It takes time to create work that’s clear, independent, and genuinely useful. If you’ve found value in this newsletter, consider becoming a paid subscriber. It helps me dive deeper into research, reach more people, stay free from ads/hidden agendas, and supports my crippling chocolate milk addiction. We run on a “pay what you can” model—so if you believe in the mission, there’s likely a plan that fits (over here).Every subscription helps me stay independent, avoid clickbait, and focus on depth over noise, and I deeply appreciate everyone who chooses to support our cult.Help me buy chocolate milkPS – Supporting this work doesn’t have to come out of your pocket. If you read this as part of your professional development, you can use this email template to request reimbursement for your subscription.Every month, the Chocolate Milk Cult reaches over a million Builders, Investors, Policy Makers, Leaders, and more. If you’d like to meet other members of our community, please fill out this contact form here (I will never sell your data nor will I make intros w/o your explicit permission)- https://forms.gle/Pi1pGLuS1FmzXoLr6Thanks to everyone for showing up the live-stream. Mark your calendars for 8 PM EST, Sundays, to make sure you can come in live and ask questions.Share^^Bring your moms and grandmoms into the Chocolate Milk Cult.Martin Ford is a futurist and author who has spent nearly two decades studying how AI and robotics reshape the labor market, starting with his first book in 2009, well before job displacement was a mainstream economic debate. In Rise of the Robots: Technology and the Threat of a Jobless Future, Ford argues that AI and robotics will eventually displace most human workers. That’s the question this livestream sits on top of: how would that impact the economy, and can we adapt if it happens. The updated edition, out June 2nd of this year, folds in everything since ChatGPT.Book: Rise of the Robots X: @MFordFutureManish Gupta is a Principal Applied Researcher at Microsoft India R&D Private Limited at Hyderabad, India. He is also an Adjunct Faculty at IIIT, Hyderabad, and a visiting faculty at ISB, Hyderabad. His YouTube channel has a ton of fantastic deep dives on ML papers. Manish’s work combines his academic expertise with his real-life experiences to provide really insightful deep dives into some overlooked papers.If you’re doing interesting work and would like to be featured in the spotlight section, just drop your introduction in the comments/by reaching out to me. There are no rules- you could talk about a paper you’ve written, an interesting project you’ve worked on, some personal challenge you’re working on, ask me to promote your company/product, or anything else you consider important. The goal is to get to know you better, and possibly connect you with interesting people in our chocolate milk cult. No costs/obligations are attached.This guide expands the core ideas and structures them for deeper reflection. Watch the full stream for tone, nuance, and side-commentary.The Event — Martin Ford, futurist and author of Rise of the Robots (2015, updated edition out June 2nd), opened by tracing 17 years of his own thesis. In 2009, warning that automation could eliminate jobs got you labeled a Luddite by economists. Today those same economists are studying the question directly, running labor statistics against the LLM rollout, and so far not finding a measurable disemployment effect. Ford’s read: everyone is measuring the wrong variable.Why this reframes everything — The debate about AI and jobs keeps hitting a category error: people ask whether AI can do a task, when the real question is whether it can do the job. A job is a bundle of tasks held together by judgment, coordination, and the capacity to get better at things you don’t already know how to do. Ford’s example is a fresh graduate: nearly useless on day one, competent within six months. The entire gap is on-the-job learning — trying something, watching it fail, absorbing the correction, doing it slightly better tomorrow. That loop is the actual moat protecting most white-collar jobs, not raw capability. Frontier models ship frozen: whatever they know at release is what they know until the next training run, months out, at enormous cost. There’s no six-months-in version of a model quietly getting better at your specific job the way a new hire does. Ford treats this as the one variable worth tracking — the moment a model can learn continuously and generalize outside its training distribution, the standard defense that AI automates tasks but not whole jobs stops working. He frames it as a condition for the “AI complements workers” era ending, not a prediction that AGI itself arrives.The Event — Dev separated the AGI question from what’s already automatable, using a burger-flipping robot: technically buildable, but the edge cases in a fast-food kitchen are numerous enough that the ROI doesn’t justify the engineering spend. A coding agent is expensive to run well but clearly worth it, given the volume and value of software work it displaces. Ford generalized the pattern: the deciding factor isn’t blue-collar versus white-collar, it’s whether a job is routine, predictable, and verifiable against a large dataset of past examples done correctly.Why this reframes everything — The old intuition was a skill ladder: physical labor is safe from software, cognitive labor is the target, and within cognitive labor, sophisticated work is safer than rote work. That model is already wrong on both axes. On the physical side, Ford points to Amazon warehouses: robots don’t need to solve the general chaos of the real world, they need a controlled environment where the messy variable — people moving unpredictably — is removed. That’s a narrower engineering problem than self-driving cars, which operate inside exactly the chaos warehouses design away, and it’s why warehouse robotics is advancing faster than autonomous driving despite both being “physical AI.” On the cognitive side, a financial analyst doing routine quantitative work is more exposed than a plumber, because the analyst’s task is narrow and repeatable with thousands of historical examples