CHAPTERS
- 0:00 – 0:49
2025 hot takes setup: a different kind of year-in-review
Michael and Dalton open with a playful “hottest takes” framing for reviewing 2025, riffing on the SNL joke-swap concept while explicitly avoiding intentionally cancel-worthy bits. They set expectations for a rapid-fire tour through big themes—tech, politics, and society.
- •A “2025 in review” episode with deliberately spicy opinions
- •Reference to the Colin Jost/Michael Che SNL end-of-year format
- •Positioning: provocative, but not trying to be outrageous for its own sake
- •They tee up a sequence of major topics to navigate
- 0:49 – 2:56
San Francisco narratives: ignore Twitter, and stop over-crediting politicians
They argue that most viral takes about San Francisco on Twitter—positive or negative—are noise rather than signal. Michael adds that big macro forces (COVID, downtown office collapse, policing politics) explain more than any single local leader’s choices.
- •Twitter memes about SF are more likely wrong than right
- •Viral narratives substitute for simpler macro explanations
- •COVID + empty offices + policing politics created a predictable city slump
- •Politicians have less agency than people assume in these cycles
- 2:56 – 5:09
Depopulation panic: from ‘population bomb’ to fertility decline
The conversation shifts to demographic decline and how the cultural story flipped from fear of overpopulation to fear of depopulation. Dalton recounts Stanford-era exposure to institutional overpopulation thinking and how powerful ideas can shape real policy for decades.
- •Past elite consensus: fewer kids to avoid food scarcity/overpopulation
- •Those ideas migrated into global public policy and culture
- •Rising living standards often reduce fertility without mandates
- •Powerful academic ideas can outlive their original evidence
- 5:09 – 7:10
Why incentives fail: immigration, environmental beliefs, and ineffective pro-natal policies
They note that “America is fine” fertility narratives often rely on immigration, which doesn’t solve global demographic decline. They also criticize common government responses (tax credits) as mostly ineffective, while highlighting that environmental beliefs still deter some people from having children.
- •Immigration can stabilize a country but doesn’t fix global depopulation
- •Overpopulation/environment narratives still influence personal choices
- •Government playbook is mostly tax credits; evidence suggests low impact
- •Having kids is a long-term commitment—hard to nudge with small incentives
- 7:10 – 9:06
AGI progress vs AGI hype: models improved fast, timelines feel similar
They assess whether 2025 moved us closer to AGI: Dalton sees faster-than-expected model improvements and a flood of strong releases, but not a clear shift in ultimate timelines. Michael remains skeptical, warning that AGI talk can become a viral concept that distorts founders’ planning.
- •Recent model releases improved quickly and are hard to evaluate by benchmarks alone
- •Better tools don’t necessarily imply nearer “AGI,” depending on definition
- •AGI discourse can “take all the oxygen” and warp decision-making
- •Founders may overreact to dramatic 6–18 month superintelligence claims
- 9:06 – 10:05
Self-driving as the cautionary parallel: right direction, wrong timeframe
They compare AGI timelines to self-driving: in 2015–2017 it felt imminent, yet broad reality arrived much later and unevenly. Dalton argues the utopians were directionally right but early; Michael emphasizes that time-to-delivery changes how people should live and build.
- •2015–2017 optimism: ubiquitous self-driving felt around the corner
- •2025 reality: meaningful progress, but geographically limited (e.g., Bay Area)
- •Timeframe errors matter for personal and startup strategy
- •Directionally correct predictions can still be operationally misleading
- 10:05 – 12:42
Humans in the loop: teleoperators, ML replacing rules, and what ‘solved’ means
They dig into whether self-driving is truly solved, raising questions about hidden human labor (teleoperation ratios) and the shift from hard-coded rules to ML approaches that weren’t mature earlier. The takeaway: systems can feel autonomous to users while still relying on human backstops—an analogy to future AI labor structures.
