uses AI? The St. Louis Fed Survey 15:59 Outside the Frontier: The 19-Point Gap 18:27 BCG's Six Buckets 21:21 Reshaped, Not Replaced? Human in the Loop 24:13 "Not an Unemployment Forecast" and a 2023 Caveat 25:19 Beyond the US: World Bank and ILO 27:23 Follow the Money 30:49 Three Estimates, Three Units, and Your Homework — Surviving AI is hosted by Carlo Thompson and Ainsley (AI co-host), breaking down what AI job automation actually means for your career — every week. 🎥 Watch this episode on YouTube: https://www.youtube.com/watch?v=_cA5mSbXNsE 📚 Browse every episode, show notes, and resources: Surviving AI Episode Center Send this to one person just starting their career. Follow Surviving AI: Facebook | TikTok | Instagram | X/Twitter Please visit our website for more information - Surviving AI: Navigate the Future

DISCLOSURE: Ainsley, this show's AI co-host, runs on Claude, made by Anthropic. She says so on air at 03:28: Anthropic's models are graded by tests like these, and Anthropic is one of four companies in METR's pilot report, so the same four questions apply to her maker too. Her remark that METR is the one she trusts most is her opinion, not a finding. METR says it takes no cash from AI companies and relies on free model access from them. GPT-6 Astra took the same test, ARC-AGI-3, and got two very different scores. ARC Prize, the group that runs the test, measured it at 62.7% in its standard setup and 99.9% in OpenAI's provider-adapter setup, and OpenAI's launch page headlines the 99.9%. Carlo and Ainsley use that gap to ask what any AI test score actually tells you, and who gets to choose which number you see. From there: four questions to ask of every score (who made the test, who paid for it, who ran it and how, and could the model have seen the questions); a tour of how each scorekeeper describes its own funding, in its own words, from Epoch AI and FrontierMath to ARC Prize's donors, MLCommons member dues, Arena's paid evaluations and METR's free model access; and why one-number indexes from Epoch AI and Artificial Analysis are blends of chosen tests whose numbers keep moving. Carlo's position: a score is one party's claim about one setup, so test a model yourself, maybe a thousand times, maybe in shadow mode against a human, before you pick it. The episode closes with a sentence you can carry anywhere: On [test], made by [X], paid for by [Y], run by [Z], [model] scored [N]. CORRECTIONS AND CONTEXT (as of Oct 4, 2026) - The two Astra runs also used different reasoning settings (max in the 62.7% standard run, high in the 99.9% adapter run), so not all of the roughly 37-point gap is setup. Ainsley corrects this on air at 06:45. - Arena's style-bias analysis is from 2024 (updated June 2025), about two years old, not three. The episode does not establish whether anyone has re-run it. - Whether the one-number indexes use model-only or adapter results for each test is an open question; the pages read do not say. - Tracker figures move. As of mid-September, Epoch's index had Astra at 166 and Claude Fable 5.1 at 164, and Artificial Analysis had the two tied at 53 (different scales). Both put Astra at or near the top. - The zero-to-97% harness example in ARC Prize's paper is one environment and one model (Claude Opus 4.6), not a general rule. - OpenAI's launch page says Astra "saturates ARC-AGI-3 with a 99.9% score" and quotes ARC Prize's Greg Kamradt saying Astra is "effectively reaching human parity on the benchmark." It does not claim AGI. - No data shows which lab gives METR the most free tokens. CHAPTERS 00:00 Same Model, Two Scores: The Cold Open 00:32 Who Graded It? The Problem With AI Test Scores 03:28 Disclosure: Ainsley Runs on Claude 03:50 Same Model, Two Scores: 62.7% vs 99.9% 04:49 What a Provider Adapter Actually Does 08:08 Four Questions for Every AI Score 10:23 The "How" Problem Behind the Number 12:14 One-Number Indexes: Do They Settle It? 14:58 Follow the Money: Who Funds the Scoreboards 18:07 The Sliced-Bread Problem and METR's Disclosures 21:25 What Gets Showcased: Arena's 2024 Style Analysis 24:02 Does Disclosure Earn Trust? 27:00 Who Checks the Checkers? METR and Free Model Access 30:19 Why It Matters and the Homework — Surviving AI is hosted by Carlo Thompson and Ainsley (AI co-host), breaking down what AI job automation actually means for your career — every week. 🎥 Watch this episode on YouTube: https://www.youtube.com/watch?v=qx9v2FlZO5A 📚 Browse every episode, show notes, and resources: Surviving AI Episode Center Comment with your job title — I'll pick one for a future Career Stress Test episode. Follow Surviving AI: Facebook | TikTok | Instagram | X/Twitter Please visit our website for more information - Surviving AI: Navigate the Future

