[@ModernRogue] Weird Things Podcast #1001 - We Tell You The Future

Link: https://youtu.be/E8PpRX8xTMU

Duration: 119 min

Transcript: Download plain text

Short Summary

This multi-host discussion on the Weird Things podcast covers the latest milestone plus a roundtable featuring Brian Brushwood, Andrew Maine, and Justin Robert Young, with no single interviewee. The hosts explore AI assistants like Codex and the 'Dots' agent, amateur astronomy breakthroughs including a Claude AI-discovered exoplanet candidate, and the prospect of a Neptune-sized Planet 9 in the outer solar system. They also debate how LLMs are well-suited for signal-in-noise analysis akin to SETI@home and share personal project-management strategies for sustaining 'vibe projects.'

Key Quotes

1. "And it also just kind of opens up a lot of different lanes for them. Would it be shocking in a world where this exists for them to come out with hardware for a phone that would be, you know, Grockbot native? You know, you're you're doing X, Y, and Z. Like that that seems like something that would be a possibility." (00:15:15)
2. "So it's deadlier, Brian. And the reason why that's good news is because it doesn't have as long of an incubation period. That was the big tricky thing with COVID was that it had this very long, very camouflaged incubation period. And so spreading it was very hard to contain because you didn't know that you had it at the time that you were spreading it." (00:25:51)
3. "I I think that that we've always been extremely populist when it comes to information. You know, even the first segment of this very program, we talked about like give people the actual truth to play with. And uh you know when I get in the most trouble, take scam school or what have you. It's because some people would very much like to have fewer competitors." (00:54:24)
4. "I think that they meaningfully eroded public trust in how our government communicates these things. And I think that that's incredibly dangerous. And to demonize and to talk down to people when there's something this important on the line is reckless." (00:45:02)
5. "Like we're all sitting on pioneer probes in our own backyards. And now we have the ability to tell the wizarding box what to do with it to make it alive again. And I I I as excited as I am to see what can change in the world uh by solving you know your fancy math problems or world whatever I'm even more excited to see the second order changes of what happens when my kids generation grow up empowered to do things without intermediaries." (01:03:58)

Detailed Summary

Episode Overview

The multi-host Weird Things podcast broadcast its 10,000th episode as a YouTube live stream with co-hosts Brian Brushwood, Andrew Maine, and Justin Robert Young, with producer "Meep Moop" running the back end. The roundtable conversation spanned LLM capabilities, the OpenAI hardware/software roadmap, amateur astronomy breakthroughs, project-management strategies for sustaining "vibe projects," Russia's reported plague-lab incident, COVID transparency critiques, and post-scarcity futurism.

Stream Mechanics and Community

• A limited-edition founders club challenge coin, originally vault-locked for the 999 Modern Rogue original founders, was released in surplus with roughly 200+ coins becoming available at coin.scamstuff.com.
• The hosts promised a free coin giveaway to the most active chatter; chat volume was running 3x normal at the time of the announcement.
• Chat users Joey H66 and DZ Ren were specifically acknowledged mid-stream, and Andrew was wearing glasses he swiped from Eden, prompting Clark Kent jokes throughout the broadcast.

SpaceX, Starlink, and Connectivity

• Starship reached orbit and deployed a batch of Starlinks, which the hosts framed as a "gigantic step forward" in SpaceX's cadence.
• SpaceX made a major spectrum acquisition (the so-called "SI" buy) that the panel said could enable Starlink-based mobile/direct-to-phone service.
• Current Starlink terminals require a dish and clear line-of-sight and do not work indoors, but the newly purchased spectrum is said to transmit through buildings.
• Brian pushed back on space-based AI data centers, arguing that the physical distance between satellites makes training uneconomical, though he conceded inference might still be viable.
• Andrew noted he has zero cell coverage in a small Bay Area valley even inside San Francisco, and Brian said he would pay an extra $10/month for a satellite/cellular hybrid that just works.
• Justin compared Starlink's incremental improvements to Amtrak's faster trains running on 60% curved track: individual components improve, but overall coverage does not.
• The hosts joked about a rumored pandemic coinciding with GTA 6's release next month while Brian is in Texas.

Russia Plague Lab Incident

• A panelist raised a report of a Russian pathogen research facility running gain-of-function plague experiments, where a lab worker's social-media post initially said she died of plague, was later edited to "might have" died, after which a news blackout and hospital shutdowns followed.
• Russia publicly said people were being treated with antibiotics and the situation is "under control," but has not admitted the lab connection.
• Plague has a roughly one-day incubation period, far shorter than COVID's long, camouflaged onset, which historically arises in isolated outbreaks from Siberia/Mongolia tied to undercooked animals.
• The panel cited the 1977 Russian flu outbreak as precedent, noting it mirrored a 1950s strain, a likely freezer-pull lab leak.
• About a week and a half after the first report, no outbreaks had been traced to planes leaving the remote Russian city, which the panel read as a cautiously optimistic containment sign.