to train and verify against, while the plumber’s job is soaked in the unbounded, case-by-case variation that breaks pattern-matching systems. The filter that predicts automation risk: how much historical data exists to train on, and how cheaply an output can be checked against ground truth. Job title, salary, prestige, years of schooling — all noise.The Event — Ford introduced a mechanism most of the conversation’s later arguments hinge on: even short of full job elimination, partial task automation triggers headcount consolidation. Take three people doing broadly similar work. Once half of each person’s task list gets automated, management doesn’t keep three people at half productivity. They merge what’s left into one or two roles.Why this reframes everything — Ford reaches for electrification as the template. It took decades for factories to gain any productivity from electric motors, because early factories replaced the steam engine with an electric one and kept the same centralized layout: the shaft-and-belt system built around a single power source. The gain only showed up once factories redesigned around distributed motors, one per machine — an organizational change, not a technology swap. Ford expects AI to move faster than that multi-decade lag, but the shape of the delay is the same: the bottleneck is organizational restructuring, not capability. It’s also why economists currently finding no disemployment effect yet aren’t wrong. They’re measuring during the shaft-and-belt phase, before most organizations have reorganized around what’s already automatable. The visible layer right now is what Ford calls bottom-up adoption: employees using the tools on their own initiative, finishing work faster, and quietly keeping the slack time for themselves. Consolidation happens once management notices and reorganizes formally around the capability instead of the old task list.The Event — Ford drew a distinction people usually collapse into one question. Everyone repeats that ATMs didn’t eliminate bank tellers, which is true but beside the point: what actually cut teller employment was mobile banking, letting the customer do the transaction themselves with no employee in the loop. Dev connected this to Irys: some users get AI-assisted verification that compresses a task they’d already pay a professional for, like a faster contract review because a first pass flags what to check. Others get access to something they’d never have paid for at all, since the old cost of even a basic legal review priced them out entirely. Getting them from zero informed decision-making to a serviceable one, for free, is a different kind of gain.Why this reframes everything — These aren’t the same displacement mechanism, and conflating them produces sloppy predictions. The first is direct substitution: technology does what an employee used to do. The second is self-service enablement: technology lets the customer route around the employee entirely, eliminating the role by making the human intermediary unnecessary rather than by replicating their output. Ford’s point is that both run in parallel inside the same organization. A company can use AI to make its analysts faster, which drives the consolidation from Section 3, while also building a self-service tool that lets a manager query the AI directly and skip the analyst altogether. Dev’s contract-review example is the self-service case at the consumer layer: nobody would previously pay $500 for a lawyer to read a routine agreement, so that market didn’t exist and no lawyer’s job was ever at stake there. Once the marginal cost of a good-enough first-pass review drops near zero, a market appears where there wasn’t one, and professional services shift toward many small transactions plus a shrinking core of large ones. Whether an industry sees mass job loss or this kind of bifurcated restructuring depends on which mechanism dominates. Regulated professions are the test case for why, covered next.The Event — Dev pushed on why radiology and law haven’t been automated despite years of confident predictions. Ford’s answer centered on liability, not technical capability: a radiologist who misses a cancer diagnosis can be sued for malpractice, and that legal exposure is load-bearing in the current system. There’s no clean equivalent once the decision-maker is a model instead of a licensed person.Why this reframes everything — The unspoken assumption behind most “AI will replace doctors and lawyers any day now” takes is that the bottleneck is model accuracy: get the diagnostic model to radiologist-level performance and the job dissolves. The real constraint is that the entire liability architecture of medicine and law is built around a human who can be individually sued, disciplined, or stripped of a license. Nobody has built the equivalent for a company whose model makes an error across thousands of patients at once, or figured out what standard applies when the error is systematic rather than one practitioner’s lapse. The useful heuristic here: the relevant axis isn’t “regulated versus unregulated,” it’s how catastrophic and attributable a single error is. Booking your own travel or taking AI meal suggestions carries close to zero downside if the model’s wrong, so self-service automation faces no liability wall there. A misdiagnosed tumor or a botched contract carries severe, attributable downside, so an accountable human stays in the loop until someone builds a liability framework that can absorb the systematic-error case. That’s a legal and institutional problem, not a model-scaling one, and better benchmarks won’t solve it.The Event — Ford named the default policy response to worker displacement: retrain, upskill, send people back to school. Then he explained why it doesn’t transfer here. The historical template assumed machines took the routine, lower-skill work while humans climbed a ladder toward more sophisticated tasks that stayed out of reach. Ford’s point is blunt: AI is currently more capable at the sophisticated, credentialed end of that ladder than at a lot of blue-collar physical work. That inverts the assumption the entire “upskill your way out of it” policy rests on.Why