In the era when development became free, we now put a price on 'taste'

Josh's NewsletterJosh

AI summary focused on the core points of the original article.

In an era where implementation has become easier, the power to decide what to make and how to show it becomes more expensive.

•

Good reading when

•

AI prototyping feels like it is overtaking team decision-making

•

You want to organize the changes as the boundaries between designers, PMs, and engineers blur

•

You want to see the direction in which agent tools like Codex are expanding into knowledge work tools

1. What has become more expensive than implementation is taste and curation

When AI lowers the cost of feature implementation, the bottleneck for product teams shifts from "can we build it?" to "what should we choose?"

•

Focus of product judgment

•

Curation that chooses the good from among multiple attempts

•

Structuring decisions about which features to combine and what to discard

•

The sense to choose the right medium among prototypes, documents, and experiments

•

What to look for in design reviews

•

Whether it is a plausible screen, rather than what stage of output it is at right now

•

Whether it is launch-ready quality, or the first trace of an exploration

2. Design in the AI era has become faster, but also easier to lock in

When highly polished prototypes appear too quickly, teams can easily mistake ideas that are still being explored for signals that are right before launch.

•

Dangerous illusions

•

The assumption that a visually ready screen is therefore product-ready

•

Situations where implementation pulls the team along before research and business judgment are complete

•

Agreements needed

•

Stage labels such as "This is for exploration"

•

Distinguishing between problems better solved by documents and problems better solved by interaction prototypes

3. Roles are not disappearing; they are changing by overlapping

The boundaries between job functions are blurring, but that does not mean the expertise and best practices of each field should be erased as well.

•

Changes in team operations

•

More situations where designers understand technical language, and PMs work with code

•

A perspective that sees people not by job boundaries but by the average of where they spend their time

•

Expertise that must remain

•

Deep decision criteria in product, design, and engineering respectively

•

The difference between "I can build it too" and "I make good product decisions"

4. Codex expands from a development tool into a starting point for work automation

The direction of Codex is less about trapping every function inside one app and more about serving as a home base that calls in the tools needed to start and finish work.

•

Interesting ways it is used

•

Creating a daily brief for product leaders by summarizing Slack channels

•

Setting scheduled tasks through voice in the app and coaching the criteria for the next run

•

Automating existing workflows by directly operating browser or desktop tools

•

Clues to the product direction

•

The boundary between developer tools and general knowledge work tools is blurring

•

You can understand why Codex and ChatGPT are getting closer from the perspective of the work surface

XPLORED View

This piece is great because it makes you think not about "how to build faster" with AI, but about what judgment to make in front of something that has been built quickly. Especially for designers, before being persuaded by the finish of a prototype, bringing this back as a lens to check the medium you need right now and the level of team alignment will be directly useful in practice.