Andy Merskin

👨‍💻 Product + UX / UI and Full-Stack Engineer
🔧 20+ years building professionally
🔭 Looking for work
✨ Let's build your dream product together!

[Andy Merskin @andymerskin]

@skrug 4th edition is almost out, I'm excited to see what's new!

https://t.co/2M8AzbZR6B

[Andy Merskin @andymerskin]

We really need everyone to read "Don't Make Me Think" by @skrug again. In the age of AI, everyone is way too verbose. Where did all these walls of text websites come from?

Why can't I figure out what your product even does, above the fold? Why do you need 50 paragraphs of slop?

[Andy Merskin @andymerskin]

@elthecreator7 @theo Did anyone really think Theo would take this well with his fragile ego, even if it was intended to be illustrative? 😅

[Andy Merskin @andymerskin]

@steventey @typesafeai @aisdk This is comedy gold 😂

[Andy Merskin @andymerskin]

@ggsimm I LOVE THESE 😂

[Andy Merskin @andymerskin]

@adamwathan Huge congrats! Tailwind is important to preserve and maintain, it's made such a massive impact on development speed and ergonomics.

[Andy Merskin @andymerskin]

@kettanaito Uhh, yes please! The worker setup is a lot.

[Andy Merskin @andymerskin]

@LucaAmesDesign @shadcn And then an accessibility nightmare was started 😮‍💨

[Andy Merskin @andymerskin]

@shadcn @aidenybai NICE! Love this. Thanks for sharing it

[Andy Merskin @andymerskin]

@MansiCodez I say do both. Staging is useful for demos and testing that go beyond feature flags.

New feature-flagged work often requires refactors to old code that isn't flagged anymore, and not all test suites are complete, so staging is the place you would catch those remaining gaps.

[Andy Merskin @andymerskin]

@theo TypeScript removes an entire category of unit tests needed to validate code works. Safety at the compilation step is just one essential layer protecting production from runtime nightmares, and saves you (and agents) from debugging nearly as many traces.

[Andy Merskin @andymerskin]

@letdarky @mattpocockuk Or put another way, I get why people don't want everything to read like an ad.

The overuse of LLMs complicates everyone's unique ways of expressing themselves if we lean too heavily on them.

[Andy Merskin @andymerskin]

@letdarky @mattpocockuk Ah I misunderstood you, you're talking more about writing and prose in general. It seems like a contentious argument, since models were trained arguably on the best writing styles we knew prior to LLMs, but now they feel overused.

Then again, so much reads like marketing speak.

[Andy Merskin @andymerskin]

@letdarky @mattpocockuk I agree. I feel like this discussion comes up every week in perpetuity, like nobody's actually solving it.

I think it's partly a symptom of each codebase having their own patterns, preferences, and established idioms, so naturally agents need to remain flexible.

[Andy Merskin @andymerskin]

@buildwith_yash @mattpocockuk Lower surface area of bad patterns for agents to repeat.

The fewer the anti-patterns, the less likely for agents to treat them as anchors across a project as it tries to honor what's already there.

[Andy Merskin @andymerskin]

@webdevcody Grouped commits is where it's at! I have a dedicated `/commit-all` skill that has the agent check for topics or phases across the working tree and the chunks it out with Conventional Commit messages so everything's nice and purdy. ✨

[Andy Merskin @andymerskin]

@AcdNrg @trikcode Speak of the devil, Grok Bot was opened to Cursor Pro+ plans ($60/mo) and now I've got access. Been playing with it today for job search stuff, it's been incredible!

[Andy Merskin @andymerskin]

@AcdNrg @trikcode Unfortunately it's only available for SuperGrok Heavy or Cursor Ultra users, but not the mid-tier plans.

[Andy Merskin @andymerskin]

@0xDesigner I want an infinite canvas with brief, generated artifacts that help me navigate projects spatially, and each artifact can be interacted with as chats.

Each artifact would maybe capture chunks or concerns across a project to group navigation points and organize larger pieces.