Divya van Mahajan

Something Big Is Happening

ai-daily-brief-podcast

Overview

This episode of the AI Daily Brief (recorded around February 16, 2026) is a deep-dive reaction to a viral essay titled “Something Big is Happening” by Matt Schumer, CEO of an LLM startup, which garnered approximately 80 million views on X (formerly Twitter). The host synthesises Schumer’s original argument alongside several notable responses and critiques, exploring the thesis that AI has crossed a meaningful capability threshold with profound implications for knowledge work across all industries. The speaker does not identify themselves by name in the transcript.

Source video URL: not provided

Prerequisites

  • Basic familiarity with large language models (LLMs) and AI assistants (e.g., ChatGPT, Claude)
  • General awareness of recent AI model releases (GPT-series, Claude/Opus series)
  • Some understanding of software development concepts (code generation, agentic workflows)
  • Familiarity with debates around technological unemployment and historical analogies (Industrial Revolution, etc.)
  • Awareness of researchers/commentators referenced: Ethan Mollick, Derek Thompson, Gary Marcus

Main Points

The Core Claim: A Qualitative Shift Has Already Occurred

  • Matt Schumer uses the COVID-19 pandemic as an analogy: most people were not paying attention before the world suddenly changed; he argues we are in the “this seems overblown” phase of something far larger.
  • Schumer, with six years of AI startup experience, states he is not making predictions — he is describing what has already happened to him and his colleagues.
  • He reports that as of early 2026 (following the release of GPT-5.3 Codex and Opus 4.5), he can describe a software project in plain English, leave for four hours, and return to find finished, production-quality work requiring no corrections.
  • The AI now autonomously tests its own output — clicking through an app, identifying issues, iterating, and only presenting the result when it meets its own standards.

Why Coding Was Targeted First — and Why That Matters to Everyone

  • AI labs deliberately prioritised coding capability because writing code accelerates the development of the next, more capable AI model — a recursive improvement loop.
  • GPT-5.3 Codex was described in its release notes as “instrumental in creating itself.”
  • The disruption to software engineers was a side effect of this strategy, not the end goal; labs have now “done it” in code and are moving on to other knowledge domains.
  • Affected sectors named include: law, finance, medicine, accounting, consulting, writing, design, analysis, and customer service — with timelines estimated at 1–5 years, possibly less.

Addressing the “I Tried It and It Wasn’t Impressive” Objection

  • Schumer (and the host, echoing Ethan Mollick) notes a significant capability gap between free-tier AI products and top-tier paid models.
  • The analogy used: judging the state of smartphones by using a flip phone.
  • The time when AI was genuinely unimpressive is described as “ancient history” relative to current model capabilities.

Schumer’s Practical Advice: What to Actually Do

  • Use AI seriously: Obtain paid access; use the best available model; apply it to hard, substantive tasks — not just as a search engine.
  • Treat this as a pivotal career moment: The person who completes three-day analyses in an hour using AI is immediately the most valuable person in the room; the advantage is time-limited as adoption spreads.
  • Abandon ego: Those who refuse to engage — dismissing AI as a fad or believing their field is immune — are most at risk.
  • Build adaptability as a habit: The specific tools matter less than developing the capacity to learn new ones quickly; the models available today will be obsolete within a year; workflows will need to be rebuilt repeatedly.
  • Engage with curiosity and urgency, not fear.

The Critique from Isaac Saul: Coding Is a Special Case

  • Saul argues that code is a highly structured, pattern-rich domain and that software engineers systematically over-generalise AI’s coding proficiency to all other work.
  • Much human knowledge work involves disorder, unpredictability, and irreducible human judgment — e.g., working a journalistic source over years, reading a jury, building a client relationship.
  • He references the Good Will Hunting “Sistine Chapel” scene as a metaphor: raw pattern-matching intelligence does not substitute for embodied, experiential understanding.
  • The host finds this critique partially valid but rejects the conclusion that it “discredits everything else.”

Will Meneides’ “Tool-Shaped Objects” Essay — and Its Rebuttal

  • Meneides uses the analogy of a centuries-old Japanese hand plane (chiazuru) — beautiful, technically extraordinary, but economically superseded — to introduce the concept of a “tool-shaped object”: something that produces the feeling of work rather than actual value.
  • He applies this to both Schumer’s essay (implying it was AI-generated and functions as engagement-bait rather than genuine insight) and to LLM deployments more broadly.
  • His central claim: AI is “everywhere in consumption and almost nowhere in output”; spending on AI systems primarily produces the experience of spending, not real economic value.
  • He concedes AI may eventually diffuse into the real economy but argues this will take “much, much longer” than AI boosters suggest.
  • Host’s rebuttal: The agent-workflow critique is actually a critique of the value of knowledge work in general (the “TPS reports” problem), not of AI specifically. Calling AI fake tools because they serve work of questionable value is a non-sequitur.
  • Jacob Franek’s response, cited approvingly: Meneides used many words to say “AI adoption won’t happen as fast as some believe” — itself a “nothing statement.” Meanwhile, Anthropic executives state 100% of their code is written by LLMs, representing real, shipped output.

