
Reality Before AI, Taste Before Consensus
In today’s newsletter, I’ve got a lot of great content for you, including:
- Why building faster can hide the real problem
- UI sounds without audio files
- When uniformity quietly kills creative distinctiveness
- And 7 more
Most Interesting

Punch Yourself in the Face with Reality
AI makes it easier to avoid the real bottleneck in startups: facing reality. The author warns founders against using AI to build endlessly without talking to users. The hard parts of success stay the same.
Design + Development
Tiks: UI Sounds Without Audio Files
Tiks generates UI sounds like click, success, and toggle on the fly with the Web Audio API, skipping audio files entirely. It’s about 2KB gzipped, has zero dependencies, and adds instant audio feedback to web interfaces.
Rogue Amoeba’s Paul Kafasis says MacOS Tahoe’s forced squircle icons hurt third-party apps. Golden Gate makes Apple’s own icons better, but everyone else’s gets flattened, losing usability and creative distinctiveness in the process.
Murphy Trueman’s take: design systems should borrow evals from LLM development, actually testing whether AI agents follow component rules, token conventions, and accessibility requirements instead of just hoping generated code respects the system.
NameThatUI Gives Every UI Element a Name
A visual dictionary of UI elements, with plain-English names for interface parts across web and macOS designs.
Tech + Innovation
The Future Worth Building is Human
Thinking Machines lays out its mission to build AI shaped by distributed human knowledge rather than centralized training. Draws on Polanyi and Hayek to argue tacit, local expertise resists aggregation, and outlines technical directions including customizable models and richer human-AI interfaces.
OpenAI Puts a Physical Keypad on Your Desk for Codex
OpenAI and Work Louder built Codex Micro, a programmable keypad for Codex coding agents. Dedicated keys, LEDs, a reasoning dial and a joystick give developers physical control over agent workflows and code review, per Engadget’s Ian Carlos Campbell.
Work + Mindset
Hardik Pandya breaks down why design carries unique costs compared to engineering and product management: unbounded inspiration, non-repeating breakthroughs, and universal legibility that invites unsolicited opinions from everyone, constantly.
Learning to Code is Still Worthwhile
Steve Krouse, founder of Val Town, argues that coding remains valuable for educational reasons even as AI writes code for us. Drawing on Seymour Papert’s LOGO language, he frames programming as a creative language like literature or math, worth learning regardless of job prospects.
A UX Collective essay arguing that AI and organizational design-by-committee share the same flaw: both excel at identifying preferences but fail at judgment. The author explains why taste requires context, intent, and someone accountable for deciding what deserves to exist, not just what’s statistically preferred.

Resources & Market Signals
Edition #120
Design Systems Meet AI, Process Evolves
Edition #144


