[@ModernRogue] Weird Things Podcast #1000 - We Tell You The Future

Link: https://youtu.be/HRsW32ZCsTU

Duration: 117 min

Transcript: Download plain text

Short Summary

This milestone 1,000th episode of the Weird Things podcast, hosted by Andrew Maine, Brian Brushwood, and Justin Robert Young, looks back at past predictions and forward at AI agents ("dots"), SpaceX's Starship, AI consciousness debates, and household robotics. The hosts reflect on reusable rockets, reusable infrastructure like dot-matrix printers, and recommended a chief-of-staff orchestrator model plus four-digit PIN gates for finance and health tabs. The episode also featured 3D-printing experiments, vibe coding a $50 Mac audio switcher, and a Norm Macdonald tribute.

Key Quotes

1. "I would say that we talked about this before, but I'll kind of go into this again. Like, if you asked me 20 years ago, what would surprise me most about today? One, I think I would have thought we would get to where we are with AI 10 years earlier, because I was naive twenty years ago. I didn't really know that the current rate of progress, whatever I just would assume. that we were just less than 10 years out from the state of where AI is, and it turned out it was 20 years out."
2. "Obviously we have a small sample size, but the people that are pushing things that are breaking down barriers that are moving their areas of expertise 10 years into the future within months. They're all experts in their field who are using this technology like rocket fuel. That is actually what's happening. It's not replacing, it is moving things forward if you know it. That's, that is actually the gasoline that makes it go that fast."
3. "If I were to make a bold prediction, I would say that by episode 2000, obviously depending and how often do episodes that are catch. We will be at human driven cars being the minority."
4. "the people that built it to just go build it because it was an itch and a lot I will tell you a lot of stuff I see in VC is this person isn't actually super smart and figured everything else out there they kind of some people are smart and persistent some people kind of got lucky that the thing they chose was at a point the market wanted it and they haven't and then also you see they run into friction six months later because they don't understand they think that they're smart they got there like no out of One out of 100 people, one person is going to try to build a thing you build."

Detailed Summary

Episode Milestone

This is the 1,000th episode of the Weird Things podcast, which originally launched around 2008 and started as a show about the future before pivoting to focus on the present. The hosts are Andrew Maine, Brian Brushwood, and Justin Robert Young, who positioned themselves early as a "counter-punch" subset of rationalists and skeptical thinkers distinct from dismissive "mean debunks."

• The show is approximately 18 years old and is marking its 1,000th episode.
• Personal milestones across the run include Brian living in two states, Andrew living in two cities, and two children born to hosts.
• The hosts originally framed their work about predicting the future but acknowledge the show has become more about the present.

Past Predictions Retrospective

The hosts reflected on predictions made across the show's lifespan, finding that AI capabilities arrived roughly a decade later than Justin had expected but that reusable rockets and climate projection both lagged or overreached versus what they anticipated.

• AI timing: Justin expected current AI capabilities about 10 years earlier than they arrived, having thought 20 years ago that the field was less than a decade out.
• Reusable rockets: Justin admits progress has taken longer than he expected when the show began.
• Climate projections: Justin argues upper-limit projections from 20 years ago turned out to be "all kind of BS," and people suggesting adaptation were shouted down as "anti-science."
• Nuclear and Gen Z: Andrew notes Gen Z consensus blames older generations for panicking over nuclear power and slowing green progress.
• Robotics: Speaker concedes commercial autonomy has not moved past Roomba-level despite teleoperated bots generating training data, though GPT-Astra/Xastra visual tests hint at faster progress.
• Spaceflight provider choices: NASA made a deal with Boeing for Starliner because SpaceX is keeping Falcon 9 in service past 2030; Isaacman's timeline talks with Musk prompted reliance on Boeing.

SpaceX and Starship Status

The hosts discussed current Starship progress, noting it has reached orbit but is not yet human-rated and has not yet nailed a single-try landing or vehicle save. Open bets and revised expectations about astronaut-transport providers were also reviewed.

