Nicholas Zinner - Future Shock Newsletter (Page 3)

Watercolor illustration of an AI workflow demo becoming a larger review room with papers, checklists, stamps, cables, and human reviewers.

AI Agents

The Agent Demo Ends Too Early

Agent demos usually stop when the artifact appears. Real deployment begins when someone has to decide whether the work is ready, reviewed, reversible, and owned.

Watercolor illustration of a research workspace at evening. A wooden desk before a wall covered in pinned papers and red string. Brass lamp, empty chair, potted plant on the windowsill.

Editorial

The Shape That Showed Up

What we found, what shifted under us, and the four shapes we're watching from here.

Watercolor and ink illustration of residential AI compute boxes mounted beside a house, wired into the power grid with a research campus in the distance.

Bright Signals

Compute Wants Out

AI infrastructure is starting to split in two directions at once: open public compute for science, and distributed compute pushed toward the walls of new houses.

Watercolor illustration of a messy back-office workflow machine with invoices, filing cabinets, conveyors, an empty chair, and a small AI terminal.

AI Agents

The AI Automation Problem Is Mostly Not AI

Most companies do not need AI agents first. They need clean intake, permissions, handoffs, receipts, escalation, and workflow ownership.

Watercolor illustration of a harbor at dawn. Some sailboats have departed toward a golden horizon while others remain moored at the dock.

what-if

The K-Curve

The gap between AI power users and everyone else is widening. The interesting question is when the gap stops being a personal choice and starts being a structural requirement.

Watercolor illustration of a figure hunched over blueprints in a construction trailer, looking out at a chaotic, nascent building site

what-if

What If AI Is Just Bad at Most Things?

AI agents in 2026 might be where smartphones were in 2008: past the proof of concept, before the mass-market product. The people saying it isn't worth the hassle aren't wrong.

Watercolor illustration of a person walking through a street overwhelmed by signs and screens all demanding attention

what-if

What If Exposure Breeds Exhaustion?

The tech industry assumes more exposure to AI converts skeptics into adopters. Four independent data sources say otherwise.

what-if

We Made AI Agents Negotiate the End of the World

The same model behaved completely differently depending on who else was in the room.

Watercolor courtroom where legal documents dissolve into teal washes

Long View

Code Is Still Law

Lawrence Lessig argued code regulates as powerfully as law. AI model weights now do the same, and nobody can read them.

Watercolor illustration of a park bandshell where a conductor leads a small ensemble, but musicians have spilled across the park in breakaway clusters

Sci-Fi Saturday

When Your Swarm Disagrees

AI agents disagree. One wants to cancel your meeting, another insists you need it. Nobody built a tiebreaker. Welcome to the coordination problem that gets harder with every agent you add.

Watercolor: person facing BI dashboards on left, lightbulb gap center, chaotic agent terminals on right

What If

The (Missing) Layer Between You and Your Agents

AI products are scattered across apps, browsers, protocols, and coding tools. As companies rush to consolidate, visibility and management of your agents will become the priority.

Watercolor illustration of a small robotic figure climbing a steep desert cliff face with layered sandstone strata, sagebrush at the base, and purple mountains in the distance

Sci-Fi Saturday

The Imperfection Engine

A robotics startup discovers that broken data produces better robots. Two Cold War-era novelists already knew why.