Andreessen Horowitz’s decentralized SF Tech Week 2025 was fascinating. I got to talk Agentic AI, Robotics, meet with exciting startups like Anthropic and SpacerRobotics, and even good old IBM. It was quite the venture spectacle and San Francisco showed out — it’s nothing short of thriving.

Big week this week, with the market buttressing worries about circular AI deals and trade war saber rattling. And the mystery over which and when economic statistics will be released isn’t helping anything.

Speaking of economics, we take a big swing at the economic and labor effects of AI, in response to a AI big report out of the US Senate. We hope it gives you a lot to think about.

Hey, take a look at our updated website. It features an ever growing repository of past notes and our searchable research database at epistrophy.beehiiv.com.

As always, I’m focused on three things:
1) Technology-driven change;
2) the latest in innovation and startup trends, and;
3) stock fraud.

Companies Discussed

In This Note:

• Companies Discussed

• Wall-E vs. Workers

• Tweet O’ The Week

• Epistrophy In The News

• 📆 of Epistrophy Events

• Availability This Week

Wall-E vs. Workers

AI Frights From A Senate Minority Report

Technology was supposed to make work easier. For most Americans over the last fifty years, it hasn’t. Since 1973, productivity has risen by more than 150% and corporate profits by over 370%, but real weekly wages have fallen by about $30, according to a new minority staff report from the Senate Health, Education, Labor and Pensions Committee, where Senator Bernie Sanders serves as ranking member. The productivity went up; the pay did not.

Artificial intelligence is the sharpest break yet. Unlike earlier technologies, AI compresses centuries of change into years. The Senate HELP staff warned that as many as 97 million jobs could be disrupted in the next decade. Their method—asking ChatGPT to classify which Bureau of Labor Statistics tasks are automatable—was crude. Labor economists would be right to call it imprecise. But knee-jerk alarmists who dismiss the report as hysteria miss the point. The risk is not the exact number but the scale of disruption.

AI could provide a world of abundance – perhaps the space utopia promised in the Pixar movie “Wall-E” is the best case scenario. Robots do the work, humans get all the leisure.

But the alarming breakdown in the Senate Report is enough to make you spit up your Wall-E Slurpee: it determined that  89% of fast-food jobs, 83% of customer service roles, 64% of accountants, 54% of software developers and even 40% of nurses are at risk of job loss. 

That last figure shows the error of a report that, ironically, used ChatGPT to compile its results. There is a nationwide shortage of nurses. Traveling nurses command outsized pay packages. The Bureau of Labor Statistics projects more than 200,000 annual openings for registered nurses through 2031. Machines may chart and transcribe, but they do not console families in intensive care. AI takes tasks, not whole occupations—but history shows that when enough tasks are automated, the jobs vanish. Elevator operators and typists were not eliminated all at once; they were eroded until nothing remained.

This week brought a cascade of billion-dollar AI deals. OpenAI signed a long-term supply pact with AMD (AMD: NASDAQ), securing access to its next-generation accelerators and receiving warrants for as much as a 10% stake in the chipmaker. Nvidia (NVDA: NASDAQ) countered with massive commitments of its own, tying OpenAI more tightly to its Vera Rubin platform and guaranteeing deployments measured in gigawatts of compute. OpenAI expanded its infrastructure with Oracle (ORCL: NYSE) in a $300 billion, five-year cloud deal, while simultaneously deepening arrangements with Google Cloud and preserving its ties to Microsoft. The “Stargate” consortium—OpenAI, Oracle and SoftBank—announced five new U.S. data center sites, pushing projected capacity toward 10 GW and a headline $500 billion commitment. The scale is breathtaking, but the focus is silicon, not people. If AI yields only cheaper compute and fatter margins while shoving millions into precarity, the backlash will be political and potentially existential. The opportunity is just as large: firms that tie their platforms to rising wages as well as rising productivity could secure not only dominance but legitimacy.

The risk of not doing so is documented by MIT economists Daron Acemoglu and Pascual Restrepo, who show that between 1987 and 2016 the displacement effect of automation exceeded reinstatement and that 50–70% of U.S. wage-structure changes stem from routine-task declines in automating industries (see Econometrica, “Tasks, Automation, and the Rise in U.S. Wage Inequality”).

Yet AI is not just another round of automation. Its comparative advantage is not following rules but approximating judgment. That makes it threatening to white-collar roles, but also promising as a way to spread expertise. MIT’s David Autor argues in a 2024 NBER working paper that AI can “extend the relevance, reach and value of human expertise.” Nurse practitioners, paralegals, teachers, technicians—non-elite workers could perform decision-heavy tasks once reserved for elite professionals if institutions allow it.

Evidence supports the case. In a randomized controlled experiment, programmers with GitHub Copilot completed a coding task 55.8% faster than those without the tool (Peng, Kalliamvakou, Cihon and Demirer, 2023). A study in Science found that access to ChatGPT cut time on professional writing tasks by 40%, with the largest gains for lower-skilled workers (Noy and Zhang, 2023). In real-world operations, an National Bureau of Economic Research field study of 5,000 call-center agents found AI assistance raised productivity by ~14% on average, with novices converging toward veterans. In each case, AI narrowed performance gaps—lifting the floor more than the ceiling.

The danger is stratification. The RAND Corporation estimates that $79 trillion has shifted from the bottom 90% of Americans to the top 1% since the 1970s. The Economic Policy Institute finds CEO pay has climbed to more than 330 times that of the average worker. Without intervention, AI could magnify that pattern: goods get cheaper, but wages stagnate, and the benefits concentrate at the top.

Senator Sanders proposes blunt remedies: shorter workweeks, profit-sharing, board seats for workers, robot taxes. Silicon Valley triumphalists scoff that markets alone will ensure fairness. Both ignore the history. Markets alone produced stagnant wages. Institutions—not technology—decide whether gains are shared. Nurse practitioners were created by regulation and training, not by invention. AI’s trajectory will be no different.

The real choice is not between doom and utopia but between stratified abundance and shared abundance. Both involve more output at lower cost. The question is whether workers share in the benefits or watch them accumulate in the hands of the few.

Artificial intelligence will deliver more, faster, cheaper. That much is certain. Whether it also delivers fairness is not. Alarmists are wrong to dismiss it as hysteria. Triumphalists are wrong to assume abundance trickles down. The Wall-E future—Slurpees for all, not just for the 1%—is possible. But it will not happen automatically. It will require deliberate institutions to make abundance shared, AI-related job upskilling and careful attention by policy makers to make sure that those of us not running Open-AI can also share in that abundance.

Tweet O’ The Week

Epistrophy In The News

On NewsNation with Connell McShane, we had an interesting conversation about the Senate Health, Education, Labor and Pensions Committee report (our conversation led to the report above.) The segment captured what’s at stake and, yes, in a moment of incivility I called a Congressman “dumb.” Good TV? Yes. Dumb of me? Also yes.

📆 of Epistrophy Events

Availability This Week

Available throughout the week, in an unusually hot San Francisco, and in the office for part of each day. Company meetings will take me down the Peninsula but I’ll be back by evening.

Written reports are available to clients, with video summaries on YouTube, and of course our popular summaries of the summaries on Instagram, TikTok, and YouTube Shorts.

Reach out soon! I’d love to discuss them further and, as always, comments, questions and ideas are appreciated.