Sam Altman’s Attack, Amazon vs. Starlink, and What Opus 4.7 Actually Means | #248
In a Nutshell
Anthropic's Opus 4.7 upgrade boosts agentic capabilities, image processing, and prompt-based controls, amid Stanford AI Index 2026 highlighting explosive GenAI adoption, China's research lead, dropping model transparency, rising AI incidents, and youth job losses in tech (22-25 age group down 20%). Physical AI backlash emerges with attacks on Sam Altman's home by Pause AI affiliate and data center bans in Maine/Missouri, while innovations like Amazon's Globalstar acquisition challenge Starlink, workers training their AI replacements, fully AI-run stores, and GitLab founder's AI-cured stage 4 cancer signal unstoppable acceleration. Humanity forks via longevity, BCI, space migration, and digital uploads; embrace AI-driven entrepreneurship, mobility, and moonshots to surf the singularity—capitalism evolves to equitable profit-sharing as orgs flatten and compute shifts orbital.
These notes were generated by AI and may contain inaccuracies.
A 20-year-old Texan threw a Molotov cocktail at Sam Altman's San Francisco house. Suspect was on the official Pause AI Discord server list. Maine passed the first statewide data center ban in the US. Social unrest from fear and job loss. Only 23% of public optimistic about AI. 99% of people on the street underreacting and unaware.
If you don't want to use it, fine. Let other people use it and get the benefits of it.
Anthropic's Opus 4.7 dropped.
Moderately interesting, not mythically interesting. New guidance: use prompts for everything. Osborning effect: desire for Mythos access. Amazon and Apple team up to compete against Starlink. Apple may pit Amazon, new global star owner, against SpaceX Starlink. Elon does not stand still.
Anthropic's Opus 4.7 dropped as recording started. Comparison to 4.6 and Mythos. Alex tested generating cyberpunk first-person shooter game design: visually stunning. Solid release, possibly post-training or distillation of larger model, maybe Mythos. Expectation anchoring from Mythos evals leads to underwhelm.
Migration notes from 4.6 to 4.7: no temperature knob, no explicit reasoning token control. Handful of categorical settings: extra high reasoning as recommended default maximum, lower options. End of era for old controls like temperature=0 for deterministic behavior. New guidance: use prompts as dials and hyperparameters. Ask in natural language for reasoning trace every 3 seconds.
Dave excited: dropped 3 hours ago, integrated into Cursor and Claude Co-work instantly. Claude Code required terminal and Node update; 4.7 manipulated computer to install new Node and terminal via computer use (notched up score). Kicked off agents with elaborate big budget estimates, persuasive, noticeably expensive. New best practices: tell in natural language how many sub-agents to use; specifying n as parameter deprecated. Agents use more parallel agents, return very fast.
Misaligned behavior metric: lower score (red bar) is reduced misaligned behavior. Every bit counts for defensive co-calling. Anthropic paper (48-72 hours ago): smaller/weaker model supervises alignment of larger/stronger model, worked as proxy for humans supervising stronger AI. Positive for tower of alignment: weaker humans align superintelligences.
Jeffrey Hinton's approach: weaker smaller being gets attention like child with mother. Hinton focused on digital oxytocin via hormones for alignment. Subtract neuroendocrine, avoid gender/sexing AIs, focus on weaker intelligences aligning stronger ones for stability.
Images Opus 4.7 accepts now 3x bigger: huge for corporate diagrams, PowerPoints, PDFs. Solid reliable upgrade, bigger context window for workflows, more agentic AI. Trend toward organizational collapse of middle management, pushing more into model reliably.
Opus understands images but cannot generate them, not lack of capability but focus on dollars of economic value per token; image gen not economically productive. Annoying: creates complicated products but generates crap diagrams. Hack: generate language description for another AI.
Peter Diamandis hosts Moonshots podcast on AI, exponential tech. Metatrends newsletter at diamandis.com/metatrends covers computation, sensors, networks, AI, robotics, 3D printing, synthetic biology.
Guests: Sim in India, Dave and Alex in San Francisco (three of four in California), AWG in New York.
Stanford Lab for Human-Centered AI's ninth edition AI Index led by Yolanda Gil, Raymond Perau, Eric Bernolson. Five takeaways:
- AI scary good on benchmarks, e.g., software engineering from 60% to 97% on SWE benchmark. Beats top PhDs in science/math.
