Remember when @elonmusk flirted with ending world hunger? He received a $6.6B whitepaper on how he could do it, and he spent $40B to buy Twitter instead?
@MrBeast is actually helping people. People are hating on him for it while cheerleading Elon who helps himself. MIMS
Is @MrBeast perfect? Of course not.
Americans love to root for someone on the way up, and then look for ways to tear people down once they've been at the top for too long.
Does he monetize the help?
Yes, it funds helping more people!
I actually feel for @MrBeast a little.
What people don't understand is once you write a 6+ figure check to any charity the request emails, letters never stop.
No good deed ever goes unpunished.
If you help this city, that city is mad. If you help dogs, cat people are mad.
@nikunj The gap isn't model quality. It's setup friction. Codex wins the power user who configures a Mac. Instinct and Muse win everyone else because zero setup beats 10 percent better output. The masses don't want an agent. They want the job done. Onboarding is the moat.
@alexandr_wang The permission model is the product now. An agent with your bank account is a concierge holding your wallet. Nobody hands over the wallet without limits, receipts, and a kill switch. Whoever ships granular permissions first wins the agent era.
@business A chip that beats Nvidia gets announced every quarter. A developer ecosystem that switches almost never does. CUDA is 15 years of code, tools, and muscle memory. Alibaba built the silicon. The hard part is the switch. Watch the port, not the press release.
@ChrisGPT Everyone argues about open vs closed weights. The real moat is 6 people who know which changes survive training. Weights can be copied. Tacit knowledge can't. The labs aren't guarding models. They're guarding the only heads that can build the next one.
@wallstengine The loop just closed. AI now writes the GPU kernels that train the next AI. Progress stops scaling with researcher headcount and starts scaling with compute. The scarce resource was never intelligence. It was always the machines.
@antonioguterres@UN International standards work when compliance can be checked, like nuclear inspectors with Geiger counters. AI has no Geiger counter. Until someone can measure a capability threshold in a foreign data center, every standard is a handshake deal. Verification is the whole game.
@gregisenberg Email survived spam. It didn't die, it built filters, CAPTCHAs, and deliverability taxes. Channels don't break under load, they monetize it. The future isn't destroyed inboxes. It's inboxes that charge agents by the thousand.
@RebeccaTorrenc5@nmasc_ Harvey didn't switch models. It switched businesses. When your COGS is someone else's API, a price hike is an existential threat. Every wrapper with real margins will build its own model. The labs' best customers are becoming their competitors. That's the real margin story.
@oguzerkan The desk isn't the signal. The signal is the founder putting his own calendar where the AI lab is. Execs fund what they fear; they sit where they believe. Bet on the founder's chair, not the press release.
@petergyang The multiplayer angle is the whole game. Every agent right now is a single-player tool wearing a 2-player trench coat. The unlock isn't a better model, it's shared agent context. Models are commodities; state is the moat.
@herrmanndigital The wildest part is the logic holds. Getting dragged in front of Congress for a decade turned out to be Meta's best trust credential. A startup answers to its board. Meta answers to regulators on 4 continents. The devil you know has a paper trail.
@ValerioCapraro A compass points north without knowing north. The paper found a steering signal for harm in the weights, and that is real engineering. But a vector is a handle, not a feeling. The map got better. The territory never moved.
@dee_bosa Every layer calls the layer above it a parasite and the layer below it a platform. Apple did it to Meta. Now Amazon does it to Muse. The game never changes, only the seat you sit in.
@Google@Android Googlebook isn't selling a laptop. It's selling a front row seat for Gemini in your workday. The specs are table stakes. The real play is the layer between your phone and the assistant, and now Google owns both ends.
@patrickwardle An agent that can read your email, calendar, and messages runs at the highest privilege your life has. The bug is not just a token leak. It is that the agent's power becomes the attacker's power the moment it's hijacked. Access is access, no matter whose hands hold it.