- •Waymo/Tesla experiences vs the question of “how autonomous” they really are
- •Teleoperation may scale into a major new job category while still being branded “self-driving”
- •Engineering evolution: deleting brittle rules and moving toward ML-heavy stacks
- •Defining the finish line (user experience vs true autonomy) changes the conclusion
- 12:42 – 14:13
Space in 2025: Starlink mainstream, funding surge, and ‘cheap launch’ hasn’t arrived yet
They argue that SpaceX’s success is driving an exponential drop in launch costs, unlocking new categories of space startups—similar to an “iPhone moment” for orbit-based applications. Dalton suggests we may still be early: what feels cheap now could look expensive later, with Starship potentially marking the true inflection point.
- •Launch cost declines resemble Moore’s-law-like curves
- •More startups and “space defense” attention follow reduced costs and reliability
- •Debate: are we at the ‘iPhone moment’ for space apps, or is it still ahead?
- •Starship could be the next major step-change in affordability and capability
- 14:13 – 14:43
Second-order effects: competing with SpaceX/Starlink for launches
They surface a practical downside for space startups: dependence on SpaceX can mean getting bumped from launches when Starlink needs capacity. This highlights platform risk—where the dominant provider is also a major competitor for scarce resources.
- •Startups face schedule risk when SpaceX prioritizes internal missions (Starlink)
- •Negotiating leverage is limited when the provider is the market leader
- •Launch delays can cascade into fundraising, customer contracts, and burn rate
- •Platform dependency becomes a strategic constraint for the ecosystem
- 14:43 – 16:14
Remote vs in-person after COVID: startups vs big-company reality
They conclude in-person is generally better for startups, while acknowledging COVID normalized useful tools like Zoom and chat. Dalton argues the debate was muddied because large companies become “remote” by default due to scale and distributed collaboration, whereas startups copying big-company practices was the real surprise.
- •In-person collaboration remains a startup advantage
- •COVID left lasting positive norms: Zoom acceptance and real-time chat
- •Large companies trend remote because close collaborators rarely sit together anyway
- •The real issue: startups adopting big-company behaviors they didn’t actually want
- 16:14 – 20:14
Are we in an AI bubble? Adoption is early, but infrastructure debt is the risk
Michael says “no” in the sense that AI is creating real value and adoption is still low, so impact could keep compounding even if development slowed. Dalton’s nuanced worry is debt-financed data center and GPU buildouts—echoing historical cycles where leverage amplifies crashes even when the underlying technology is real.
- •AI isn’t a hoax; value creation is real and adoption remains early
- •Risk concentrates in the infrastructure layer: debt financing for compute buildout
- •Historical analogy: fiber overbuild—society benefits while builders/financiers can get wiped
- •Potential for public market turmoil and legacy software repricing even if AI ‘works’
- 20:14 – 24:37
Abundance vs affordability: governance truth vs campaign message
They contrast technology-driven abundance (better goods, healthcare, consumer products) with acute affordability crises (especially housing). Michael argues affordability wins as a campaign slogan, while abundance is the harder, longer-term governance agenda—highlighting society’s tendency toward short-term optimization.
- •We live amid technological abundance, but many don’t feel it day-to-day
- •Housing is the standout scarcity/affordability failure
- •Affordability is politically compelling; abundance is better long-term governance
- •Humans struggle with long-term planning, yet systems can still ‘sort of work’
- 24:37 – 26:18
2026 political prediction and closing: dueling populisms and AGI jokes
Michael predicts Democrats will lean into a populist affordability narrative blaming “bad people” for high prices, paralleling Republican populism around immigration. They end on a humorous note about AI/AGI eventually producing the show, with Michael joking he’s polite to models to stay safe.
- •Forecast: competing populist messages may dominate 2026 politics
- •Concern: populism can hide underlying facts and trade-offs
- •Perspective: the U.S. has navigated worse turbulence historically
- •Wrap-up: lighthearted AGI takeover jokes and farewell
