Dear Craig,
I had a hotdog wrap which was tortilla with 2 hotdogs, relish and for some reason mashed potatoes inside. Weird but in a good way
Also waiting for the train for hours and hours as is everyone’s favorite pastime in Germany.
I enjoyed the hotdog wrap more than the rest
Of course AI is a fundamental change in the industry and an accelerant, and how much AI-generated code actually gets checked and read is a hot topic. Same with people feeling less valid because some or most of their code was generated.
But I don't feel like that is fully new, we have been using libraries, copying snippets, even more so with shadcn-style component approaches. Most don't write 100% from scratch.
Its not a 1:1 comparison. Ai code is fresh every time, so it doesn't fully translate to libraries. But there can also be bad or poorly maintained libraries and even more so when we copy code from other sources. How much can we trust how much do we validate and what are the validation mechanisms.
Same issue really, where do we put focus and do we have systems in place with the limits of our attention and time.
So this doesn't feel like a new issue to me, it has basically been like this forever. Approaches, checks and validations might change, but that's fine.
Great fireside chat and very down to earth AI takes by Werner Vogels at @WeAreDevs
Web development has always been a learning journey and AI is a great turning point but it also doesn't change the importance of fundamentals and decision making or responsibility. And also enjoying the process.
@aurorascharff@sebsilbermann I checked again and it is only an issue on ios mobile, it seems to be missing `ReadableStream[Symbol.asyncIterator]` specifically when using the sandbox and will likely be fixed in next safari release. Works everywhere else
Even though they have been around for a while suspense and "use" patterns are often not clear in teams
For this and SSR PPR ISR, cache tag etc patterns there is def room for some better dev tooling / visualization
I've been working on the React Suspense docs: there's now a full list of what activates a Suspense boundary, with a live example for each trigger!
Thank you @sebsilbermann and Dan Abramov for reviews!
https://t.co/nblR4BcgHr
Demos below ↓
I do love the idea of approving plans with all tool calls and permissions listed. Right now yolo mode or step by step approval babysitting are both not ideal.
Permission management will change for sure a lot with more deeper integrations and automations
Second time after Vercel Ship in Berlin now for WeAreDevelopers
They really were trying to answer the question how many talks can we do at the same time, answer is a bazillion
Able to easily start with grill-me for people new to working with AI in general and having your vids to go along with it. Then obvious step to add more on top and explore the other already installed skills.
Not overcomplicated, system agnostic and skills are short.
Just a very natural transition throughout them for all level of experience