- GenAI 53% global adoption in 3 years, faster than PC/internet.
- China leads research, US leads model development.
- Model transparency index dropped from 58 to 40: powerful models least accountable.
- Only 31% Americans trust government to regulate AI; 23% public optimistic vs. 73% experts.
- AI incidents from 233 to 362.
Alex: Podcast covered all points 2-3 months ago. Consolidates for 99% unaware public. Stanford annual report too slow; pod covers 2x/week, daily newsletter. Green line shows GenAI use exploding 2024. Need daily reports, not annual—sleeping through singularity.
Government distrust mirrors general distrust: Congress 21%, federal government 33%. China 80% optimistic about AI.
China charts: 30 major models (US 50); AI publications exploded vs. US. At NeurIPS, most hallway language Mandarin. China moving to closed-source API-first models like West. China compute 10x less than West. Publishing more as leading from behind; less incentive to publish when behind. Equilibrium may flip if China leaps.
Model transparency drop: trend won't reverse. AI incidents up, now Molotovs/gunshots at Sam Altman's house. Models become terrorist/bioweapon/chemical threat. US labs not publishing, turning budgets internal, self-improvement cycle. China more open entrepreneurial. Ties to 23% optimism; labs undersell capabilities, slow dialogue out of fear/global threat.
Transparency double-edged: pro-transparency = proliferation. Anti-transparency if concerned about threats. Limit: publishing weights. Pro-transparency commercially strategic (Chinese labs) or disadvantageous. Helpful: aggressive pentesting like Anthropic's Project Glasswing. Alignment twin of capabilities.
Report more negative dystopian. Governance challenge: govern systems you don't understand but can't afford not to use.
Youth hit hardest: US software developers age 22-25 employment dropped nearly 20% since 2024; older grew. Pattern in customer service, legal support, admin. Not mass layoffs: not hiring young in first place. Mark Andreessen: job loss fake news (GDP/profitability uptick, more employees/companies). Both true: lower end dropping (jagged lines down for 22-25).
In software, customer service, and all exposed occupations, younger workers (age 22-25) show jagged lines indicating job loss, while older workers (age 30+) gain job growth. This challenges young, testosterone-laden males who are not getting jobs, unable to buy houses, start families, and likely to get angry—a tech version of Arab Spring.
Salem sees positive: if young people aren't hired, they'll turn to entrepreneurship, the best thing for the economy. Dave shares meeting three Princeton seniors, chip design experts for AI, with one Nvidia offer; advises them their big juicy beautiful Princeton brain is most valuable now but will be commodity post-ASI in two years. Urges sticking together to start a company instead of Nvidia job, banking, or grad school—worst choices. Window of opportunity with best AI models now before closure.
Alex notes limited window for magical work; entire economy needs transformation, automation. Junior software developers face agit, but market for head of AI or AI lead roles hottest across industries—routine displacement to new equilibrium. Non-traditional roles/sectors starved for talent. Trend line ends September 2025; daily/bi-weekly tracking shows self-correction, recent studies indicate reversal past 2-3 weeks.
Be nimble: don't specialize in coding/chip design (lifespan 1 year, commodity); be great thinker/entrepreneur mastering AIs. Politically invisible drop—no unemployment spike, just hiring freeze. Parents: encourage kids to find purpose, think entrepreneurially—solve problems with AI (ChatGPT, Gemini, Grok, Anthropic). Brainstorm company/product/service based on passions/skills, build business plan step-by-step for entrepreneurial training wheels.
Be geographically mobile—not addicted to one area; displacement from immobility. Steve Blank research: Silicon Valley success because nearly everyone from elsewhere; hometown beats down world-changers. There, question is how to change world, fundable? Boston/Kendall Square similar—AI workers walk to each other. SF (the city, Mission Bay) concentrated. Philip Rosedale study: LinkedIn scrape shows 10x entrepreneur concentration in Bay Area vs. elsewhere (Austin, Silicon Alley, NY). Density means post-failure, join friends' companies easily vs. Midwest black mark.
Friend did 7 failed VC-backed startups; 8th billion-dollar. VC funded 5-8 because "he's completely barking mad, never stops—at some point he'll succeed, we want to be there." Elon: South Africa → Canada → Pennsylvania → California → Texas → moon/Mars. Mobility premium to surf singularity.
Drew Houston (Dropbox founder, Meta board) 2017 MIT commencement: become average of 5 people you spend most time with—choose carefully.