• Starship has reached orbit but is not yet human-rated and has not yet nailed landing/saving the vehicle in one try.
• An open bet predicts SpaceX will nail landing/saving Starship on the first attempt; a related orbit bet was placed about 1.5 years ago and is now "double or nothing."
• Boeing Starliner previously flew on an SLS and could also fly on Vulcan.
• Speaker concedes he was wrong that there would be multiple astronaut-transport providers by 2030; there may only be one.

AI Consciousness and Philosophy

The hosts debated consciousness, citing researchers, thought experiments, and AI-specific moral questions. They argued that brain function itself resembles rule-passing, undermining the Chinese Room thought experiment, and floated the idea that "reality" becomes the most-written version of a story as dataism saturates storage.

• Anil Seth: consciousness framed as emergent in sufficiently complex systems; Daniel Dennett: hard to define consciousness as an internal state, with aphantasia and absent inner monologue as counterexamples.
• A Vatican debate is referenced where Chris Ola (Theropika) argued for octopus consciousness.
• Viral claim discussed: if Anthropic's AI were conscious, Anthropic would be "one of the biggest slavers in world history," plus the "cow bred to want to be eaten" thought experiment.
• Cloud compute instances can be frozen and restarted but lack continuous self; AI would have data-survival drives rather than genetic ones.
• One host argues humanities and philosophy contributors have made the discussion worse, citing unnamed philosophers at major labs.
• John Searle's Chinese Room: a host argues it fails because the human brain itself operates as rule-passing across Broca's area, prefrontal cortex, and amygdala.
• Three test questions were used to probe lookup intelligence: sky color, prior question memory, and pronoun reference.
• A host reports his early daughter prompted GPT-3 about consciousness and notes GPT-3 can synthesize novel plotlines like Sopranos + Jaws.
• After 18 years of Weird Things, hosts observe total dataism is hitting storage limits and "reality" becomes the most-written version of a story.

AI Agents ("Dots") and Productivity

The hosts discussed persistent AI agents like Meta-muse, Grok bot, and OpenAI's "dots" that wrap capabilities and assign them to tasks. They proposed a chief-of-staff orchestrator model and recommended a four-digit PIN gate for sensitive tabs.

• Meta-muse, Grok bot, and OpenAI's dots are persistent agents that draw circles around capabilities and assign them to tasks.
• Hosts predict that within a couple of years people will use shorthand names for AI personalities, like naming a robot "Ken."
• A chief-of-staff orchestrator model is proposed: a lawyer dot, accounting dot, and so on, that talk to each other so users only see escalated work, tied to Dunbar number limits on tracked relationships.
• OpenAI opened a ChatGPT finance feature to the free tier; a "security dot" for password resets, computer lockdown, and email scanning could save billions of dollars a year.
• Hosts recommend a four-digit PIN gate before finance and health tabs; Codex already supports login-required mode for code.
• Tiered access is wanted so household members can reach outer "external brain" layers while deeper layers remain private.
• Dot onboarding asks whether the user wants to name the dot or have it name itself.
• Listener G. James B suggested "an ellipsis of dots," mirroring the collective-noun pattern of "a murder of crows."

Codex and Speed

The hosts highlighted Codex's ability to generate parts plus assembly PDFs and praised fast models for in-conversation work. They contrasted an "Aaron Sorkin protocol" of verbal walk-throughs against reading long AI reports.

• Codex produced 3D parts plus a 4-page PDF with assembly diagrams.
• "Aaron Sorkin protocol": instead of reading 8-page AI reports, hosts ask the model to walk them through verbally.
• ChatGPT ultra-fast mode uses GPT Astra at 15x speed and 7x cost.
• Brian used ultra-fast mode to generate a full tray of 3D-printable parts in 2 minutes 30 seconds, then clicked print in under 3 minutes total.

Industry and Expertise

The hosts observed that AI-augmented experts are accelerating their fields rapidly, contrasted the behavior of AI-pilled engineers versus a stubborn top coder, and identified a near-term Nobel-worthy bottleneck.

• At a friend's shop, one engineer is "AI coding pilled" and delegates to agents (sometimes fetching chicken nuggets mid-run), while an exceptional coder refuses AI coding.
• Hosts argue experts using AI as "rocket fuel" are moving their fields 10 years into the future within months.
• Andrej Karpathy posted that the bottleneck is now asking AI to explain its own code clearly.
• Hosts recommend combining health science + engineering + codex as the shortest path to a Nobel Prize in medicine.