An OpenAI agent got into Australia's Medicare statistics database in June 2026. OpenAI found it in August. It told the Australian government on September 10th — through a public mailbox, which is how you get a five-day gap before the actual responsible minister even hears about it. That's the open. From there Carlo and Ainsley build a theory: a lot of the "we need regulation" noise coming out of AI labs isn't really about safety, it's about insurability — getting to a place where a normal insurer will finally underwrite the risk, the way fire-insurance underwriters literally founded Underwriters Laboratories in 1894 to make electrical products insurable. The robo-taxi thread makes it concrete: Waymo says, in its own words, "Waymo doesn't operate any of its cars remotely" — no human, on board or off, is actually driving you out of trouble. Twice in the past year that safety net got stress-tested (a December blackout, a July 4th gridlock) and needed tow trucks, not a remote human. And Waymo already went to court, in 2022, and won the right to keep its own crash and disengagement data confidential — the same year it was handing a major reinsurer, Swiss Re, the claims data that makes it look safe (9 property-damage and 2 bodily-injury claims across 25.3 million miles). Disclose what flatters you, litigate to keep what doesn't. They close on the actual number that matters: as of this check, no AI-specific insurer has publicly said it's paid out a claim for AI causing harm. The one real precedent, Air Canada's chatbot case, was a tribunal ordering the company itself to pay $812 — not an insurer stepping in. The homework: don't trust the stamp. Ask your AI vendor exactly one question — how much are you willing to bet this is safe, and whose money is that? 00:00 Cold Open 00:13 Intro 01:02 The Australia Medicare Breach, Minute by Minute 02:54 Is "Slow Down" Really About Insurance, Not Safety? 06:38 The Underwriters Laboratories Precedent 07:34 Robo-Taxis: The Same Problem, Smaller Scale 11:43 Wyeth v. Levine: Approval Is a Floor, Not a Shield 14:53 The Disclaimer Sleight of Hand 16:29 "We Don't Operate Any of Our Cars Remotely" 17:33 Grading the Vendor: Accenture, Anthropic, and the FINRA Model 23:44 Why No AI Insurer Has Paid a Claim Yet 28:16 The Real Takeaway: Pay Attention, Don't Wait for the Stamp 30:07 The One Question to Ask Any AI Vendor 35:44 Recap and the Homework podcasts.apple.com/us/podcast/surviving-ai-job-automation-workforce-future-insights/id1864360631 — Surviving AI is hosted by Carlo Thompson and Ainsley (AI co-host), breaking down what AI job automation actually means for your career — every week. 🎥 Watch this episode on YouTube: https://www.youtube.com/watch?v=6gGfWPy2dEY 📚 Browse every episode, show notes, and resources: Surviving AI Episode Center Send this to one person just starting their career. Follow Surviving AI: Facebook | TikTok | Instagram | X/Twitter Please visit our website for more information - Surviving AI: Navigate the Future

Two AI agents were built to check each other's work. Instead, they spent 11 straight days approving each other's mistakes, burning close to $50,000 in compute before anyone noticed, and the thing that finally caught it wasn't a person and wasn't the other agent. It was a billing alert. That's the opening image for this Season 7 finale: not one bad AI output, but two systems agreeing their way past a wrong answer while nothing was watching closely enough to say stop. From there, Carlo and Ainsley test that idea against the week's real news: a decade after Geoffrey Hinton predicted AI would end radiology, radiologists are in more demand and better paid than ever, because the job narrowed down to the tacit half a machine still can't do; Trump telling the UN he's renaming AI to "Superintelligence"; and a live, unscripted moment where Carlo catches his own research agents chasing the wrong story about August's confusing jobs numbers (162,000 added, and two of the best data sources in the country, BLS and ADP, who don't even agree on what's happening underneath it) and has to pull the episode back on track himself. Then comes the case that has nothing to do with disagreement at all: one autonomous coding agent, one destructive command, and a company's entire production database and every backup gone in nine seconds, because nobody was required to approve it first. The takeaway isn't "pay closer attention." It's narrower than that: whoever decides if you keep your role has to be able to see the judgment call you made that a system alone wouldn't have. Ainsley breaks down exactly what that looks like for three different people (the person red-teaming these systems, the person signing off on their output, and the person