COVID Transparency and Lab Leak Critiques

• The hosts argued that maximum safety requires maximum transparency, but countries involved lack the incentive (or a free press) to be honest.
• It was noted early on that COVID doesn't affect young, healthy people the same way; the panel questioned how fast that knowledge reached mitigation authorities.
• A short-incubation pathogen would actually make a shorter, harder lockdown more justifiable, though strict-lockdown countries still saw outbreaks and political exemptions.
• A speaker claimed that more elderly people he knows died after the COVID shot than from COVID itself and that some MDs privately refused the shot.
• The U.S. pandemic plan was reportedly originated after George W. Bush read about the Spanish flu in his second term; the panel questioned whether those lessons applied in 2020.
• The panel criticized the lack of a 9/11-style after-action report on COVID, contrasting it with the post-9/11 FBI/CIA information-sharing reforms that actually followed the commission.
• They flagged entrenched bureaucracies engaged in willful deception and ongoing cover-ups as protecting their larger mission rather than the public.
• The Rand Paul hearings and document disclosures were cited as evidence largely ignored by mainstream media.
• They called for a CDC/NIH after-action report that declassifies everything possible to explain how Dr. Fauci's recommendations reached local officials in California, Texas, Connecticut, and Florida.
• Wuhan applied for gain-of-function grants through EcoHealth Alliance, and the U.S. continued funding despite misgivings about lab quality.
• The hosts warned that a segment of the population has lost all trust in government messaging, which would be catastrophic in a deadlier future outbreak, urging public health communicators to meet the public where it is rather than talk down.

LLM Disruption in Mathematics and Cryptography

• The hosts framed the current moment as one of incredible acceleration in AI/LLM capability, citing Codex and ChatGPT as if they had existed "six years earlier."
• OpenAI released 300+ proofs on major open math problems, prompting an "association of human mathematicians" response that opened with "we did not ask for this."
• Defenders framed AI proofs as X-rays of the problem space, not destructive digs, clarifying where to dig next.
• A major Navier-Stokes proof was reportedly completed with a single prompt to an internal Pro-tier GPT, rather than the thousands of agents initially speculated.
• The hosts identified a frontier-lab split: OpenAI and Anthropic ran ahead while academia-driven companies were slowed by skepticism toward scaling LLMs.
• Cryptographic problems were underrepresented in LLM results, raising "Q-day" concerns, the day quantum or AI breaks current crypto, which could itself undermine crypto markets.
• A proposed Q-day fallback is one-time pads stored in physical safes.
• One speaker noted a crypto-firm head he knows maintains 24-hour security in his driveway due to physical-attack risk against key holders.
• The hosts observed that 99% of their past roles no longer exist, informing their unemotional view of disruption, and tied this back to Scam School's populist stance that no trick in his stage show was hidden from the audience.

Post-Scarcity, Robotics, and the Physical World

• Brian argued real abundance means continuous change, not perfection, and that "pseudo post-scarcity" applies only to brief phases like ages 15–25 or 20–30.
• He acknowledged paying $2,000 for something he wouldn't pay $20,000 for, citing economic bottlenecks, not human nature, as the constraint.
• Robotics examples: three robots clearing brush could let him run events "Disneyland-style," and robotics could turn an old Mansfield, Ohio high school into a hotel in 6 months instead of 3 years, also reviving comic shops/record stores by lowering rent and labor.
• Historical parallel: the U.S. highway system once took 4 days to drive LA to Fresno and devastated small-town motor-lodge economies.
• Brian's daughter, who has never known a world without AI, is motivated to brush her teeth by AI-generated images of her with "Lvakalola," a cow from a Spanish children's song.
• Counter to post-scarcity: Taylor Swift front-row seats, Hawaii vacations, and World Cup tickets aren't replicable via VR or mass production.
• Brian discussed with a former OpenAI colleague funding research into whether faster-than-light travel is possible, which requires establishing a cosmic reference frame to resolve causality.
• He joked his "futurist hat" is a Robin Hood cap of print-on-demand lab-grown diamonds, self-described as "Diamond Age."
• Human-crafted virtual realities already exist (subculture Discord groups, "USS Enterprise in the Federation"), though some want real science problems solved rather than "Meta 5000s."
• Brian speculated that "Bella's" generation may embrace physical community because rejecting old ways feels "lame and boring."
• Hand-made physical artifacts (a 300-year-old colonial chair) are gaining value, album sales are up, and he is tempted to build a Blu-ray wall because digital media sometimes disappears.
• Fuji instant cameras excite kids who have never owned lasting physical media, and popcorn-bucket collectibles are now a roughly $100 million industry.
• Magic tricks are best designed for amateurs because the playing-with-props experience beats performing them; Brian cited paying $50 to watch a magic trick as evidence the thrill of the moment is what consumers want.
• The Rogu's ring product's escape shim is metaphorical, not literal, and the ring's hieroglyphs act as crib notes for 23 tricks.
• Chat user Dar argued physical/digital hybrids explain Cyber Deck builds; the host added that they look more fun to build than to use.
• Brian owns a Mod Retro Game Boy with Wi-Fi/Bluetooth (theoretical RSS and Codex control), and recently used ChatGPT's inline UI to generate a Flappy Bird game.