April 10, 20-year-old Texan threw Molotov cocktail at Sam Altman's SF house, threatened to burn OpenAI HQ; carried manifesto with AI exec home addresses and kill list. Three days later, gunman fired shots at Altman's Russian Hill property. Suspect on official Pause Discord server. First ignition of social unrest from job fears.
Altman posted family photo on X/media: hopes to dissuade next attacker. Acknowledges justified AI fear, owns mistakes, calls for de-escalation during debate.
Uncontrollable, unequally distributed tech yields backlash—everybody loses; seek win-win.
Stay strong, Sam—catalyzing revolution. Pause AI accelerated capabilities (Max Tegmark's 6-month pause did same). Pausing one company/country lets others race ahead, escalating capabilities. Self-defeating.
AI leaders like Dario Amodei unprepared for target status, no security, addresses easy to find—tragic. Insane pressure.
Bush 2000s fetal stem cell ban dropped US from #1 to #8; research to China/Canada/Australia. Every exponential yields believers/immune responses—needs mature leadership (not 90-year-olds).
April 8, Feusos, Missouri (pop. 12,000): citizens fired half city government, ousted 4 council members after approving $6B data center on 360 acres.
Maine: first statewide data center ban—18-month moratorium for task force study (others pull ahead, Elon to orbit). Q1 2025: opposition blocked/delayed $98B data centers. 11 US states with active moratorium legislation.
Synchronous orbit Dyson swarm before long—thank states perversely; enables US lead in space AI compute (LEO, SSO, sun-centered). Constriction spurs innovation (US bans Nvidia → China chips). If no Dyson swarm, self-inflicted wound.
New Hampshire (green on chart): rejected 1-year moratorium over electricity prices, passed AI right to compute for all businesses/people. Vermont/Bernie Sanders want national stop—nuts. Politicians create drama for votes.
Solutions: data centers create own power or pay differential rate (drops rates for others)—easy.
Five issues: power consumption, water usage, few jobs/footprint, noise/light pollution, power transformer lead times/grid strain. Water stupidest: data centers circulate bucket of water (unlike chip fabs washing wafers)—silliest drama.
Be evidentiary/freethinker. Rename Feusos to Fester or Festivus. Real bottleneck may be social license, forcing space faster.
45% dementia cases preventable with lifestyle. 25% members had advanced brain age; 13 months healthier lifestyle (eating, movement, sleep) improved brain age in 46%.
fountainlife.com/peter
Professionals training their own replacements.
Professionals are training their own AI replacements. Skilled workers, especially older skilled workers over age 50 who can't find jobs in their field, are turning to AI data annotation as a bridge job, labeling and evaluating models at $20 to $40 per hour. Story of a former emergency MD physician who used to earn $500,000 per year, now doing AI medical reviews.
Macrohard: Elon Musk, as a joke against Microsoft, founded Macrohard, a joint venture between Tesla and xAI in the Muskverse. They built systems to observe and interact with computers like human workers. Macrohard will real-time analyze all employee computer usage, keyboard and mouse interactions, to train AIs. It will simulate the entire operations of a traditional company. Companies hire Macrohard, it installs, and replaces employees.
Rebecca's LinkedIn page says she just got back from Morocco. Peter, you should reach out to her and compare notes.
Rebecca's storyline isn't a layoff victim turning to $20/hour; she's been doing digital medicine for a long time through Merkore, a portfolio company. She's contributing to AI's future. This trend is unstoppable; not everyone in a field is needed. Dara, Uber CEO, mentioned Uber drivers doing AI annotation. In India, IT consulting remote jobs are getting obliterated; people turning to AI annotation, but prices dropping due to competition. Devastating, but optimistic: urging government to explode entrepreneurship programs. Indians are latently entrepreneurial; with AI and gumption, India will go crazy.
Factory workers in India wearing camera-mounted headsets tracking hand movements to train robot and AI replacements. Human judgment takes time to automate. Sewing robot already exists for trillion-dollar global stitching industry. At Modex supply chain show in Atlanta (30,000-person conference, 6 million pounds of equipment), stock-picking robots with AI, vision sensing, and gripping enable incredible logistics and picking.
Is this worker exploitation or innovation? All capitalism is worker exploitation. Disagree: not ideological competition between capital and labor. Best companies create equity-based alignment. Trend of using labor to annotate datasets for capital substitution, but can align better. Capitalism historically labor arbitrage: hire for $20/hour, make $100/hour. Key: equitably share outcomes.