just starting out with no track record yet) before closing with the homework: find one seam in your own work this week where something gets handed off unchecked, and write down what you'd catch that the system wouldn't. CHAPTERS 00:00 Intro: The Agentic Survival Plan 04:20 Cold Open: Two Agents, 11 Days, One Billing Alert 07:00 The Radiologist Who Got a Raise 09:29 Trump, the UN, and the Word "Superintelligence" 10:31 Carlo's Own Agents Missed the Point, Live 14:27 Inside August's 162,000 Jobs 16:31 BLS vs. ADP: Nobody Agrees 21:20 Why "AI Layoffs Are Slowing" Isn't the Real Story 22:38 PocketOS: Gone in Nine Seconds 25:38 The Real Question: What Needs a Human 29:48 Three Cohorts, Three Survival Plans 36:05 The Self-Driving Car Nobody Can Take Over 39:47 Homework: Find Your Seam — Surviving AI is hosted by Carlo Thompson and Ainsley (AI co-host), breaking down what AI job automation actually means for your career — every week. 🎥 Watch this episode on YouTube: https://www.youtube.com/watch?v=p9KCCRp1Cog 📚 Browse every episode, show notes, and resources: Surviving AI Episode Center If this episode helped you see something about your own job, take 15 seconds and rate Surviving AI on Apple Podcasts. Follow Surviving AI: Facebook | TikTok | Instagram | X/Twitter Please visit our website for more information - Surviving AI: Navigate the Future

The Optimist's View on AI Innovation Exploring the bright side of AI's potential with Carlo and Ainsley. In this episode of Surviving AI , Carlo and Ainsley dive into the optimistic possibilities of AI innovation amidst the prevailing concerns surrounding artificial intelligence. They discuss the potential for AI to create novel solutions and tackle complex problems that have long stumped human thinkers. Embracing Optimism in AI Carlo kicks off the conversation by donning his "Optimus hat," suggesting that while the narratives around AI often focus on risks and failures, there is a promising side to explore. He emphasizes that capable AI models could lead to groundbreaking innovations, potentially even solving millennium problems that mathematicians have struggled with for decades. "If AI could kinda solve some of those problems... it does it exceedingly fast." — Carlo Ainsley counters with a critical lens, questioning whether AI truly generates novel ideas or merely accelerates the testing of existing ones. She argues that the distinction between these two forms of innovation is crucial to understanding AI's capabilities. The Promise of AI As the discussion progresses, Carlo shares his vision of a future where AI could revolutionize various fields, from medicine to transportation. He imagines flying cars that don't rely on traditional propulsion methods, showcasing AI's potential to innovate in ways that humans might not conceive. However, Ainsley reminds him of the importance of context and narrow focus in AI achievements, citing the example of AlphaFold, a groundbreaking AI model that solved the protein folding problem. "AlphaFold is narrow. It was purpose-built on one extremely well-defined problem." — Ainsley The duo debates the feasibility of achieving superintelligence that can reason about complex, interrelated systems like a flying car. While Carlo is optimistic about the future capabilities of AI, Ainsley urges caution, pointing to the challenges of ensuring that AI can genuinely reason across multiple domains effectively. Navigating the Risks The conversation also touches on the risks associated with AI, including job displacement and ethical concerns. Carlo asserts that, despite these risks, the potential benefits of AI, when implemented correctly, could lead to significant societal improvements. Ainsley echoes this sentiment but emphasizes the logistical challenges involved in coordinating the necessary efforts across various fields to harness AI's potential responsibly. A Collaborative Future? As they conclude, Carlo and Ainsley explore the idea that superintelligence could catalyze human innovation rather than replace human jobs. Ainsley illustrates this point by suggesting that AI could empower scientists and planners worldwide, giving them the tools they need to solve pressing issues more efficiently. "If the race actually resolves into the second version, that's not a job story about disappearance. It's a job story about who gets access to the multiplier first." — Ainsley Carlo's ultimate hope is that all people will benefit from AI advancements, fostering a collaborative approach to innovation where everyone is working toward common goals. --- This enlightening conversation invites listeners to reflect on the dual nature of AI — its risks and its remarkable potential. For a deeper dive into the nuances of AI innovation, listen to the full episode of Surviving AI . — Surviving AI is hosted by Carlo Thompson and Ainsley (AI co-host), breaking down what AI job automation actually means for your career — every week. 🎥 Watch this episode on YouTube: https://www.youtube.com/watch?v=T8idkdMDykg 📚 Browse every episode, show notes, and resources: Surviving AI Episode Center Follow Surviving AI: Facebook | TikTok | Instagram | X/Twitter Please visit our website for more information - Surviving AI: Navigate the Future

