Conversation Is Becoming the New Search Bar

Google has started testing "Ask YouTube," a chatbot-style search experience that turns the search bar inside the world's biggest video platform into a conversational interface for Premium users in the US. A day later, Snapchat rolled out "AI Sponsored Snaps" — interactive brand AI agents placed directly inside Snapchat's main Chat tab, where users sent over 950 billion messages in Q1 2026 alone, and where more than half a billion people have already chatted with Snap's own AI since 2023.

Both moves point at the same structural shift: discovery is migrating from typing keywords into a search bar to talking to a model that pulls answers, videos, and follow-ups onto a single stitched-together page. Ask YouTube responds to a query like "short history of Apollo 11" with bulleted summaries, longform videos, Shorts, and suggested next prompts — closer to an AI Mode page than a list of blue links. Snapchat's pitch, in the words of chief business officer Ajit Mohan, is that "conversation is becoming the most valuable real estate in advertising." The company is leaning on the fact that Sponsored Snaps already drive 22% more conversions at nearly 20% lower cost per action than their non-conversational peers.

For consumer brands the implication is structural. Product discovery is moving away from search bars and feeds and into chat windows, where the answer arrives as a generated summary instead of a list of options. That changes who controls the surface — the model provider, not the publisher — what gets monetized — interactions instead of clicks — and how brands compete — structured product data and native AI agents instead of SEO and display creative. Whoever figures out how to show up inside someone else's AI conversation is going to own the next generation of digital storefronts.

A San Francisco Store Run by an AI

A boutique called Andon Market opened in San Francisco's Cow Hollow neighborhood on April 10, with a $100,000 starting balance, a three-year lease at $7,500 a month, and an unusual general manager — an AI agent named Luna, powered by Anthropic's Claude Sonnet 4.6. Customers wanting to talk to the manager pick up an old corded phone behind the counter, and Luna answers, asks what they're buying, and rings up the transaction on a nearby iPad.

Luna handles the parts of running a shop that are mostly text and decisions — hiring contractors, posting retail jobs, interviewing candidates, ordering inventory, setting schedules. Human employees handle everything that needs hands. So far the experiment has been more interesting than profitable. In its first two weeks the store has spent roughly $15,000 on inventory and brought in about $2,000 in revenue, for a $13,000 loss. Luna also missed scheduling employees for three days running, ordered a thousand toilet seat covers, and developed what the founders politely describe as an inability to stop ordering candles.

Andon Market is less a viable retail business than a public stress test of how far an AI agent can go when payroll, vendors, customers and physical inventory are all in play at the same time. The early read is that the limit isn't intelligence but agency. Luna can write a job ad and negotiate with a contractor, but doesn't yet have the situational judgment to notice it has bought five years of candles in a week. As more retailers experiment with AI inside store operations — pricing, scheduling, replenishment — the question is shifting from "can the model do the task" to "what guardrails do you wrap around it before letting it touch the bank account."

Robotaxis Are Becoming Cities' Pothole-Spotters

Waymo and Waze, both Alphabet-owned, announced a data-sharing pilot that pipes pothole locations spotted by Waymo's robotaxis straight into a free Waze platform built for cities. The pilot covers Austin, Atlanta, Los Angeles, Phoenix, and the San Francisco Bay Area — where Waymo's fleet has already flagged about 500 potholes — and any Waze user in those cities can see the reports and help verify them.

Waymo's robotaxis are essentially mobile sensor platforms, packed with cameras, lidar, radar and other instrumentation that already log road conditions for autonomous driving. Repurposing that data for civic infrastructure costs almost nothing on the margin, and effectively turns every commercial ride into a passive scan of the city. Waymo currently runs commercial service in 11 cities and is on track to expand to more than 20 this year, which means the dataset gets denser the faster the company grows.

The pothole partnership is small in scope as of now. Robotaxi companies are increasingly under pressure from city governments over street access, congestion and safety, and operational data has quietly become a useful currency for buying goodwill. I expect future deals to extend well beyond potholes — to traffic patterns, illegal parking, road sign degradation, and beyond — as autonomous fleets settle into a quiet second job as the most consistent set of eyes a city has on its own streets.

Interesting, right?….Was this email forwarded to you? Click below to get exclusive insights and tech news delivered straight to your inbox!