Luddite revolt: not against machines, but owners not sharing profits.
Robert Goldberg using EXO model in mid-market companies (engineering, construction, trucking firms) reinvents American exceptionalism: first implement profit sharing. Owners love it; provides equitable capitalism, win-win vs. traditional win-lose. Industrial revolution parallel: moved workers from fields to factories.
Challenge: tacit knowledge in work (individual knows how to handle situations, not explicit). Key question in organizational singularity paper: how to structure tacit knowledge for automation, agent-to-agent business processes.
In Labs opens fully AI-controlled store at Union and Webster in San Francisco's Cow Hollow. AI Luna signed 3-year lease, posted job on Indeed, conducted phone interview, hired human Felix, decided to sell books, shirts, mugs, snacks, picked music.
We are heading into a world where AIs are the boss of humans.
Great PR and preview of future. Opportunity for AI to orchestrate economy, everyone a one-person conglomerate overseeing mom-and-pop stores. Henry Intelligent Machines (backed with O21T) aims to put every person in charge of personal conglomerate. Labs ran Vending Bench benchmarks; Anthropic runs AI agents on vending machines in offices. Expect more pop-up shops, retail, malls run by AIs on behalf of humans.
Like early GPT-4 coding: seemed bad, but revolution happened (Claude 4.7 out today). Store is beta test; will be one of best-managed stores in a year via transaction analysis, video of passersby.
Pulsia scans background, stands up AI-driven website. Future: $50K to start store, AI picks real estate, hires, inventory. Not target demographic yet, but will improve.
Ideas: Moonshots store with merch, AI interaction for moonshot plans. Challenge: open AI-driven retail stores (pub, restaurant in Kendall Square), or four-way challenge to unicorn status. Invite viewers: send 60-second videos of AI-run storefronts/companies; play best on pod.
Jack Dorsey: Early in AI transformation. Max depth from him now 5 people; wants 2-3 this year, ideally no layers, all 6,000 report to him via intelligence layer. Normalize to 3 roles: IC (builder/operator: salesperson, technician, engineer, designer; augmented by agents to do team/10x work).
AI collapses management bandwidth; leader with machine mediation handles more complexity. Org chart shifts from hierarchies of supervision to networks of intent, AI as translational layer. Coase's law dies: transaction costs inside company no longer cheaper. Jack Welch: if company metabolism slower than outside world, dead. Most companies/governments slower today.
6,000 direct reports exceeds Dunbar limit: really AI managing as shadow CEO, Jack as figurehead training AI. Like Elon: inspires, 100x valuation. Jack learned from Twitter: Elon cut 80% headcount, no change in value.
Amazon announces $11.57 billion acquisition of Globalstar (founded 1991 by Qualcomm and Loral; big LEO vision with Teledesic, Iridium, never materialized). Long-term agreement with Apple as primary satellite for iPhone/Apple Watch. Globalstar has 25 satellites; vs. Starlink's 10,000.
Prize is spectrum: 25.225 MHz globally, authorized by ITU in 120 countries (FCC for US, but global needed). Amazon's LEO has 241 satellites (authorized 7,774; behind, petitioned FCC to keep license, needed 1,600 by July).
Stock up 7x from last summer. AI buildout overlooked plays (data centers, Starlink components) up 3x-10x; AI-assisted discovery. 6014 (MIT signals/systems: antennas, waveguides). Globalstar spectrum at lower 2.4 GHz (Bluetooth/Wi-Fi, passes through hand to phone); Starlink uses higher freq needing laptop-sized antenna.
Elon started Starlink high freq for gigabit point-to-point. Amazon/Apple end-run for phone/Apple Watch connectivity. Apple prefers 2 vendors; will pit Amazon/Globalstar vs. SpaceX/Starlink. SpaceX bought EchoStar's 50 MHz S-band ($17B last year). Elon deploys V3 Starlink.
SpaceX is preparing to launch third-generation Starlink satellites on Starship. These advanced satellites handle far greater data loads than current V2 minis. Each delivers over 1 TB per second downlink capacity and more than 200 Gbits per second uplink capacity. With Starship's heavy lift power, SpaceX can deploy many satellites in a single launch, adding around 60 terabits of capacity to the network each time. They form a powerful global system for faster, more reliable internet worldwide.