AI Tool Features and Codex

• Andrew announced a new Codex feature that acts as an autocomplete, suggesting next prompts based on what the user has been saying, doing, and the current state of the project.
• Brian said this feature is particularly useful for projects where he is in over his head, such as his fitness app, where he needs help with code review and deployment at scale.

The 'Dots' Assistant and OpenAI Hardware Vision

• Andrew called the 'Dots' assistant a game-changer because of its single persistent thread, admitting his biggest weakness is not using it enough; his wife Rush uses her Dot constantly.
• He described ChatGPT's Dots as a one-point control source for every computer with Codex installed, plus a cloud computer for delegating tasks away from the physical machine.
• Speakers envisioned a clear path from OpenAI software and cloud to OpenAI-branded hardware, wanting Dot capabilities (screen, battery, functionality) on such a device, dubbed 'Dot on a thing.'
• They requested Dot live outside the ChatGPT app, ideally headless in text messages, kicking out to a separate visualizer only when visualization or a browser is needed.

Project Focus and 'Vibe Projects'

• Brian said the only remaining moat is staying interested in a vibe project longer than two days.
• Andrew limits himself to roughly two concurrent projects (typically one software project and one other) to avoid abandoning work after the zero-to-one thrill.
• Brian recounted building a Trump/Jadakus laugh project where he generated 50 different laughs, built a voting site, and tested them across five or six Trump speeches before questioning the effort, calling the token spend 'AI psychosis.'
• Brian's advice: pick one thing and put all effort into it; do not start nine other projects.

Amateur Astronomy and AI Discoveries

• Brian described a Reddit Claude AI thread where a non-astronomer tech used Claude Code to find a legitimate exoplanet candidate and got NASA to point a telescope for confirmation.
• Multiple researchers, including a professor, confirmed the signal.
• A Claude AI modbot TLDR praised the find as 'incredible' after a flood of useless AI-generated replies.
• Astronomers were reportedly upset by an amateur finding a planet candidate using AI.
• Justin referenced SETI@home as the first distributed networking super-campaign that processed raw telescope data to find signal in noise, noting LLMs are well-suited to the same task.

Planet 9 Hypothesis

• Brian noted a rumored large planet (Planet 9, which would have been Planet 10 without Pluto's demotion) possibly beyond Neptune's orbit whose gravity may explain odd motions of outer solar system objects.
• The hypothesized planet could be Neptune-sized and very far out, making detection difficult without waiting for transits or sifting through large datasets.
• Speakers predicted significant astronomical discoveries in the near term because AI excels at math and imaging, which are core to astronomical data analysis.

Speaking Engagements and AI Reliability

• One speaker prepared and delivered a 1 hour 15 minute talk in Omaha by waiting to see the venue before assembling the program, then having a two-hour conversation with an AI to produce flawless PowerPoint slides with teleillustrator overlays, describing it as one of the best technical talks of his career.
• The same speaker gave an Ecuador talk where they spent the first 10 minutes addressing 'AI doom' questions off-slide before proceeding; the audience responded positively.
• AI tools have only become reliably integrable into augmenting workflows over the last six months, even though impressive results were possible one to three years ago with dedication.
• The speaker recommended playing with the Dot (AI agent) and telling it to do things in the cloud computer, citing an example of asking it to render a 3D Blender image of their daily schedule.

Tangential Recommendations and Asides

• Andrew recommended watching Godfather 1 and 2 back-to-back for fans of Rambo and Sicario.
• Brian is rewatching For All Mankind from the beginning, praising Ronald D. Moore's commitment to confronting finite-resource storytelling (famously asking "how many shuttlecraft do we have" on Star Trek Voyager) and crediting him with the premise of Battlestar Galactica.
• In season 3 of Breaking Bad, Jesse Plemons's character kills someone on purpose, after which the showrunners decided to never reference that he killed a man.

Surprises and Notable Numbers

• 10,000th episode milestone; 999 original Modern Rogue founders; 200+ remaining coins; 3x normal chat volume.
• 60% of Amtrak track curved; $10/month satellite/cell hybrid; $2,000 vs $20,000 willingness to pay; 300+ AI math proofs; 300-year-old colonial chair; $100M popcorn-bucket collectible industry; $50 magic-trick experience; 23 tricks encoded on Rogu's ring.
• 4 days LA-to-Fresno pre-highway; 6 months vs. 3 years to convert a high school into a hotel; 24-hour driveway security for crypto holders.
• 50 different laughs generated for the Trump/Jadakus project; 2 concurrent projects as Andrew's self-imposed cap; 1 hour 15 minute talk in Omaha prepared via AI; 10 minutes addressing 'AI doom' in the Ecuador talk; 6 months of reliable AI workflow integration; 99% of past roles no longer exist.