this reframes everything — If the ladder itself is being automated from the top down, “climb higher” isn’t a strategy. It’s a description of moving toward the part of the economy under the most pressure. Ford’s book has always argued for universal basic income as the least-bad solution, not an ideal one, because the retraining paradigm depends on a next rung that stays human-only long enough to retrain into, and that assumption is exactly what’s failing. He’s equally direct about UBI’s failure mode: strip out incentive design and you get a real dependency problem, sharpest for teenagers who see no reason to finish high school if the payout is identical whether they graduate or drop out. His proposed patch tiers the basic income higher for people who complete school, keep learning, or do recognized community work, preserving the incentive a paycheck used to provide without requiring an actual job to exist. Ford treats a second point as almost as important as the income question: a job isn’t just money, it’s identity and a socially legible answer to “what do you do.” A policy response that solves income while ignoring that solves half the problem.The Event — This is where the conversation shifted from Ford’s established territory into a live disagreement with Dev. Ford’s model of economic collapse from AI is a demand-side story: consumers are the ultimate source of all revenue, even for a company like Boeing that never sells to an individual, because somewhere downstream an airline only buys planes if consumers are buying tickets. If AI concentrates income into fewer hands, the median consumer’s spending power collapses, and the whole chain down to Boeing eventually feels it, regardless of how much wealth exists in aggregate. Dev pushed back that this framing undersells a faster-moving problem: asset and perception-based wealth. His example was Elon Musk’s SpaceX IPO, a valuation surge driven not by new revenue or margin but by index funds getting structurally forced to buy the stock through rule changes at CRSP, Nasdaq, FTSE Russell, and S&P. That paper wealth converts into real political power. Musk can borrow against it, spend it, and by Ford’s own admission, plausibly influence a presidential election with it, without the underlying business having created any additional value at all.Why this reframes everything — Ford held the line that these are two distinct problems, and neither substitutes for the other. Wealth concentration and political capture are real and will worsen with AI, but they’re separable from the income-concentration mechanism that actually breaks consumer demand and triggers a recession-style spiral. That distinction matters because the two problems call for different interventions: antitrust and campaign finance reform address the first, income redistribution addresses the second. Collapsing them into one narrative risks solving neither well. Dev’s extension is the sharper move: once a startup’s real audience becomes the investors who set its valuation rather than the customers who use its product, decision-making warps toward whatever appeals to that audience. That’s the benchmark-maxing, the researchers privately admitting they’re chasing metrics they don’t believe in, the institutional aversion to shipping anything that could dent valuation even if it would help users. Guy Debord’s society of the spectacle names this exactly: under enough media saturation, the packaging and perception of a thing displaces the thing itself as the product. It’s a dynamic anyone inside a modern AI lab has felt but rarely has vocabulary for. The appearance of progress toward investors becomes a competing objective against actual progress toward users, and the two increasingly diverge without anyone deciding they should.Ford’s calibration for how early this still is: the iPhone launched in 2007, and roughly three and a half years after ChatGPT’s release puts us at the rough equivalent of 2011 in smartphone years. Nobody in 2011 could have named social media’s effect on attention or fertility as a coming consequence of the device in their pocket. His own track record backs the caution. Writing the original Rise of the Robots in 2015, right after deep learning’s 2012 breakthrough in image recognition, he expected a Turing-test-passing system 20 to 30 years out, and expected whatever automation arrived to come from narrow systems built for a single domain like accounting. Instead we got one general model fine-tuned into a hundred vertical applications — exactly the pattern Iqidis and every other applied-AI company now runs on. Ford is the one person in this conversation who’s already had a forecast graded, and he undershot it by a wide margin.Subscribe to support AI Made Simple and help us deliver more quality information to you-Flexible pricing available—pay what matches your budget here.Thank you for being here, and I hope you have a wonderful day.Dev <3If you liked this article and wish to share it, please refer to the following guidelines.ShareThat is it for this piece. I appreciate your time. As always, if you’re interested in working with me or checking out my other work, my links will be at the end of this email/post. And if you found value in this write-up, I would appreciate you sharing it with more people. It is word-of-mouth referrals like yours that help me grow. The best way to share testimonials is to share articles and tag me in your post so I can see/share it.Use the links below to check out my other content, learn more about tutoring, reach out to me about projects, or just to say hi.Small Snippets about Tech, AI and Machine Learning over hereAI Newsletter- https://artificialintelligencemadesimple.substack.com/My grandma’s favorite Tech Newsletter- https://codinginterviewsmadesimple.substack.com/My (imaginary) sister’s favorite MLOps Podcast-Check out my other articles on Medium. :https://machine-learning-made-simple.medium.com/My YouTube: https://www.youtube.com/@ChocolateMilkCultLeader/Reach out to me on LinkedIn. Let’s connect: https://www.linkedin.com/in/devansh-devansh-516004168/My Instagram: https://www.instagram.com/iseethings404/My Twitter: https://twitter.com/Machine01776819