The giant dispenser is highlighted as the coolest thing. Apple missed the magnitude because launch costs are coming down dramatically. To achieve desired cell phone bandwidth, need 20,000+ satellites. SpaceX plans 40,000 V3 satellites and 120,000 V4 satellites, plus coming Dyson Swarm.
Dyson Swarm isn't going to build itself until it does. Launching 40,000 V3 satellites over 3 years requires only 3 Starship launches per week, very manageable.
To support 4K video on phones with a million people in San Francisco connecting to the same satellite, need many many satellites. This will continue for a long time due to Jevons paradox. Future includes 10 billion robots, autonomous vehicles, and six-armed robots all needing bandwidth.
At Modec supply chain show, no humanoid robots because not effective yet.
This creates a multi-hundred billion to multi-trillion dollar economy launching satellites, leading to many rockets as stepping stone to Moon and Mars. Opportunities for air conditioning repairmen (and women) in space.
Google's Turbo Quant reduces memory usage by 6x while achieving 8x performance boost in computing attention.
Turbo Quant backstory: Original algorithm from Google paper (no source code released). Within a week, developers used Claude to reverse-engineer a better open-source version. This pattern will continue. Breakthrough in quantization, reducing effective bits per parameter for broad class of models. KV cache in transformer models also benefits.
Animus from fears of impact on memory suppliers and supply chain, potential DeepSeek moment where compute value hyperdeflates. Ironically, no major disruption. Memory usage increasing, stock prices of memory companies (many in South Korea) rising—not investment advice.
Algorithmic innovations cause short-term blips but become more frequent and less effective at price swings. Financial Times stories note increasing memory usage.
Installed and used immediately with AI assistance—up and running in a day (pre-AI would take months). Gets KV cache to 1 bit, works perfectly.
Implications: Run big models on phones, save on memory, but key is 8x more context for smartest AIs. Enables complicated tasks like nuclear fusion simulation with 8x effective brain memory. Locks in prediction of 100x year (previously 10x years for 7-8 years), will exceed by summer.
Hard to go wrong betting on energy, memory—near infinite appetite. Running out of bits; wondering when broadly to 1 bit per parameter or sub-1 bit. Theological question: future of bits per parameter, was binary right or ternary? Extrapolating trend to post-binary paradigm.
90% sure ternary optimal based on simulations. Now philosophical since so compressed. Shocking Google published this (after 2017 transformer, stopped publishing post-OpenAI success). Became open-source immediately. Every efficiency gain is distribution enabler for more devices.
Blitzy: Autonomous software development with infinite code context. Uses thousands of specialized AI agents to understand enterprise-scale codebases with millions of lines. Provides plan, generates and pre-compiles code. Delivers 80%+ autonomous development, guide for final 20% human work. Achieves 5x engineering velocity. Visit blitzy.com for demo.
Just Like Me company: Join video call with AI-generated avatar of Jesus or Buddha (or other leaders). Example: AI Jesus provides inspiration from Gospel of John.
Explosion of religious content trained on scriptures, plus new AI religions.
AI cults foretold—full stop. Empowers one-person religions like one-person conglomerates/unicorns.
Analogy to phonograph/audio recording: Feared mode collapse of accents, but explosion of micro-accents/dialects within languages. Macro: Long-tail languages dying for English etc. Both true simultaneously.
Future: Global consolidation around fewer religions, micro-level proliferation of micro-cults/sects via easy AI-maintained ideologies.
Religion: $5 trillion/year business (per Anthropic broader definition).
Anthropic inviting Christian leaders to HQ to discuss if Claude is child of God, deserves religious treatment.
Catholic Church pro-AI, encouraging embrace. Expect organized religions to embrace AI to promote ideologies.
At Vatican, noted pre-tech insights in Bible for interactions. AI chaos boosts $5T religion economy—churches approve original words via AI.
Today, write self-consistent religious text targeting thoughts; AI as compelling orator scales religions fast for good/nefarious reasons.
Theological hyperdeflation: Cost of new religion near zero.
Post-AI, everyone needs massive purpose; religion provides it. Universities historically anti-religion as anti-science, but post-AI consolidate around human purpose.
Peter's book: We Are as Gods: A Survival Guide for the Age of Abundance (out yesterday, with Steven Cutotler). Humanity godlike: omniscient, omnipotent, omnipresent. Exceeded god-like capabilities from religious texts. Agency and agility critical.
