Overview
Today’s thread runs through risk and reward in the AI era. Markets flirt with automation as Grok’s demo portfolio outpaces the S&P, while Taleb warns that popular trading crutches can magnify tails. Hardware firms pivot to data centres, platforms jostle over ads, and policymakers and media trust come under strain. Alongside, a few grounded stories from customs desks to film sets remind us how culture and rules shape the pace of change.
The big picture
Taleb’s new paper: stop-losses clip variance in theory, inflate tails in reality
Nassim Nicholas Taleb teases an updated paper on stop-loss strategies that uses barrier option maths to show why they tidy up noise in Gaussian models yet worsen tail risk in real markets. He highlights slippage and altered stochastic integrals that can make dynamic stops riskier than just holding the asset, reviving ideas he has pushed since the 1990s.
Grok’s $10k portfolio beats the S&P, but can it survive a drawdown?
A tracker account says Grok’s February-launched $10,000 portfolio returned 24% by December, topping the S&P’s roughly 18%. December tilts include AMZN, VRT, NFLX and tech-heavy bets via MU, AVGO and XLY, with replies pointing out that momentum wins in uptrends then hurts in bears.
Micron exits consumer RAM to chase HBM and servers
Micron is winding down Crucial-branded consumer RAM and SSD sales through retail, with shipments running until the end of fiscal Q2 2026. The move reallocates capacity to higher-growth data centre products like HBM, as commenters brace for pricier consumer memory and a bout of panic buying.
Google’s no-ads play for Gemini
A commentator argues Google should keep Gemini ad-free, eat the cost, and force rivals to lean on ads to fund their models. With AI overviews buoying search revenues, Google can cross-subsidise, while OpenAI’s ad experiments face sceptical users.
Jensen Huang: AI to generate 90% of “knowledge” within 2-3 years
NVIDIA’s CEO tells Joe Rogan that AI will produce most of the world’s knowledge soon, framing it as resynthesis of verified data. The clip reignites concerns around hallucinations and error rates, even as NVIDIA’s kit underpins much of the current boom.
Dwarkesh Patel: if AI needs this much manual graft, it is not AGI yet
Responding to a lawyer touting gains, Dwarkesh says the need for prompts, few-shot examples and tool wiring shows current models are far from AGI. He cites his podcast workflow to argue chat-based boosts feel like better search, not a labour replacement that would spread through firms at warp speed.
OpenAI Foundation’s People-First AI Fund backs 208 nonprofits
OpenAI announces $40.5 million in unrestricted grants for community groups focused on workforce, youth and rural health programmes, under a pledge set in July. The timing draws replies tying the philanthropy to user anger over GPT-4o deprecation and safety routing.
Balaji on nations, empires and dilution
Balaji sketches a trap: small countries risk absorption by larger powers, while sprawling empires dilute their core culture. He nods to his network state idea as an alternative that uses selective membership to avoid both fates.
Media blind spot on Minnesota fraud, says Turley
Law professor Jonathan Turley slams major networks for overlooking a Minnesota child-nutrition scandal, where hundreds of millions were siphoned during the pandemic. He frames the silence as partisan, noting years of warnings and dozens of convictions.
Nigeria’s import maze goes viral over an iPhone
A viral quip about the NDLEA and an iPhone 17 Pro Max taps public frustration with customs, fees and agency overlap. Replies stack up with tales of valuation disputes and surprise charges that can turn gifts into headaches.
Knife rules in China, a traveller’s lesson
Pieter Levels tries to ship a Japanese kitchen knife from China to Europe, only to hit customs bans that even block outbound parcels. Workarounds flood in, reflecting tight controls on blades since high-profile attacks.
DiCaprio to young actors: build for the long run
Leonardo DiCaprio urges newcomers to treat acting as a marathon, choosing roles that sustain a career rather than chasing quick heat. The advice lands as he promotes a new Paul Thomas Anderson project with a satirical bent.
Why it matters
- Risk is not tidy: Taleb’s note is a timely reminder that tidy backtests can hide fragile tails. In a cycle where AI-fuelled momentum performs, that blind spot can encourage sizing that looks smart until it does not.
- Supply pivots ripple: Micron’s move signals where margins lie, namely HBM and servers. Consumers may face higher memory prices just as AI PCs and local inference push demand up.
- Platforms set the rules: Whether Google runs Gemini without ads or OpenAI funds community groups, platform choices shape the market’s economics and public trust. Users care about speed, costs and control more than press releases.
- Capability over hype: Huang’s claim meets Dwarkesh’s caution. If models still need heavy handholding, the bottleneck is capability, not adoption. That gap will decide how fast AI replaces tasks rather than just assisting them.
- Borders still bite: From Lagos fees to Beijing knife bans, trade frictions and safety rules remain stubbornly offline, reminding us that the future arrives in uneven patches.
- Culture compounds: DiCaprio’s long-game advice is a useful counterpoint to short-term market wins. Staying power, in art or tech, comes from choices that survive the next turn in the cycle.