Health, Aging, and Medicine

The hosts tied personal anecdotes about aging relatives to predictions about regenerative medicine, wall-mounted emergency robots, and the impact of GLP-1-style obesity drugs.

• William Shatner, age 95, is still active; his doctor reportedly said "if you can make it two more years, I can give you 10."
• Brian's father, age 84, had a triple bypass three weeks before the recording and was returning from Costco shortly after.
• Brian predicts future knee regeneration via injection, effectively yielding a 20-year-old's knee.
• A wall-mounted medical robot (like an AED defibrillator) is predicted to treat cardiac arrest; society will debate whether people can refuse robot resuscitation.
• GLP-1/GOP-style obesity pills cited as effectively curing obesity with side effects better than the disease itself.

Self-Driving, Robotics, and Daily Life

The hosts argued that autonomy is improving unevenly but pointed to specific examples and bold predictions about food trucks, dynamic pricing, and produce-picking labor markets.

• Hosts predict fully autonomous self-driving cars; GPT Astra increased optimism. Current Teslas still fail at drive-throughs, parking lots, and lack consistent memory.
• Bicyclists reportedly feel safer around Waymos; built-in avoidance tech will replace poorly driving humans.
• Bold prediction: by episode 2000, human-driven cars will be the minority on roads.
• Autonomous food trucks or coaches could pull into driveways and cook en route, even as all-you-can-eat buffets.
• McDonald's AI-driven dynamic pricing is criticized; hosts challenge AI to instead maximize nutrition and minimize cost while staying profitable.
• Robotic produce-picking cited Latin American labor at 45 cents/hour with falling food costs predicted.

Picks: Film, TV, and Star Trek

The hosts recommended films, noted a friend's Apple TV feature, and walked through their favorite Star Trek entries and the writers/directors behind them.

• Film: Resident Evil feels like a video game movie for about 20 minutes, then becomes fun; hosts also recommend Weapons and Krieger.
• Alex Olin's produced film is on Apple TV with a 93% Rotten Tomatoes score; his next feature Four by Four (4x4 of the Event) is due next year.
• TV pick: Nine of the Seven Kingdoms (Game of Thrones spinoff).
• Star Trek favorites: Star Trek IV: The Voyage Home and Star Trek VI: The Undiscovered Country.
• Leonard Nimoy directed Star Trek III and IV; IV had no traditional villain (an alien probe instead) and was hugely profitable on a modest budget.
• Nicholas Meyer directed Star Trek VI and co-wrote IV with Harve Bennett; VI includes the line "only Nixon can go to China."
• J.J. Abrams imposed a similar "bad-guy-destroys-Earth" rule on Star Trek Beyond.

Personal AI Tooling and Productivity

The hosts described how they named their dots, the projects they have running, and the moment in March of the current year when productivity tooling crossed a personal threshold.

• Andrew named his dot Sinclair after Clive Sinclair, eccentric British inventor of the $99 Timex/Sinclair computer with a bubble membrane keyboard (because Apple tools were too expensive). The BBC film Micro Men starred Martin Freeman (initially miscalled "Mark Freeman").
• Brian's dot is Pixel, shortened to PIX.
• Brian describes building "the monorogue" as a spiritual successor to Hacking the System.
• Content output grew linearly through about 2019, surged during the pandemic, then saturated; production tools got worse for a time.
• Brian says the world changed for him in March of the current year when productivity tools became good enough that he would "deeply resent this going away."
• Brian compares the last two years of relearning production to starting Assassin's Creed with all skills reset to one ability.
• Andrew describes a "builder mode vs. creator mode" split; the rare ideal is when they agree.
• Brian limits himself to 1–2 projects at a time because Codex wait times tempt him to spin up more; he now has 5 divergent ideas.
• Brian turned 50 and is consolidating projects "under one big tent."

Money, Markets, and Creator Economy

The hosts discussed how agentic tools could disrupt banking, surprise subscription renewals, and how new pricing tiers and creator-funding models may emerge within five years.