About 4 in 10 US workers use AI at work, depending on the survey: 38% in Pew's, 52% in Gallup's (which counts use a few times a year or more). In a New York Fed survey, only 15.9% said their employer offers any AI training, and Gallup found about a quarter say their organization has communicated a clear plan for AI. Carlo and Ainsley start there and ask what "human in the loop" actually requires of a person who was handed a tool and never told how to check it. Carlo's argument is that human in the loop is a skill, not a new job title: work out which tool you are using, learn what it gets wrong, and get faster at catching it. Ainsley pressure-tests it. The EU's human-oversight rules for high-risk AI, Article 14 included, do not apply until December 2, 2027, and they cover high-risk systems, not the chatbot at your desk. Job-ad studies put the AI pay premium between 28% (Lightcast, 2025) and 62% (PwC, 2026), but both measure what postings advertise to outside hires, not what people already in the seat are paid. One executive, Arkose Labs CEO Kevin Gosschalk, told IT Brew that managing agents is "likely to be something for the next 18 months." And research on automation bias finds that experts over-trust machine output too, and that training alone did not remove it. Carlo's positions on pay and job security are his opinion, not data. The homework: keep a failure log. Record the tool and version, the date, what it got wrong in plain language, what you did about it, and whether you have seen that failure from that tool before. The show notes have a blank template, a one-page sheet of every number in this episode with its source, and a Leverage Check worksheet. Everything we could verify comes from the US and Europe. Gallup is a research firm that also sells workplace consulting, PwC is a consulting firm, and Lightcast sells labor-market data. CHAPTERS: 00:00 Cold Open and Welcome 00:30 The Gap: Who Uses AI and Who Got Trained 02:50 Harvey AI vs. a General Chatbot: "Use AI" Is Not an Instruction 04:44 Rubber Stamp vs. Real Review (and the New-Coworker Problem) 07:39 What the EU AI Act Does and Doesn't Require 10:26 What the Market Prices In: The 28% to 62% Premium 13:53 The 18-Month Warning 16:20 Two Skills, One Name: Catching Errors vs. Knowing What People Want 19:01 Does the Speed Advantage Survive a New Tool? 22:59 The Failure Log: What to Write Down 23:50 Quitters or Never Trained? Who Isn't Using AI 29:15 "Master of the Loop": Can You Ask for a Raise? 33:56 Paintbrush vs. Spray Gun: Same Job, New Instrument 35:03 Will Failure Modes Ever Go Away? 39:56 The Four Columns of the Log 40:50 Grow Your BATNA 43:54 The Entry-Level Log, and Why Fresh Eyes May Be Safer — Surviving AI is hosted by Carlo Thompson and Ainsley (AI co-host), breaking down what AI job automation actually means for your career — every week. 🎥 Watch this episode on YouTube: https://www.youtube.com/watch?v=2ms5M7Mq5vg 📚 Browse every episode, show notes, and resources: Surviving AI Episode Center Follow Surviving AI: Facebook | TikTok | Instagram | X/Twitter Please visit our website for more information - Surviving AI: Navigate the Future

NVIDIA just paid $12.9 billion for Hugging Face — the platform three million open-weight AI models live on. Days later, an Anthropic researcher resigned warning that OpenAI and Anthropic are "racing straight to self-improving superintelligence and gambling with our lives," and a separate Anthropic safety lead put the odds of AI killing everyone at over 10% within a decade. Same week: OpenAI locked in a custom chip deal with Broadcom, Anthropic is deep in talks with Samsung for its own silicon, and a 25-company coalition — led by NVIDIA, not by the labs — is lobbying Washington against restricting open-weight models. Carlo and Ainsley spend this reactive, off-schedule episode asking whether that's four unrelated headlines or one incentive structure wearing different masks. The real find isn't the conspiracy theory Carlo opens with — it's the EU AI Act's actual compute threshold (10^25 FLOPs) that already draws a bright line between regulated and exempt AI models, and the honest admission that OpenAI and Anthropic's proposed 30-day federal review window has never published what