Ethan Mollick’s Framing: Two Symmetric Errors

  • Mollick identifies two equally common mistakes:
    1. Vastly underestimating what AI can do and the scale of inevitable workplace impact with today’s models.
    2. Underestimating the real-world friction involved in actually extracting value from AI systems.
  • The host argues the asymmetry of costs matters: the cost of overestimating AI (some over-preparation) is far lower than the cost of underestimating it (potential professional extinction).

Connor Boyack’s Response: The Seen vs. the Unseen

  • Boyack invokes Frédéric Bastiat’s 1850 distinction between the “seen” effects of economic change (visible job losses) and the “unseen” effects (new industries, new possibilities, unlocked creativity).
  • Historical examples: the knitting machine, the power loom, and the computer each appeared to threaten destruction but ultimately expanded economic activity vastly.
  • The “seen” from AI: jobs automated, tasks displaced. The “unseen”: work that becomes possible only when costs drop, creative output unlocked when drudgery disappears, solo entrepreneurs who can now build what formerly required teams of twenty.
  • The real risk is not AI but a fixed-pie mindset — staring at the seen and missing the explosion of opportunity forming outside one’s field of vision.
  • The host endorses this view but adds one caveat: optimism about long-run resolution does not eliminate the possibility of “utter carnage in the liminal period of transition.”

Key Concepts

  • Something Big is Happening: The title of Matt Schumer’s viral essay (~80 million views) arguing that AI has crossed a capability threshold already experienced in the tech industry that is about to propagate across all knowledge work.
  • Agentic AI / Autonomous AI workflow: AI systems that do not merely respond to prompts but autonomously plan, execute, test, and iterate on complex tasks without ongoing human guidance.
  • Tool-shaped object: Will Meneides’ term for something that produces the feeling of work or utility without generating real output value; applied both to AI deployments and to the Schumer essay itself.
  • The seen vs. the unseen: Bastiat’s economic principle (1850), as applied by Connor Boyack, distinguishing visible first-order effects of technological disruption (job losses) from harder-to-see second-order effects (new industries and opportunities created).
  • Fixed-pie mindset: The assumption that there is a fixed quantity of work in the world, such that AI performing more of it necessarily leaves less for humans — a framing the host and Boyack both reject.
  • Capability gap (free vs. paid AI): The significant performance difference between publicly available free-tier AI models and top-tier paid models, which causes many users to underestimate AI’s current capabilities.
  • Recursive self-improvement loop: The dynamic in which AI capable of writing high-quality code can contribute to building the next, more capable AI model, accelerating the pace of progress.
  • Liminal transition period: The host’s term for the potentially disruptive intermediate period between the old economic order and a new equilibrium, even if the long-run outcome is positive.
  • Meter autonomy study: An ongoing study tracking the degree of autonomous capability demonstrated by AI systems over time, referenced by Schumer to illustrate accelerating progress.

Summary

The episode centres on the viral essay “Something Big is Happening” by Matt Schumer, which argues that AI crossed a meaningful capability threshold in early 2026 — one already experienced by technology workers — and is now propagating rapidly toward all knowledge work professions, with timelines of one to five years or less. The host uses Schumer’s essay as a springboard to survey the wider conversation it provoked: Isaac Saul’s reasonable caution that coding is a uniquely structured domain and that AI’s success there may not generalise cleanly to more human-centric work; Will Meneides’ more dismissive “tool-shaped objects” argument (which the host largely rejects as clever condescension that inadvertently critiques knowledge work in general rather than AI specifically); and Connor Boyack’s Bastiat-informed reminder that technological disruption always produces unseen opportunities that doom-focused analysis misses. The host’s own position is that the asymmetry of risks strongly favours taking AI seriously and adapting early — the cost of over-preparation is trivial compared to the cost of being professionally displaced — while acknowledging that even an ultimately positive long-run outcome does not rule out significant disruption in the transition period. The episode concludes that whatever one thinks of Schumer’s specific claims, the public conversation his essay generated is itself valuable, and the most important question anyone can ask right now is: what is this making possible?