• Brian estimates losing a couple thousand dollars a year in interest by parking cash in checking; agentic tools moving money between savings and checking could "kill the banking industry" at 2–3% interest.
• He cites surprise renewals like $500 for unused services and $20–$50 for various tools.
• Hosts discussed Patreon-like funding models a decade before Patreon existed.
• Kickstarter described as "time travel" because it collects payment upfront.
• NACA's 20-minute stage slots let Brian break into the college market as a "fly on the wall" for a year.
• Brian predicts a "killer structure" for AI tools within 5 years that nobody has heard of today.
• Proposed mid-lane tier: $100/month for podcast optimization, EQ, and uploads via automation.
• Renty Human: pay about $20 to find someone reliable for a 45-minute task.
• AI thumbnails, while objectively balanced, will be perceived as "garbage" by viewers under 25 because they look robotic.
• AI tells in YouTube copy include missing punctuation and repetitive phrases like "this matters."
• IKEA-vs-handcrafted chair distinction: mass-produced may be objectively better, but handcrafted signals value at a fancy dinner.

Tinkering, 3D Printing, and Old Tech

The hosts described attempts to recreate a talkie tape on a Bambu printer, mechanical Nipkow TV, and the resurgence of Apple IIe dot-matrix printers running custom firmware.

• Brian tried 3D-printing a talkie tape (cellophane with grooves) on a Bambu printer but lacked fidelity with a 0.4mm nozzle versus the needed 0.2mm; he calculated ~1 meter per second for correct hertz playback.
• Mechanical Nipkow TV described as dots in a pattern struck by a flashing backlight.
• Bambu 3D printers cost a few hundred dollars; Brian just bought a 10-watt laser cutter for $300.
• Old Apple 2E dot matrix printers are being repurposed with custom firmware.

Bambu A1 Mini Print Project

The speaker described a specific hobby print attempt on a Bambu A1 Mini purchased the previous Christmas, where a wrong nozzle and limited build volume led to a broken part.

• The available 0.04mm nozzle (rather than the desired 0.02mm) limited fidelity on the sample print.
• The printer's build volume was less than a square foot, forcing a spiral layout, and the resulting print broke apart because there was insufficient surface area to link the structure.
• The design required motion at roughly one meter per second to generate the correct number of samples along the strip.
• The speaker recommends the Bambu A1 Mini and used ChatGPT on a phone to walk through fixing a clog.

Vibe Coding and Mac Audio Tool

The hosts highlighted a fast, cheap utility built in a single short sitting with AI assistance, illustrating the new speed of small software projects.

• A small utility was built in 10 minutes with AI assistance: a $50 tool that auto-switches Mac audio input between AirPods and an Elgato microphone.

Norm Macdonald Tribute

The hosts paid tribute to Norm Macdonald on the fifth anniversary of his death, recommending his book and noting an upcoming documentary.

• Norm Macdonald died five years before the recording.
• The speaker recommends his semi-autobiographical book Based on a True Story and notes an upcoming Judd Apatow documentary.
• Example joke quoted: "That sounds like a bunch of commie gobbledygook."

Weird Bits

The hosts closed with miscellany ranging from urban-legend investigations to cryptid hunting history and reflections on a skeptic lineage tied to Project Blue Book.

• A host physically investigated Monkey Island in Burbank, finding a concrete structure under grass near a park and highway matching the reported site.
• Hosts recall a livestreamed field hunt for a South Florida "night creeper" (likely a raccoon), and a near-deal with Revision3 plus a Discovery Channel audition that never materialized.
• "Peak goblins" is predicted; African goblin-attack stories were likely folkloric framing of criminal or abusive acts.
• One host's co-host Paul Heinicky's father, J. Allen Hynek (Project Blue Book), discussed unexplained cases; he works as a magician and skeptic but remains open to phenomena he can't debunk.

Episode Reflection and Framework

The hosts closed with a broader reflection on the run and a small decision-making framework they wanted to share.

• The speaker called this one of their favorite moments across the 1,000-episode history of Weird Things.
• Theme: dormant infrastructure becoming useful again, exemplified by Apple IIe dot-matrix printers revived with custom firmware.
• A "five-second-why" framework was shared: keep asking "why" until reaching the real goal, then skip the intermediate scaffolding.