Agile vs agile vs the world. REAL-LIFE SECRETS OF BUILDING SOFTWARE THAT DOESN'T SUCK The fastest way to build ANYTHING is to just build SOMETHING. The fastest way to build SOMETHING GOOD is to build it, then loop on using it (or get someone to use it) while building it better

Wildfires suck and should be systematically avoided whenever possible. Check. If your best idea is to shut the grid down for the entire foothills then you are underthinking the solution.

So many angry friends don't understand why anyone would target NCAR. It's not a question of whether NCAR has deep science chops, they do. But, it stops being science when it stops being skeptical and only pursues one narrative.

Geniuses here in Colorado shut down the entire electric grid across the foothills for what turned out to be a calm/slightly breezy day w/a random 53 MPH gust. 1-3 days to reenergize. Intersections=dark, traffic=snarled, food=spoiling, home healthcare=off, ... Dumbasses

As an Nvidia shareholder, pivoting production away from game GPUs to AI infrastructure is tactically smart. As someone who uses a lot of "game" GPUs for cloudless AI development...grrr! Nvidia might one day and find the server market cold and its former gamer customers colder.

Ukraine just took out a Russian submarine (smaller one) in port using underwater "baby" drones. What the U.S. should take from this is that expensive strategic assets that were reliable instruments of power in the 20th century are efficiently offset by 21st century technology.

The federal government plans to hire 1,000 AI engineers for 2-year stints with promises of jobs at Apple, Google, and Nvidia after serving. PREDICTION: Fail. The government doesn't know what to do with AI engineers. It'll be lots of POCs with nothing accomplished.

Sadly, this guy is right. The types of 20th century “hard power” systems that made the U.S. a global superpower are increasingly offset by wide-proliferation 21st century technologies. The U.S. continues to prepare for 20th century warfare when that page has already turned.

Major portions of the new U.S. National Security Strategy were clearly AI-generated (not that there's anything wrong with that). Although, it opens the administration to questions about whether AI is making U.S. national security policy. Can you spot the clues?

🇺🇸🔥 BREAKING: America’s 2025 National Security Strategy Just Dropped - It’s a Full Ideological Reset The White House has released the 2025 National Security Strategy and to me it’s one of the clearest windows yet into how Washington actually sees the world. This isn’t the usual

This guy is awesome. N'uff said.

All these tech companies are seeing the datacenter as an endless money printing machine. Nvidia no longer takes game hardware seriously. Crucial is winding down operations. In two years, after the datacenter crash, they will be groveling for any consumer business they can get.

$IBM CEO says that at today’s costs it takes about $80B to build & fill a 1 GW AI data center, so the ~100 GW of announced capacity implies roughly $8T of capex & “no way you’re going to get a return on that,” since you’d need “about $800B of profit just to pay for the interest”

The need is to compete with China on a balance of power level. Does this answer that need? Ahem...China is mastering physical AI to build fully autonomous battle droids. Genesis Mission feels more like bringing a knife to a gunfight.

The Secretary of Energy will be in charge. This gets the national labs involved. That seems okay until you look under the hood and realize how little national labs get done.

This violates my first rule in AI development. RULE #1: Never start a new AI initiative with an infrastructure project. Infrastructure projects are easy and distract from doing the hard part. Do we really need this dataset? I doubt it.