Notes
The last 20% is the whole point
AI nails the first 80% of a product in minutes. The last 20% — hierarchy, edge cases, copy, point of view — is where taste lives, and it is still your job.
Somebody still has to care
AI can write your code, but it cannot care. The human filter — nitpicking, obsessing, fixing — is what makes an app memorable.
AI Design Engineering: How Product Teams Ship Faster Without Losing Quality
A practical guide to LLM workflows, agent-safe design systems, production UI, and human quality gates.
Why I left my job to start my own company
Why I left a comfortable job to build independently — and what I gained and lost.
What I learned from my first year of independent work
Hustle myths, money cycles, loneliness, scope creep, and the golden rule.
How to Export Metadata from MDX for Next.js SEO
Export metadata from MDX so Next.js can drive titles, descriptions, and canonicals.
The design engineering toolstack is multiplying
Designers now use twice as many tools as last year. Here's what that means for solo builders and design engineers.
Stay at the Great Line
A self-reminder about consistency. Showing up every day is the only thing that actually works.
The best AI tools are boring
Claude Code ships Bun rewritten in Rust — a boring, invisible upgrade. Here is why the best AI tools are the ones you do not notice.
Why I stopped chasing AI models
New models drop every week. Qwen 3.8, GPT-5.6 variants, Kimi K3. Here is why the solo builders who win are the ones who pick a stack and ship.
Design taste is the new moat for AI startups
Vibe coding means anyone can ship a prototype in hours. Design taste — knowing what to build and what to leave out — is now the only durable advantage.
AI slop is drowning the web
Merriam-Webster made it word of the year. 52% of new articles are AI-generated. Here is what AI slop is, why it is toxic, and how to avoid making it.
Why AI startups fail at design (and it is not about lack of talent)
Most AI startups fail at design because they treat it as the last step. Here is what actually goes wrong and how solo founders can fix it without hiring a designer.
The vibe-code gap is real
Vibe coding gets you a prototype in hours. Getting that prototype to production takes weeks. Here is why the gap is widening and how design engineers close it.
Taste engineering is the skill after vibe coding
Prompting got us an answer. Context engineering made it relevant. Taste engineering decides whether the answer is actually good.
The hardest part of building with AI is not the building
AI tools let you build anything in days. Validation, distribution, and judgment have not gotten any easier. Here is what solo founders should actually spend their time on.
Design taste infrastructure is the missing AI stack
New tools package design taste into files AI agents can read. Taste infrastructure is the layer every solo founder needs in their stack.
Distribution is the real bottleneck for solo founders
AI made building almost free. Getting users is the hard part now. Here is what actually works for solo founders in 2026.
What taste tools miss about taste
Taste tools capture your preferences in machine-readable files. But the hardest part of building a good product has never been knowing what good looks like — it is holding that standard across every surface, every edge case, every day.
AI has no taste — and that is your advantage
AI-generated products all look the same because models average what they have seen. Solo founders who override the default win on taste — not speed.
Your AI eval set is your taste, made measurable
An eval set is a list of real inputs with the outputs you consider good. It is also the closest thing AI product development has to written taste — and the discipline most solo founders skip.
AI products fail at the interface, not the model
Users rarely abandon an AI product because the model was wrong. They abandon it because the interface hides uncertainty, offers no way to steer, and treats failure like a bug.
Every solo founder is a design engineer now
Solo founders are 63% of new startups, and the best ones are design engineers who never got the title. Here is how to close the last 20% of the work AI cannot.
The AI aesthetic is a costume. Copy the two parts that are real
The AI aesthetic is now a named thing: sparkles, streaming text, thin icons, beige. Two idioms encode real constraints. The other four are a costume. Here is what to copy and what to skip.
Start Today.
Deciding to write every day. Not because I have something to say. Because I want to get better at saying it.
mosh desk
One person. Multiple tools. This is the machine.
When I became ramen profitable
Ramen profitable means covering the basics on your own terms — slowly, one step at a time.
The Founder Floor
The solo founder's biggest unlock in 2026 isn't a new tool. It's learning to operate on briefs and approvals instead of keystrokes.
AI-generated code is cheap. Production engineering is not.
Vibe coding gets you a prototype in hours. Making it reliable, secure, and maintainable takes a different kind of work — and AI has not made that work any faster.
Taste tools are the most important AI category in 2026
Half a dozen companies shipped taste infrastructure in a single window. That density tells you something about where AI building is headed.
Generative UI is a design problem
Interfaces that assemble themselves at runtime change what design means. Here is how design engineers should think about generative UI.
One-person startups are rewriting the rules of product taste
More than one in three new startups are solo founders. That changes what good design means — because taste can't be a team sport anymore.