The marginal cost of energy going to zero is a national security issue.

Something is clearly broken. It’s time those of us who were able to live this out, figure out what to do to help the millions who aren’t anymore.

I’m very biased but I think @BrianYutko is a star. He’s technically brilliant, business savvy, thoughtful, funny and kind. A better person Boeing could not have found. Very bullish with him in the seat.

Jason brings up a really important point that we’ve found to be very true. That’s why we built a series of background agents into 8090’s Software Factory. Their goal is to constantly observe any changes, minimize drift those changes may cause and keep tasks at hand in sight at all times. For example: The system detects a spec change. A background agent then scopes an update, maps the code files and suggests a fix. We’ve implemented a quiet layer of agents to have your back throughout the process. It’s less about AI and more about foresight. These background agents are the quiet hero’s of Software Factory. Yesterday we started to release Software Factory to the thousands of users already on the waitlist. We want their feedback on how to make it more useful before Beta. Please give it a try and tell us if you like it. 👉

What is the correlation of public consumer brands between their stock price and their use of diversity in advertising? Cite hard data and be brutally honest.

Very useful to have Grok tackle questions in a deescalated, truthful and scholarly way… Dear Grok: What can explain why some people are so offended by ads featuring classically beautiful women? Be brutally honest and cite hard data.

In a parallel universe: - Tasks are scoped - Specs are linked - Code reuse is automatic - Agents do the grunt work In our universe? That’s just Software Factory. Alpha rollout starts Friday. 👉

Energy is the big bottle neck for AI. In both physical and software AI, it’s not just that the ingredients need to change but the entire recipe does as well: 1) we need infinite and marginally costless energy. This will mean an ensemble of energy sources working together that can produce energy TODAY (ie nuclear not really an option before 2032, building a Nat gas or coal plant with a multi year backlog for parts also isn’t a near term option until 2030+) which means we will need Solar + Storage because it can go online 12-17 months from being greenlit. No way around it. 2) but for energy storage to scale economically in light of Foreign Entity of Concern (FEOC)/Prohibited Foreign Entity (PFE) you will need to find domestic LFP CAM providers for the ESS supply chain. There are very few. 3) you will need to step down the overall power footprint of the data centers which means HVAC needs to be rethought - an entirely new kind of heat pump must be invented. This new device while having a superior profile will also need to eliminate the forever chemicals which are now outlawed and must be sunset. 4) the chips themselves need to be re architected for performant, power efficient inference. Memory design, c2c, cabling and all manner of other design decisions that work well for training won’t likely scale for inference if inference is 100x+ bigger than training. 4) in physical AI, after storage (see above) abundant REs are essential for any form of motion/actuation. But getting REs out of the ground, into an oxide then into an alloy that can be made into Permanent Magnets….are a huge exercise in energy. And the list goes on and on… . . . My point is that if you are focused on AI, you should start paying attention to energy as it will be the gatekeeper of progress/change over the next few years in AI.

Energy is the big bottle neck for AI. In both physical and software AI, it’s not just that the ingredients need to change but the entire recipe does as well: 1) we need infinite and marginally costless energy. This will mean an ensemble of energy sources working together that can produce energy TODAY (ie nuclear not really an option before 2032, building a Nat gas or coal plant with a multi year backlog for parts also isn’t a near term option until 2030+) which means we will need Solar + Storage because it can go online 12-17 months from being greenlit. No way around it. 2) but for energy storage to scale economically in light of Foreign Entity of Concern (FEOC)/Prohibited Foreign Entity (PFE) you will need to find a domestic LFP CAM providers for the ESS supply chain. There are very few. 3) you will need to step down the overall power footprint of the data centers which means HVAC needs to be rethought - an entirely new kind of heat pump must be invented. This new device while having a superior profile will also need to eliminate the forever chemicals which are now outlawed and must be sunset. 4) the chips themselves need to be re architected for performant, power efficient inference. Memory design, c2c, cabling and all manner of other design decisions that work well for training won’t likely scale for inference if inference is 100x+ bigger than training. 4) in physical AI, after storage (see above) abundant REs are essential for any form of motion/actuation. But getting REs out of the ground, into an oxide then into an alloy that can be made into Permanent Magnets….are a huge exercise in energy. And the list goes on and on… . . . My point is that if you are focused on AI, you should start focusing attention on energy as it will be the gatekeeper of progress/change over the next few years in the AI market.

I’ve had every car imaginable: Range Rover Mercedes Bentley Toyota Ferrari Volkswagen Honda And my Model Y is the absolute best, no close competitor.