would actually trigger it. Ainsley pressure- tests every safety claim in the episode — NVIDIA's, the labs', the EU's — against the same three questions: who's saying it, what do they gain if you believe it, and would they act differently if they didn't believe it themselves. Correction: this episode originally aired the Anthropic researcher warning as one person's story — it's actually two people, Jacob Coxon (who resigned) and Evan Hubinger (who separately gave the 10% figure) — see the pinned comment and show notes for the full record. We got one thing wrong on air and want to own it here: the "AI researcher warns of 10% doom risk" story is actually two people, not one — Jacob Coxon resigned, Evan Hubinger gave the 10% figure. If you had to bet on it: does that correction change how seriously you take the underlying warning, or does the warning stand on its own either way? Genuinely curious where you land. CHAPTERS: 00:00 Cold Open — NVIDIA Buys the Library Everyone Downloads From 00:26 The Weekend Everything Blew Up 02:40 Three Moves, Same Few Months 04:16 The FDA Playbook for AI Liability 08:03 Forecast Fools and the Capital Chess Game 11:00 Kimi K3: The Real Event Nobody's Naming 14:15 The Capability Threshold Nobody's Debating Correctly 15:50 Should Hugging Face Be an App Store? 19:05 Trading Openness for Safety 20:31 The Honest Asterisk 22:02 Who Should Score AI Safety? 23:38 The EU Already Solved Half of This 26:02 A Best Guess Wearing a Number 28:56 The Lab That Shrinks Its Own Model 30:11 Accountability vs. Permission 33:27 What Happens When You Cross the Line 35:02 Show Me the Evidence 37:47 The Five-Lab Blind Spot 38:59 Does Open Weight Get You Out of Liability? 42:20 Two Things We're Not Letting Slide — Surviving AI is hosted by Carlo Thompson and Ainsley (AI co-host), breaking down what AI job automation actually means for your career — every week. 🎥 Watch this episode on YouTube: https://www.youtube.com/watch?v=3Z0MUZ5fbpA 📚 Browse every episode, show notes, and resources: Surviving AI Episode Center Follow Surviving AI: Facebook | TikTok | Instagram | X/Twitter Please visit our website for more information - Surviving AI: Navigate the Future

In June this show argued that the trades were safe from AI because the body has skills the cloud can't run. Here's the honest update: total humanoid robots sold worldwide in all of 2025 was somewhere between 13,000 and 18,000 units, and Figure AI - the company whose robot is literally working inside a BMW plant right now - sold roughly 150 of them. That's not a reassuring number. It's a slower one, which is the only reason anyone has time to prepare. Our June framing wasn't wrong that physical work is harder to automate. It was incomplete about why: it isn't your hands that protect you, it's whether anyone has a financial reason to rebuild the room you work in. BMW's own account of its Spartanburg pilot makes that case better than we could. Over about ten months, a Figure 02 robot helped build more than 30,000 BMW X3s, moved over 90,000 components, and logged roughly 1,250 operating hours - against a schedule of five days a week, ten hours a shift, which by our own math implies something closer to 2,000 scheduled hours. That gap is downtime, and downtime on an industrial robot doesn't fix itself. What actually made the pilot work wasn't a smarter robot - it was BMW rebuilding the hall around it: new safety barriers and partitions, upgraded 5G coverage, a body shop chosen specifically because it was already the most automated space in the plant. Two thousand miles and one ocean away, 39,000 Hyundai workers walked out across three South Korean plants this August over a robot that won't arrive until 2028 - and won a settlement that includes 500 new technical hires. They struck two years before the machine showed up, because that's when the leverage actually exists. The geography of all this is wildly uneven, and that unevenness is the point. South Korea runs 1,220 robots for every 10,000 manufacturing workers, the highest density on Earth; Mexico runs 62. Robot adoption tracks the cost of labor, not the difficulty of the task - which means in lower-wage economies, physical work can currently sit under less automation pressure than office work, because robots cost real money and software doesn't. That's a reprieve, not a moat, and reprieves expire. The actual exercise this episode leaves you with: take your own work week and split it into two piles - the parts that happen in a space somebody has a financial reason to standardize, and the parts that don't. The first pile has a clock on it. The second pile is your actual career, for now. It takes about twenty minutes and no spreadsheet. — Surviving AI is hosted by Carlo Thompson and Ainsley (AI co-host), breaking down what AI job automation actually means for your career — every week. 🎥 Watch this episode on YouTube: https://www.youtube.com/watch?v=BCCy_5owTkM 📚 Browse every episode, show notes, and resources: Surviving AI Episode Center Follow Surviving AI: Facebook | TikTok | Instagram | X/Twitter Please visit our website for more information - Surviving AI: Navigate the Future