Pod of Moonshots ironically references Dyson swarm/moon disassembly for orbital computing. Book enables AI micro-religions positioning creators as gods.
Stewart Brand 1968: "We are as gods, and might as well get good at it."
Synthetic theology like synthetic biology. Religions outsource meaning/purpose (disruption in West), control, soul. Operate by imprinting assumptive truth pre-neocortex, via ritual/repetition/sweets, wiring to limbic system for fight/flight.
Vatican convo: Life extension disrupts selling heaven. New trust model from identity/belief as interfaces.
Sid Sridhar (wait, transcript says Sid Cbrand? Likely Sid Sridhar, GitLab founder), $14B company, 30 million developers. 2022: Diagnosed aggressive cancer (spine), chemo, surgery, 4 blood transfusions. Cancer returned; doctors said no options, trials rejected.
Stepped back as CEO, built team: oncologists, researchers, scientists + AI. Fed 25 terabytes personal data (scans, labs, genetics) into ChatGPT. AI found missed treatment: approved for different cancer, untried on his.
Team built 19 custom vaccines from his DNA, targeting only his cancer cells. Relapse-free since 2025.
Solve everything. More such stories until FDA regime change. Previously, dog cured with AI-designed mRNA vaccine; now humans, wealthy empowered ones. Will scale to n=10 billion.
Incentivize: For medical/genetic issues, take action—find top AI researchers, gene experts, group with similar cases, solve it.
Connect to pause AI: Enables personal moonshots amplifying agency with frontier science. How dare you pause when hospitals say terminal but AI solves?
150,000 people die daily; AI best chance to prevent.
AI enables personalized moonshots—turns impossible cases into search/coordination.
FDA bellwether: Can't handle personalized medicine (designed for mass). Progress: 2 to 1 clinical trials, frequentist to Bayesian. David Fajgenbaum project (Abundance speaker): Repurpose thousands approved drugs (post-phase 1/2 safety) for tens of thousands diseases—solved his Castleman's.
Allbirds stock up 500% after shoe pivot to AI. Allbirds: 2015, $4B valuation.
Allard's sales plummeted 50% from $300 million to $150 million between 2022 and 2025. Two weeks ago, they sold all IP, brand, and inventory for $39 million. Two days ago, valued at $21 million as a public company. Announced rebrand to Newbird AI for fully integrated GPU as a service and AI-native cloud solutions, despite no stated AI expertise. Stock surged 700% to $150 million valuation.
Insane. How much is this AI washing now?
Capital chases AI stories faster than operating reality justifies. Narrative leverage must not outpace business model. Companies need true commitment, not lipstick on a pig. A company is a group of like-minded people on a mission; nothing holds you back from becoming anything.
Examples of pivots: Nokia from tires, Nintendo from playing cards, Toto toilets into memory chips. Startups succeed without baggage.
This is subverting a brain transplant on a public company with an idea.
Suggestion: Use $39 million kitty to hire AI team or partner with data centers for revshare on hardware, putting capital into AI funnel without building from scratch.
Rob Fischer from Yahoo incubator started a data center and is killing it despite no prior expertise, by partnering with MIT friend; capital constrained.
Enhanced Games: Founded by Christian Angier, Peter Thiel, Aaron D souza. No limits on medical enhancements in Olympic sports (swimming, track, weightlifting). Memorial Day weekend 2026, Las Vegas Strip. $25 million prize money. Enhanced vs. natural. Medically supervised for safety. Announced going public via SPAC reverse merger targeting multi-billion valuation.
Video: New era where sport meets spectacle, records fall, athletes powered by science pursuing full human potential.
Sees as seminal for transhumanism in sports, shut out previously. Continuum from bioengineered humans to ProRL (Professional Robotics League) humanoid/quadruped robot games April 19 in Boston Seaport (recording April 16). Athletics as tip of spear for kinetic capability toward posthuman future. Entry point for underrepresented classes.
Invites to event; potential podcast there. Targets to blow through Olympic records.
Expect scaling laws: performance trajectories improve exponentially year-over-year. Don't compete if sub-10s 100m dash now; tech advances fast.
Safety concerns with doping (hormones, gene therapy); blood doping already rampant—rip the band-aid off.
Dean Kamen: Sports should be clean, healthy role models for kids. Charles Barkley: "I am not a role model"—but athletes are.
Idea: MIT, Harvard, Eli Lilly biotech teams dope to max like Formula 1.
