Microsoft writes 20-30% of its code using AI. Coinbase is at 40%. Content, design, and marketing copy are heading the same direction. "We build fast with AI" is no longer a differentiator. It's everyone's baseline.
And yet something strange is happening. Speed has equalized, but performance gaps are widening. Something else is deciding who wins.
The belief: AI = Speed = Competitive Advantage
After the AI tool explosion, the productivity narrative ran one way: the faster you build with AI, the more ahead you are. And there was data to back it up — a Workday study found 85% of employees using AI tools saved 1-7 hours per week.
So teams optimized for output. Code faster, more content, quicker designs. Speed was the winning strategy, or so the thinking went.
But the numbers point the other way
The same study buried an uncomfortable finding. 40% of those time savings vanish immediately into AI error correction and rework. Only 14% consistently report a net-positive outcome.
Then there's the trust data. More than 50% of US workers have received low-quality AI-generated content from a manager.
A term emerged: "AI Slop." Low-effort AI-generated content dressed up in quality's clothing. Built with speed, but destroying trust in the process. And trust, once broken, doesn't come back fast.
The ripple effect of leadership AI slop
When leaders let low-quality AI content slide: 57% of employees experience weakened AI trust, 51% see productivity decline, 46% report company reputation damage.
Speed is already free. The remaining bottleneck is the ability to judge what to build, whether the output is good enough, and when to stop. That's taste.
The core shift
The competitive bottleneck moved from "production" to "judgment." Filters are rarer than generators. Curators are rarer than creators. Editors are rarer than engineers.
Why taste is the last moat
Code can be cloned overnight. But a trusted experience built on consistent aesthetic principles can't. Apple has maintained "technology as humane" for decades. Airbnb manages everything from receipt font to error message tone intentionally. Every Figma animation and icon is deliberate.
Healthcare AI offers the clearest example. Epic and Abridge are both AI medical record solutions. Epic earned ridicule from clinicians for its complex, low-trust UI. Abridge received praise for being "beautiful and trustworthy." The difference wasn't in the technology specs — it was in design taste.
"AI can amplify output. But it cannot amplify judgment. That remains a human domain."
— The VC Corner
Taste operates on three levels. Recognition is knowing which of two things is better before you can explain why. Compass is having an intuitive sense of the right direction for something that doesn't yet exist. Vision is predicting what will matter in two years and calibrating current decisions accordingly.
| Speed-first | Taste-first | |
|---|---|---|
| Strategy | More, faster | Less, better |
| AI usage | Maximize output | Curate by judgment |
| Output risk | AI slop | Trust compounds |
| Replicability | Immediately cloneable | Embedded in culture, uncopiable |
| Long-term moat | None | Compound growth |
How to build taste — 90-day playbook
OpenAI's Emma Tang put it plainly: "Anyone can become a 10x engineer — if they have good software taste." Taste isn't innate. It's practiced.
- Study the gold standard (weeks 1-4)
Each week, pick one product, piece of content, or service you consider best-in-class and write down why in specific, mechanical terms — not just "it feels good." After 30 days, patterns start to emerge. - Put them side by side (weeks 5-8)
Once a week, pick two similar things — meeting notes, app onboarding flows, marketing emails — and write 500 words comparing them. The goal isn't "this is better" but "this specific mechanism makes it better." - Edit before you add (always)
Before adding anything, ask "what happens if we remove this instead?" One engineer removed an entire search feature and served core customer needs better. Editing is harder than creating — and that's exactly what makes it taste. - Be able to explain your intent (weeks 9-12)
If you can't immediately answer "why did you do it this way?", you built from habit, not taste. In week 9, redesign one thing you made — from first principles, not convention. - Share your taste publicly
You only truly own it when you can teach it. Short post, team share, internal doc — any format works. Taste becomes clearer when articulated, and becomes culture when shared.
Want to go deeper?
Taste Is the New 10x The original piece on why judgment, not speed, is the competitive edge in AI-era engineering parul.substack.com
Why Taste Is the New Moat The Epic vs Abridge case study and economic analysis of the aesthetic moat thevccorner.com
How to Be a 30x AI Engineer with a Taste The three layers of taste and the original 90-day development playbook pakodas.substack.com
Leaders who hand AI-generated workslop to employees may risk eroding trust Concrete data on how AI slop erodes workplace trust hrdive.com
The Last Competitive Advantage In Software Is Not Software Forbes analysis of why taste is the last defensible moat in software forbes.com
Why Taste Is The One Skill AI Cannot Copy Analysis of taste as the final human competitive advantage that AI cannot replicate forbes.com