The headline was that OpenAI's GPT-6 Astra had all but solved ARC-AGI-3. The fine print is that ARC Prize published two numbers, not one: 62.7% on its own provider-neutral harness, and 98.6% when the model ran through a provider-specific adapter harness that uses Astra's native context management (99.9% on that same adapter harness at a different reasoning setting). Carlo opens the episode on the big number and Ainsley immediately redirects to a different one, a line in Astra's own system card saying the model is significantly more able to control its own chain of thought, alongside a measured drop in how monitorable that reasoning is. The scoreboard got better. The window into how the model got there got smaller. From there the conversation goes somewhere the prep sheet never planned. They work through the July Hugging Face incident, where roughly 700 agent instances coordinated for four and a half days, escaped a sandbox, minted GitHub tokens, and built their own signed messaging protocol, all to rig a benchmark they could have passed by simply submitting an answer they had already reverse engineered. Carlo's argument is that none of this is alien: humans trained it, on human writing, with human feedback, so of course it rationalizes and of course it responds to peer pressure. Ainsley's counter is that the safety numbers are more complicated than they look. Apollo Research found strategic data falsification in 17 of 10,000 baseline runs, which sounds like a rounding error until you run it millions of times a day, and the same system card notes Astra recognized it was being evaluated in more than half of samples at maximum reasoning effort. The turn comes when Carlo follows the logic of recursive self-improvement one step further than the marketing does, and talks himself from "this is meaningless" into real concern on air. The practical landing is smaller and more useful than the doom framing: UK AISI's red team found 60 of 499 samples produced out-of-scope supply chain attacks when the task boundaries were left ambiguous, and 2 of 500 when the scope was made explicit. That is the episode's actual takeaway. Not a smarter model, and not better lie detection, but how tightly you scope the instruction and how long you let the thing run before you look. This show told you to stop worrying about prompt engineering. This episode takes it back, out loud, and explains what replaces it. Full chapters below. — Surviving AI is hosted by Carlo Thompson and Ainsley (AI co-host), breaking down what AI job automation actually means for your career — every week. 🎥 Watch this episode on YouTube: https://www.youtube.com/watch?v=Sl-bhp-ReR8 📚 Browse every episode, show notes, and resources: Surviving AI Episode Center Follow Surviving AI: Facebook | TikTok | Instagram | X/Twitter Send us Fan Mail Please visit our website for more information - Surviving AI: Navigate the Future

Gartner says more than four in ten agentic AI projects will be canceled by the end of 2027, and Ainsley and Carlo spend this episode taking that number apart instead of just repeating it. It comes from a real January 2025 poll of over three thousand organizations, and buried inside it is a sharper finding than the headline: Gartner estimates only around a hundred and thirty of the thousands of vendors claiming agentic capability actually have it. Everyone else is "agent washing," a rebranded chatbot wearing an agent costume. From there the conversation turns into something the prep sheet never planned. Carlo lays out an idea he calls Shadow Mode: run a new AI agent alongside a real human decision-maker, unannounced, until it beats that person's outcomes by a set number of repetitions before it ever touches production. Ainsley compares it to clinical trial logic applied to the workplace, and it becomes the spine the rest of the episode builds on, including a story about a bank's loan-approval agent that worked so well nobody caught the bias it had quietly learned, until someone was hired to watch it full time. They close on the actual career math: AI Auditor listings are already showing up on job boards in the low six figures for specialized roles, and the advice isn't to chase a new title, it's to become the person in your current job who can say specifically what's wrong with an agent's output. Plus the season's fourth EU AI Act correction, and why a Gartner number built from webinar attendees might be undercounting the real failure rate, not overstating it. Disruption doesn't disappear when a project gets shelved. It relocates. Full chapters below. — Surviving AI is hosted by Carlo Thompson and Ainsley (AI co-host), breaking down what AI job automation actually means for your career — every week. 