Pixel 11 costs $100 more with the same screen. Google is selling software now

Google shipped four new phones this week, and the hardware barely moved. Same screen sizes, a chip that is more efficient, a slightly thinner camera bar. The prices went up $100 across the board. So why launch at all? Because the real product this year is software you can turn on, not glass you can hold. Let me catch you up.

The deep cut

• A mature product bets on software, not specs. Google raised Pixel 11 prices $100 while shipping the same screen sizes as the Pixel 10.
• Style beats perfection when the market shifts taste. Isaac Reynolds built Camera Looks to give people worse photos on purpose.
• Accessibility can be a real feature, not a footnote. Google's SL2T turns ASL into text on the Pixel camera for 70 million Deaf users.

The spec sheet ran out of road

Look at what actually changed in the metal. The Verge called it "one of those refinement years," and the numbers back that up. The Pro display went from 3,300 nits to 3,600. The chip is 20 percent more efficient. The base storage doubled to 256GB, which is why the Pixel 11 now starts at $899 instead of $799.

TechCrunch put it plainly: fewer hardware changes, a lot more Gemini. When your hardware is good enough, you stop selling the screen and start selling what the phone does. That is the whole pivot, and it is worth studying no matter what you ship.

Making the camera worse on purpose

The camera story is the sharpest lesson here. Phone photos got so good they got boring. Bright faces, blue skies, no noise. Reynolds, who leads the Pixel camera team, said the gap between what two people want from a camera "is growing dramatically." Younger buyers are paying real money for cheap old digicams. Search for "digicam" is up 500 percent in five years, per Google Trends.

So Google built Camera Looks. It changes how the image gets processed at the sensor level, not just a filter slapped on top. Pick a minimal look and the phone does less frame stacking, less sharpening. The photo is technically worse and closer to what the person actually wanted.

The team knew when to stop simplifying. Reynolds said he had "an escalation meeting" with his PM and UX team about how many controls to add, and still landed on six or more levers per look. He kept the old default first in the list, named "Original," because studies said it was still right for the base case.

Chasing the person filming alone

Google is also building for a specific buyer: the solo creator with a full-time job. The Creator Suite puts a teleprompter, a vocal enhancer, frame guides, and auto folder sorting right in the camera app. Reynolds pointed at the "hundreds of thousands of creators" at the bottom of the market who want fewer retakes and faster publishing.

This is a clean read on who to build for. Not the top creators with teams and gear, but the ones striving to post daily around a day job. Pick a real person, learn their workflow, remove the friction points. That is a roadmap move you can copy on Monday.

The feature nobody else shipped

The piece worth bringing to your next review is the sign language work. Google DeepMind's SL2T model turns American Sign Language into text right in the camera and Gboard, so a Deaf user can sign anywhere they would normally type. It runs on 100,000 hours of training data across 50-plus sign languages and hits a 70 BLEURT score, well past prior work.

What matters is how they built it. Deaf perspectives shaped every stage, from a Deaf Googler conceiving it to a formal advisory committee governing the release. They also drew a hard line: Google calls it "strictly an assistive input tool for low-stakes communication," not a replacement for a human interpreter, and named the failure modes out loud, from ghost text to low-light drops.

That honesty is the design move. They shipped a feature that helps 70 million people and told you exactly where it breaks. Compare that to how you scope and caveat your own AI features.

Three questions for your team

• Where is our hardware or core spec already good enough, so the next win has to come from software the customer turns on?
• Are we still optimizing for a "perfect" default when a growing slice of our users want something rawer, and would we let them choose the worse-looking option?
• When we ship an AI feature, do we name where it breaks as clearly as Google named ghost text and low-light failure, or do we bury the limits?