#875: The Random Show — Tim and Kevin Talk Retreats, Mortality, AI Predictions, Supplements, Rock Climbing at (Almost) 50, and Not Waiting for “Someday”

We open with clinking glasses and goofy sci‑fi banter, then settle into, yeah, tonight a small drink is the move.

Getting older means paying a bigger tax on booze, but I’m fine with a little edge‑softener, and the ketamine jokes only count if it’s done safely under medical supervision.

I sprinted my old dog Toaster to the ER after violent shaking and collapse, got sprayed in the car, and none of it mattered because I’d do anything for him; it turned out to be stress from hours of standing at a prior vet visit, and he’s okay.

Dogs are such rare companions—co‑evolved partners who even understand pointing—and it’s wild we share our homes with fanged, sixty‑pound friends who feel like family.

Life’s been life‑ing: we just lost a dear colleague, and I got the news mid silent retreat, which made me see grief as the shape love leaves behind; it hurts because we were lucky to love that much.

I’m grateful to Matt Mullenweg for an Antarctica trip that gave me real time with Om, and Tim Urban’s The Tail End pushed me to take family trips before it was too late; Sam Harris’s The Last Time is a sharp reminder that you rarely know you’re doing anything for the last time.

At a five‑day Zen retreat with Henry Shukman and Yamada Roshi, I had a brief, clear taste of ‘nothing lacking’—not an emotion, just a complete suchness—while working my koan and got rushed into dokusan to check it.

On The Way app, meditations like Whole Earth Is Medicine and This Too Is Me help me include whatever arises and relax my grip, even down to dropping the jaw; Henry’s voice is soothing, and I’m genuinely happy you’ve stayed with practice for years.

As I near fifty, I’m done stockpiling ‘someday’ hobbies and want to actually pick a few and go deep.

I’m choosing rock climbing—multi‑pitch in Yosemite gives me a deadline and a why, it’s social and puzzle‑rich, and seeing strong climbers in their sixties and seventies sold me on the long game.

I’m training grip with Michael Eckert’s finger protocol and this little wooden ‘Nug’ you turned me on to, because lean and strong beats big for the next decade.

My road kit is the Nug, a band, an Alpha Ball, and Emil Abrahamsson’s easy ‘A‑hangs’—ten seconds on, fifty off, ten minutes—twice a day; the forearm endurance jumps fast, even if Free Solo still makes my hands sweat.

Quick nutrition note: I’m liking an A2 whey shake that’s easy on the gut, and I’m curious how you think about processed meats and nitrates.

I travel with Maui Nui venison and disclose I’m involved; I use a shelf‑life heuristic—short is good, ultra‑processed is not—and treat organ‑meat sticks like a multivitamin a few times a week.

Exogenous ketones with one‑three‑butanediol work in a pinch but I use them sparingly, especially with older folks given balance risks; the cognitive lift can be striking, but moderation matters.

My mom’s in a safer place now with fall‑detection tech, and I’m focusing on showing up; I’m grateful she still recognizes me.

For brain health I finally found a stationary bike setup that lets me do Norwegian 4x4s comfortably; three sessions a week for months shows hippocampal gains that can last years, and I stack in sauna or even a hot bath when time’s tight.

My ‘good enough’ cardio is treadmill incline at an easy pace while playing Duolingo chess so the minutes disappear.

A JAMA trial of MM‑120 (a lysergide compound) for generalized anxiety showed clear dose‑response out to twelve weeks, with 100 micrograms looking like a minimum effective dose, and Neuroscape at UCSF is in the mix.

The dementia angle is fascinating and fraught—there are case reports of high‑dose psilocybin or microdosed lysergamides boosting fluency briefly, but consent and the risk of a bad experience make it a hard call.

On a lighter nootropic note, I’ve been using caffeinated toothpicks to pace my intake so I don’t chug six coffees without noticing.

AI at home is getting fun: my Ubiquiti cameras and API hooks now recognize faces and plates, trigger a speaker to shoo off alley loiterers, open the gate for my car, and even read me my team’s score if I walk in wearing the hat.

I’m running Claude with Gmail APIs for a twenty‑year investing post‑mortem to test my stories against data, and the big win with LLMs in health has been fast checks on meds, supplements, and side effects—still cross‑verified to avoid hallucinations.

Most automation just speeds up unimportant work, and seeing models write in my voice can sap motivation, but asking open‑ended, personal questions like ‘what should I explore in the next five years?’ produced scarily good, actually useful ideas.

I’m prototyping ‘Bond’: swipe‑based values picking, then mutual commitments tied to those values with agreed stakes, so we surface the ‘dark information’ of trust and reliability, track where we flake, and course‑correct—like a living 360 without the corporate varnish.