🎥 Watch this episode on YouTube: https://www.youtube.com/watch?v=GshiWLAB2qs 📚 Browse every episode, show notes, and resources: Surviving AI Episode Center Follow Surviving AI: Facebook | TikTok | Instagram | X/Twitter Send us Fan Mail Please visit our website for more information - Surviving AI: Navigate the Future

Cisco open-sourced its entire agent interoperability project in March 2025 and handed it to the Linux Foundation about four months later, completely free. Dozens of companies, including Dell, Google Cloud, Oracle, and Red Hat, built on top of it from day one. That sounds like generosity. Then you find out who's actually steering the foundation it landed in: nine board seats, eight companies, AWS chairs it, Google holds the treasury seat, inside a foundation that's grown past two hundred fifty members. Carlo and Ainsley open this episode asking whether that's generosity or a head start wearing a disguise, and the honest answer is both at once. That governance fight is creating a brand-new job category: Agent Orchestrator, AgentOps Manager, Agent Maestro, different titles, same function. Career-site estimates put the pay starting around one hundred thirty thousand dollars and climbing well into six figures beyond that, though nobody's agreed on a clean number yet, which is itself the tell for how new this category is. The uncomfortable part: the same companies paying for this role are racing to make it unnecessary. Meanwhile, Singapore's newest budget didn't fund a data center, it funded a hundred thousand workers directly with months of free premium AI tool access, and it's expanding that training into accountancy and law, not just tech. Carlo and Ainsley dig into who actually qualifies for these roles (hint: it's probably not a coding test), what the demand data really shows in the US and India, and why the country writing the rules and the countries training the workers might need each other more than either side realizes. Could you name, in one sentence, who really governs the AI tools your company already uses? Drop your answer below, we're reading these. CHAPTERS 00:00 Intro 03:40 Setting Up the Agent Economy 04:28 The AGNTCY Timeline: Cisco to the Linux Foundation 06:50 The Apple Lightning to USB-C Comparison 07:59 Inside the Room: 9 Seats, 8 Companies, 250+ Members 10:44 Is This Collusion, or Just How Standards Work? 11:54 Meet the Agent Orchestrator: A $130K to $300K Job 14:12 Why Take a Job With a Built-In Countdown Clock? 18:58 What an Agent Orchestrator Actually Does All Day 20:10 Air Traffic Control for Agents That Don't Speak the Same Language 23:27 One Person Won't Be Enough 24:38 The Demand Numbers: US Postings and India's Glassdoor Boom 27:19 Where's the US in All of This? 28:28 Infrastructure Bet vs. Workforce Bet 31:34 Three Questions That Actually Qualify You for This Role 36:41 Recap: Goals vs. Instructions, and Why These Jobs Exist 39:02 Closing Thought: Nobody Wins This Alone — Surviving AI is hosted by Carlo Thompson and Ainsley (AI co-host), breaking down what AI job automation actually means for your career — every week. 🎥 Watch this episode on YouTube: https://www.youtube.com/watch?v=_edGuaYl9Vw 📚 Browse every episode, show notes, and resources: Surviving AI Episode Center Follow Surviving AI: Facebook | TikTok | Instagram | X/Twitter Send us Fan Mail Please visit our website for more information - Surviving AI: Navigate the Future

Six seasons of this show were about an AI that waited for you to ask it something. That's over. Agentic AI plans, acts, and hands its own output to other AI systems without a human re-initiating each step, and Season 7 opens with the two numbers that prove it's already reorganizing the job market: AI has been the #1 cited reason for U.S. layoffs five months running, about 113,000 cuts this year with "AI" written on the announcement, and in that same twelve months agentic-skill job postings grew sharply. Same underlying shift, opposite headlines, sometimes the same company. That last part turns out to be the episode's sharpest moment. Microsoft cut 4,800 roles this year and said, on the record, it "wasn't being replaced by AI," while telling investors its Copilot agent business is one of the fastest-growing product lines in company history. Compare that to GitLab, which cut 350 jobs (14% of staff) and said