Written from the Oakland airport, on my way to L.A. to attend a live show of Dimension 20 with my good friend Patrick.

God, time has flown by. Is this really the end of my fifth week at Vercel? Feels like a few days. That’s a good sign, I suppose!

Work

That said, I’ve noted a tendency to work late the last couple weeks, with the intention of making sure I make a notable impact in my first few (higly-leveraged and impression-making) weeks at work. This is a trade-off that I’m consciously and deliberately making, and I’m alright with it; but I do want to keep an eye on it and make sure that it doesn’t become a habit. As a wise woman once told me, “work isn’t real life” - and work will naturally expand to fill all the time you give it. It’s a gift to have a job that is compelling enough that you want to spend time on it; but it’s also important to me to focus on, and fence off time for, the other aspects of my life.

All About AI

Tool Use

The other day, I threw a personal project (an ELO tracker for Magic EDH decks) through Vercel’s v0 tool, asking it to convert the website from Python to TypeScript (so I could use it as a testbed for experimenting with Next.js and the Vercel platform). The results were…alright, to be honest. Despite being given a link to the source code, it wasn’t able to read the code, but instead just screenshotted the actual deployed site, and the resulting code was more of a generic “EDH tracker” than “a translation of this particular site” - some intentional specific design decisions were ignored (e.g. ELO score for a deck, not for a player), and the visual design looked completely different (admittedly - much better, but different!).

I spent an evening tweaking and shaping it with Claude-Sonnet-on-Cursor¹, and, honestly, the experience was a mix of inspiring and depressing. Inspiring with how much the tooling has improved in a couple years since I first tried it; depressing for two reasons:

• How obsolete my knowledge and experience will soon be. This is something I can realistically change and work on - adapting to learning a new skillset is a challenge, but a worthwhile one!
• How obsolete “understanding of the underlying systems” will soon be to the act of creation. This one’s a little more complicated - I don’t philosophically agree with anything which gatekeeps or obstructs the act of creation, but I’m also uncomfortable with a deliberate celebration of ignorant creation². Wiser minds than I have spilled miles of pixel-ink on how AI amplifies experience, with all that that implies about “if you are inexperienced, it will amplify your inexperience”. Leaving aside any questions about whether AI will, in fact, “take our jobs” (my bet is - some, but not most), I do wonder about how maintainable the outputs of Vibe Coding will be. Especially relevant as I’ve been coaching a friend through the process of learning coding in order to build a system to drive her own medical care, where the AI’s positively encouraged her into some…interesting…choices (mostly, I suspect, driven by the tooling’s complaisant agreeability). We shall see!

Attitude

Driven by this article, I had an interesting realization - after Crypto and NFTs and Web3 (and probably others I’ve forgotten), I have adopted an irrationally-hostile perspective to any claims about technology from any company. That is - if a corporation had tried telling me “ChatGPT is able to search the web for new information, post-training”, I would simply assume that that was a lie. Even for a proud anti-corporatist, this isn’t a healthy or productive mindset - it’s a core value of mine to make sure that my criticisms are accurate and founded in fact, lest I undermine legitimate critism by mixing it in with ill-informed supposition.

• AI
• EDH-ELO
• Vercel
• Weeknotes