For writing, I dump thoughts into WhisperFlow and have Claude shape them into clean emails, and for travel audio I love Shokz bone‑conduction headsets plus a tiny Sennheiser lav; Ferrite on iOS makes hotel recordings sound way better than they should.

I popped in with the occasional chaos—ketamine ribbing, a last‑minute trampoline flip at forty‑nine, and the running joke that if there’s a weird cut and we come back sounding way smarter, you know why.

Quick gear note before we dive in: I’ll drop the link to that lossless recorder in the show notes—the quality’s wild—and AI cleanup in tools like Descript can make it sound studio‑grade. I’ve also been leaning on Gemini more lately because it just snaps into G Suite.

You can do so much now with multimodal models; Gemini’s solid, and 3.5 Flash is strong too, while Anthropic just unveiled a new model I still need to test. The tools are good enough that you don’t need a dozen apps to get pro results.

Zooming out, what does the next couple of years look like with Anthropic and OpenAI racing and that whole Fable drama bouncing from “national security threat” to “fine” overnight?

I see a big three: Google, Anthropic, and OpenAI, with X in the mix because I’d never count out Elon. Grok has tight access to the X firehose, feels grounded on current events, and we lean on it at DIGG when we want high‑fidelity facts.

Play futurist for a second—two years out in AI time, which is basically dog years—what changes?

Google’s set up well because they own the stack, right down to chips engineered for massive memory bandwidth, which points to continuous learning. If self‑improving models arrive in roughly twelve to eighteen months, always‑on training favors whoever has the best silicon, systems, and talent.

TSMC will bottleneck supply, but the spend and recruiting are intense—some packages are jaw‑dropping—while Meta’s great at consumer apps, I don’t see them catching the true model leaders.

So what do those leaders look like in two years, given Google’s data centers and Android footprint?

Android gives Google a giant install base, and inference costs probably settle into a Netflix‑style subscription. I wouldn’t write off Apple either, and I don’t think AI is winner‑take‑all unless someone hits real escape velocity.

Consumer mindshare still screams ChatGPT, even though ads are hard and Anthropic’s enterprise engine is ripping, at least per ARR headlines. Meanwhile, the Gmail API flow for Claude is brutal on the Google side, and I doubt frontier labs avoid heavy regulation for long.

China will pressure that—lots of open‑source models and a “run it yourself” posture that spreads their tech. Even AMD’s new roughly four‑thousand‑dollar box can run hefty models locally, so some teams will replace big cloud bills with a single machine and accept an eight‑month lag.

From a business lens, where does that leave Google?

Word is Google has models beyond Fable but hasn’t shipped due to cost and likely government brakes. Having seen Google X years ago with Sergey, I’d say never underestimate the moonshots under the hood; as an investor, owning a basket of the top names can be saner than chasing one hero.

You tease me about my swings, but I did a simple thing: I picked a short list—Google, Anthropic, OpenAI, Waymo, Vercel, Crusoe—and asked a few killer investors and their tech analysts to sanity‑check it, which helped me pull the trigger on Google.

For most people, the good news is many of these companies will be public soon. Prices may look rich—same story people told about Amazon back in the day—but if you’re investing, think long holds, not day trades.

Public markets still spook me with sharks, shorts, and derivatives, so I try to keep it simple.

If I owned SpaceX, I’d plan to hold for a decade; that mindset has served me best. Big lesson: I’ve lost more by selling too early than I’ve made from the initial buy.

Same here—let winners run, which is easy to say and hard to do. Also, some data shows retail investors who buy at IPO and just hold for a decade can match or beat early fund returns; not investment advice, just perspective.

A practical rule: if you love a product, consider owning the business behind it instead of only buying the product. Buffett’s portfolio echoes that—durable brands with strong margins tend to ride out storms.

He plays the aw‑shucks grandpa well, but he’s a stone‑cold dealmaker when markets seize. For regular folks, one heuristic is to scan your credit card statement and ask if you’ll spend more or less on those brands in five years.

Before we wrap, quick plug: we relaunched DIGG, and weekly users jumped from around twenty thousand to nearly five hundred thousand. It pulls the tech and AI zeitgeist from X, YouTube, TikTok, and more—check DIGG dot com and find me at Kevin Rose on Instagram.

One last note on AI’s ripple effects: my nonfiction book sales fell steadily after ChatGPT 3.5—chart and thoughts are in my post “AI, nonfiction, Tim Ferriss,” and it’s sobering but not doomsday. You can find a thousand‑plus posts at my blog, I’m light on social these days, and if you want something fun and useful each week, Five‑Bullet Friday is still rolling.

Now hand over the tequila.

You’re over‑served, old man—let’s land this plane before you go peel an orange on the mic.