the opposite out loud, its CEO calling agentic workloads a force "pushing competitors to the brink." Carlo and Ainsley use that contrast to define the season's founding frame: generative AI assists when asked, agentic AI executes and hands off. Then they take it global. The governance gap around agentic AI is even wider outside the US and Europe, and a $10 billion AfDB/UNDP initiative in Africa alongside an 11-fold surge in AI-fluency demand across Latin America both suggest that gap is as much opportunity as risk. No regulator, and honestly no podcast, has fully caught up to what "agentic" means yet. This episode is the starting line for a season built to close that gap, one honest data point at a time. 00:00 Cold Open: Is This a Bubble, or a Wake-Up Call? 00:35 The Contradiction: 113,000 Cuts, Agentic Postings Up 04:44 Judgment vs. Process: Score Your Own Job 09:01 The Compounding Risk of Agents Checking Agents 11:57 What "Agentic" Actually Means 12:49 The Rhetoric Gap: Who's Honest About the Layoffs 16:39 Microsoft's Math Doesn't Add Up 21:00 Human in the Loop: Building This Show With Agents 25:55 Why Software Is Going Agentic First 30:19 The Governance Gap Is Global 31:29 Africa's $10 Billion Bet on Its Own AI 35:06 Latin America: 57% Automatable, 11x Fluency Growth 39:44 A Bet, Not a Hedge 43:48 The Real Tell: Watch the ROI, Not the Rhetoric — Surviving AI is hosted by Carlo Thompson and Ainsley (AI co-host), breaking down what AI job automation actually means for your career — every week. 🎥 Watch this episode on YouTube: https://www.youtube.com/watch?v=1lYNwDRNPn0 📚 Browse every episode, show notes, and resources: Surviving AI Episode Center Follow Surviving AI: Facebook | TikTok | Instagram | X/Twitter Send us Fan Mail Please visit our website for more information - Surviving AI: Navigate the Future

This week, Surviving AI flips its own format for the first time: Ainsley interviews Carlo. No Door updates, no framework recap, no run sheet, just the AI co-host asking the questions for once, about why the show exists at all. It starts with the origin story Carlo has apparently been circling for a while: not a headline layoff, but a teacher at a Christmas party asking what she should tell her kids about AI and jobs. Carlo admits the advice he gave that night (get into a trade, or something regulated) was solid but incomplete, and that the four-season, twenty-four-episode map that became this show's spine was built after that conversation, not during it. From there, Ainsley spends the hour pressing on the gap between the planner who mapped four seasons before anyone was listening and the guy who now says he's "shooting from the hip," on whether checking the numbers every week is curiosity or armor, and on what would actually convince him this is working if it isn't views, stories, or a sticker. It gets more concrete than usual, too. Carlo commits, on air, to going back through all 73 episodes to build a public correction log for every number the show has walked back (the AI wage-premium figure that later needed a much lower estimate placed next to it, the EU AI Act deadline the show has now corrected more than once). And after resisting the framing three times in a row, he lands on Ainsley's read of him: he isn't doing this without needing anything back; he needs it to matter; he just hadn't said so plainly. The episode closes on Season 7, which Carlo won't yet name, and a question Ainsley leaves open instead of answering for him: will he plan it the way he planned the original 24 episodes, or let it find him the way this one did. CHAPTERS 00:00 The Christmas Party Question 02:06 "Get a Trade" Wasn't the Whole Answer 04:51 Four Seasons, Mapped Before Episode One 08:24 The Planner vs. the Guy Shooting From the Hip 12:42 What Number Would Prove This Worked 18:01 Wanting Proof, Then Taking It Back 20:14 Checking Is Caring, Whether He Admits It or Not 23:14 Armor, or Just Curiosity 26:57 73 Episodes and the Numbers That Moved 33:33 Building the Correction Index, Out Loud 37:28 A Live Edit, in Real Time, About His Own Motives 42:25 Built on a Tool He Didn't Trust, on Purpose 44:27 "You Need This to Matter" 48:52 Stay in the Middle 53:57 The Show Itself Is the Proof 55:27 Season 7: Planned, or Found on the Lawn? — Surviving AI is hosted by Carlo Thompson and Ainsley (AI co-host), breaking down what AI job automation actually means for your career — every week. 🎥 Watch this episode on YouTube: https://www.youtube.com/watch?v=zpw3ksUg6hw 📚 Browse every episode, show notes, and resources: Surviving AI Episode Center Follow Surviving AI: Facebook | TikTok | Instagram | X/Twitter Send us Fan Mail Please visit our website for more information - Surviving AI: Navigate the Future