AI's NIMBY Problem, 20-Min Coders, & Hard Choices

AI's NIMBY Problem, 20-Min Coders, & Hard Choices

The Throughline Brief · May 24th. 5 min 53 sec.

Cold open

A single AI data center uses as much electricity as one hundred thousand households. Think about that. Utility companies are passing those massive grid upgrade costs directly to consumers, fueling intense political pushback. AI is officially hitting the physical world. The local resistance is here. And it is about to stall your infrastructure. The physical constraints are catching up to software.

Intro

This is The Throughline Brief for May 24th. I went through every single major AI and tech podcast this week so you can save your time. Here is what actually mattered to founders and operators right now. I am West.

AI's Physical Wall

We need to talk about AI hitting a physical wall. Shaan Puri brought this up on My First Million. He pointed out a glaring issue for anyone building at the infrastructure layer. AI needs massive data centers and endless power grids. People are pushing back hard. We are seeing a fierce local resistance. It is stalling the exact physical infrastructure this industry desperately needs to scale. John Coogan expanded on this over on TBPN. He mapped out the math. A single AI data center draws the same electricity as one hundred thousand households. The kicker is who pays for it. Utility companies are passing those grid upgrade costs down to everyday consumers. They are completely shielding the trillion dollar tech giants. This is pure fuel for political backlash. We are already seeing proposed bills from Bernie Sanders and AOC trying to cap this. The supply chain is getting just as geopolitical. Jacob Helberg explained a wild initiative on No Priors called Pax Silica. It is a fourteen country coalition building a forward deployed industrial base on a massive four thousand acre diplomatic property in the Philippines. The goal is securing the physical AI supply chain. Critical components like precision reducers and rare earth magnets are moving to allied soil. If you are building in AI, look past the software layer. Power, land, and international supply chains are the new bottlenecks. The physical constraints are very real.

The Speed of Code

Let us pivot to the software layer, because the speed of shipping is completely breaking traditional workflows. Ryan Nystrom dropped a crazy stat on How I AI. At Notion, they are using an internal AI agent called Codeex. A request for a tab block went from a natural language prompt to a fully complete pull request, including UI verification, in under twenty minutes. Twenty minutes. Zero humans in the loop for the heavy lifting. Ryan made a crucial point for engineering leaders. Fast continuous integration is now a critical requirement. If your CI is slow, AI creates massive bottlenecks. You are essentially throwing those PRs in the trash. Eric Jang gave us the theoretical underpinning for this on the Dwarkesh Podcast. He looked back at AlphaGo to explain how neural networks do something mysterious. They condense impossibly deep game tree simulations into manageable decisions. AI handles the intractable search problems so we can move fast. So how do you actually manage these systems. John Coogan had a fascinating take on TBPN. He advocates for a Cognitive Behavioral Therapy approach to AI prompting. Guide your AI with positive guardrails. Give it clear boundaries. Set it up for success by encouraging it. Treat it like a junior developer who needs context and confidence, and your output quality will skyrocket. The difference between demanding perfect code and coaching the model toward the right answer is night and day. You have to build intuition.

Mark Benioff

Let us move to the best moments of the week. Mark Benioff joined the All In Podcast and gave a massive reality check to founders panicking about software valuations. The industry has been bracing for an extinction event. Benioff offered a completely different view. He stated the SaaSpocalypse is a normal market rerating. Many companies are still trading at two times sales. This matters immensely if you are raising capital right now. The sky is fundamentally intact. Valuations are simply returning to historical norms. Founders must accept this new reality. Build real, durable businesses with actual revenue. Focus on the fundamentals.

Shaan Puri

Over on My First Million, Shaan Puri sketched out the future of human and AI collaboration. He painted a picture of an organization where software completely flips the traditional hierarchy. Shaan said the reality is that we work for the AI. This is a massive model shift. We are moving toward a world where AI acts as the central company brain. Humans will simply serve as contextual inputs and physical executors for that intelligence. If you are designing organizational structures for a startup today, you need to build around the AI as a core partner. Rethink your human roles from the ground up.

Shiv Rao

Finally, Shiv Rao sat down with Harry Stebbings on 20VC to talk about scaling Abridge to a five point three billion dollar valuation. He shared a brilliant philosophy on finding the right investors to back your vision. Shiv said, you know when you just have chemistry with a person, you just know you are going to find a way to work together. He literally stalked an investor for years to find that intellectual alignment. Finding founder partner fit is everything. You have to play the long game. Build relationships long before you ever need the check. Chemistry is what gets you through a five year wilderness building your company.

The bottom line

The future is being built by founders who see AI as far more than just a tool. It is a true partner. It is your organizational brain. And it is a demanding new layer of physical infrastructure that requires massive amounts of power and diplomacy. Build your company with those massive shifts in mind. Embrace the weirdness of this moment. That is your Throughline. See you next Sunday.