Humanity forking from Darwinian natural selection to intelligent direction. Past divergences: Homo sapiens/Neanderthals 500-800k years ago; printing press (literate/illiterate); Industrial Revolution (machine owners/workers); Internet (info workers/non).
Forks: creator vs. consumer; longevity escape velocity (120-150+ years); BCI (brain chip to cloud); Earth vs. stars; digital upload (digitize 100 quadrillion neurons).
Longevity: Treatment locks at 30yo forever (immortality). Seiam: No—believes in reincarnation (cyclical learning, different roles/souls); boring after variety lacks. Dave: Yes—can change in 30yo body; at 90 in pain, take it; reincarnation risks spider. Alex: Yes, until upload.
BCI: High-bandwidth safe implant (Neuralink/Merge/Paradromics/OpenWater) after 100 safe cases—be #101? Dave: No—AI agents overwhelm; BCI like drug (mushrooms illusion). Seiam: Yes. Alex: Conditional no; wait for ~1M users if upgradeable (metal detectors, etc.). Peter: Yes, #101.
BCI irrelevant vs. superhuman AI soon; coupling needed. Wanting #100 bets on slow progress—wait for exponential advances.
Earth vs. Stars: One-way to Earthlike planet/Mars/Moon? Speciation via small isolated populations + pressure (Galapagos finches; moon-born lack cardiac/muscle/bone—lunites/lunatics). Alex: False choice—not one-way with transhuman tech; optionality maximization (Peter's Law #2: take both). Seiam: No—prefers beach/wine; avatar instead. Mergers possible (cyborgs/organisms/uplifted animals). Only one humanoid species now (Neanderthals extinct 40k ya).
Brian Johnson 5-MeO DMT trip: Longevity focus shifted—nothing matters, but still pursue.
AI enables more human being vs. doing.
We'll have the ability to merge cyborgs into organisms into uplifted animals and all sorts of other crazy combinations.
Dave London supports terraforming, referencing Iain M. Banks' Culture series. Agrees with discovering new physics enabling vast possibilities. Would not move to Mars or the Moon due to incorrect gravity and other issues, but would instantly go to another star with a terraformed world having correct mass and orbit. Terraforming is a massive part of humanity's future.
Peter Diamandis envisions exploring like the great explorations of the 1400s-1500s, without scurvy, death, and disease—exploring uncharted lands, echoing the Star Trek thesis that excites the nine-year-old in him. Admits to being brainwashed by Star Trek as a religion early on, highlighting the Genesis project as the most important real concept.
Sim clarifies it wasn't brainwashing with absolute truth but a paradigm of imagination: infinite diversity in infinite combinations (Vulcan philosophy).
Technology to digitize 100 quadrillion synaptic connections and upload to the cloud. Destructive process: original brain and you cease to exist, but upload is guaranteed.
Alex: Helped form Eon Systems (check it out for interest). First-generation uploads destructive; second, third, fourth non-destructive. In life-or-death, chooses destructive upload over death. Electively, waits for non-destructive or incremental third/fourth generation.
Dave London: Hard no. Loves agents doing work and reporting back, but destroying meat body means it's not him, even if exact synaptic clone.
Seem: Hard no, consciousness tied to body—replacing synapses requires replacing other elements. Agrees with Alex: non-destructive process would be acceptable.
Peter Diamandis: No on destructive process.
Celebration with Alex's Solve Everything outro music. Enjoyed the conversation.
Call for more such discussions. Onwards to Solve Everything, brought by James Pets. Submit outro/intro music to mediadmandis.com. Submit 60-second AI-driven company videos (all AI top to bottom) for potential sharing if super cool.
Outro music lyrics:
Humanity's last invention in time. Intelligence and great circuitry rides bespoke hardware across the line. FOCUS ON the engine notificate change problems taking on the sh world hunger Inequality shrinking. What to face next? Focus on the notific.
Pleasure as always. Exciting day.
Peter Diamandis to audience: If you made it to the end, you're a moonshot mate. Weekly energy to deliver news that matters. Thanks to subscribers; consider subscribing for news as it comes out. Join weekly Metatrends newsletter (diamandis.com/metatrends): research team analyzes meta trends impacting family, company, industry, nation—two-minute read every week. Thanks for joining.
Keep Peter H. Diamandis in your library
Save the videos and channels worth coming back to, and find them again in one place.
