Own the stack or own the problem.

That tension is showing up everywhere this week: startups building chips to escape supply risk, law firms pricing outcomes instead of hours, and record checks chasing anything that looks like infrastructure.

The companies pulling ahead aren't picking between software and service, they're collapsing the boundary. What's left open is whether full-stack means defensibility or just operational drag dressed up as moat.

Every missed onboarding step, delayed offboard, or broken provisioning handoff has a root cause: HR and IT aren't aligned. This guide gives both teams a shared framework for the full employee lifecycle.

Norm just quietly rewrote the law firm P&L. The nearly three‑year‑old startup raised $120M Series C led by Khosla Ventures at a $1.2B valuation, joining Harvey and Legora in the legal AI pile‑up.

Its bet: a full‑stack, AI‑native firm (Norm Law) that uses proprietary agents as first line, human attorneys as supervisors, and outcome‑based pricing instead of billable hours. It’s also building AI agents to supervise other AI agents, pushing “partner track” down into the model layer.

With Bain, Craft, Coatue, and large institutional LPs backing this hits $1.2B valuation, the signal is clear: vertical, regulated‑industry AI will be capital intensive and winner‑take‑most. Anyone selling generic “AI for lawyers” now competes with a firm that’s building AI-native law and the services business around it.

🎬 Why do Netflix shows still have black bars... on a widescreen TV?

You bought a 16:9 television. Streaming platforms make content specifically for TVs. So why are they still wasting part of your screen?

In this episode, Caya uncovers the 70-year format war between Hollywood and television—from the rise of widescreen movies and anamorphic lenses to IMAX, VistaVision, and the psychological trick behind Netflix's "cinematic" look.

In AI, the real moat is increasingly measured in nanometers, not parameters. DeepSeek is reportedly moving from model tuning to silicon spec, with a home-grown inference chip fabbed at SMIC, according to a reported chip design push that hasn’t been confirmed by the lab. The goal: cut dependence on Nvidia-class hardware that U.S. export rules can throttle at will.

Strategically, it’s the obvious next step after a shift toward domestic silicon like Huawei’s Ascend: move from adapting to local chips to defining them, starting where Chinese fabs are least behind; high-volume inference, not bleeding-edge training. The risk is execution, not intent; SMIC’s process lag and yields turn every design win into a manufacturing question.

For founders, the pattern is clear: own the chokepoint. DeepSeek’s bet is that, under rising export-control pressure, the real platform shift is vertical, from model to metal.

Scale your growth: Join names like HubSpot, Ahrefs, and Deel. Put your product in front of 85,000+ startup founders. Let's talk.