🚨 Microsoft signs back-to-back multibillion-dollar AI infrastructure deals with Lambda and IREN within hours—revealing the frantic race to lock down compute capacity. The speed of these deals shows just how desperate the hyperscalers are to secure power.

The sexiest part of building an app is the build. The boring part nobody talks about? The documents that come before.

Specifically: ADR to PRD to plan.

ADR = Architecture Decision Record. PRD = Product Requirements Document. The plan is... well, the plan.

Sounds corporate and soul-crushing, I know.

But here's what I've noticed after watching 50+ app builds over the past year: The ones that actually ship and get used start with this sequence. The ones that fizzle out or get rebuilt 3 times?

They skipped straight to coding.

An ADR captures why you're making technical decisions. "We're using Supabase instead of Firebase because our team knows SQL and we need row-level security." It prevents you from forgetting your own reasoning three weeks later.

A PRD defines what you're building and for whom. Not features -- problems. "Customer service reps waste 2 hours daily searching for order history across 3 systems. We're building a unified search that pulls from all three."

The plan maps out how you'll actually execute. Breaking the PRD into buildable chunks with dependencies mapped out.

The ratio that keeps showing up? 60-70% planning, 30-40% execution.

One founder I know spent 6 hours documenting his workflow before writing a single line of code. The build itself? Done in 4 hours using Cursor.

... aren't moving faster. They're coding confidently in the wrong direction, then rewriting when reality hits.

"We'll figure it out as we build" works for personal projects. For anything that involves users, money, or multiple people? You're just procrastinating the hard thinking.

The counterintuitive part: Good planning documents make AI coding tools way more effective. When you can point Cursor at a detailed PRD, it builds exactly what you need. When you feed it vague vibes, you get vague code.

What's next: Tools like Notion and Gamma are adding AI features specifically for turning voice dumps into structured PRDs. The format is becoming standardized even as the creation process gets faster.

BTW: The best PRDs I've seen are 1-2 pages max. If you need 20 pages to explain your v1, you're not building v1 -- you're building v3 without having learned from v1 and v2.

There is no better time to learn how to use AI to build solutions businesses want. I’m hosting a workshop that will show you, step-by-step, how to build tools and solutions using AI without having to know how to code.

If you’re interested, reply back to this email with the keyword BUILD and we’ll send you more info.

We are disclosing the dates soon. So if you’re interested make it known.

The Recap: OpenAI CEO Sam Altman and Microsoft CEO Satya Nadella publicly admitted they don't know how much electrical power AI development will actually require—yet are committing tens of billions to infrastructure anyway. This stunning confession reveals that the industry's massive capital deployment is essentially a bet made blind, potentially exposing investors to unprecedented risk.

Unpacked:

Bottom line: When the leaders of the two most influential AI companies admit they're guessing on their biggest cost driver, it signals either remarkable honesty or a fundamental market immaturity. The real question isn't whether AI will need more power—it's whether the industry can credibly plan infrastructure spending without understanding the basic resource requirements. This knowledge gap could turn into a massive capital allocation disaster.

Elad Gil maps which AI markets already have clear winners versus the vast territory still wide open—essential reading for anyone placing bets in the space.

a16z pauses its diversity-focused TxO fund and lays off staff—signaling a strategic retreat from underserved founder programs amid market pressure.

Studio Ghibli and Japanese publishers demand OpenAI stop training on their work—escalating the global copyright battle beyond US borders.

GM joins the 'eyes-off driving' movement without clarifying liability—automakers are racing toward Level 3 autonomy with the legal framework still undefined.

Waymo accelerates robotaxi expansion to Detroit, Las Vegas, and San Diego—Alphabet is aggressively scaling autonomous vehicles while competitors stumble.

Analysis suggests AI is still in its 'dial-up era'—a contrarian view that current infrastructure constraints parallel the pre-broadband internet, implying decades of buildout ahead.

P.S. Reply back to this email and let us know what area of AI you are struggling with the most. We are going to be hosting some free trainings and want your input.

Best of AI™ Team

Was this email forwarded to you? Sign up here.