Claude Fable 5 Is a Mythos-Class Model for General Use

Claude introduced Fable 5 - a Mythos-class model they made safe for general use, with capabilities above any model they have ever made generally available. Fifty-seven million views on a model drop, and if Mythos-class is the new floor, the next decade of software gets written by something that already exceeds everything you have been allowed to touch.

Check the post here.

Elon said Tesla AI chip design reviews are great, the team is awesome, and AI6 might set a record for usable intelligence from a wafer when you factor in yield. Fifteen million people saw the chip post, and wafer-level intelligence per yield is how Tesla stops buying someone else's brain over the next ten years.

Check the post here.

Elon told Jon Erlichman he thinks SpaceX might reach about $1 trillion in revenue in 2030. A rocket company at a trillion of revenue in four years is an industrial state, and that number rewrites what defense and compute even mean.

Check the post here.

I said the American empire is just getting started: the US has a comically large SpaceX lead on the next defense frontier that nobody closes for at least 20 years, it leads AI in models and chips, and free markets plus capitalism plus free speech are the mechanism that keeps that lead. China can catch models - it cannot copy brutal company competition and AIs that are allowed to become smarter than the humans who built them.

Check the post here.

I said SpaceX should put Starlinks around the moon, drop a Cybertruck on the surface with a retractable solar panel, let it FSD anywhere it wants, and livestream it back to Earth. A roving truck on the moon with a live feed is how a civilian audience starts treating cislunar as real estate, not a NASA special.

Check the post here.

I said if you are angry Elon is now a trillionaire, you have the wrong mindset - the global economy is about $100 trillion a year, global assets are about $1 quadrillion, and his net worth is 0.1% of that asset base. Anger at 0.1% is a hope problem, and it will keep people on the sidelines for the entire physical-AI buildout.

Check the post here.

I said the first phase of Tesla's planned fab is bigger than the entire CHIPS Act budget, the AI5 chip that just taped out will not be sold outside Tesla, and that is the part of vertical integration NVIDIA cannot price into forward guidance. A captive fab that size, over the next ten years, is how Tesla's inference cost falls on a curve no merchant chip vendor can match.

Check the post here.

I said merge SpaceX and Tesla, make Gwynne Shotwell CEO of the new entity, and let Elon go ultra deep into psycho engineering. One operator across the stack and one engineer at the limit is how you get a twenty-year industrial empire instead of two public-company calendars fighting each other.

Check the post here.