If you already have a GraphQL API, you’re closer to AI agents than you think.
Apollo MCP Server exposes your existing operations as MCP tools — with schema validation, protocol translation, and secure execution built in. No backend changes required.
https://apollographql.pulse.ly/2r6ufpirne
Gotta take another look a those :) Thanks!
Oh, to be clear:
I'm not mad at the training. They can of course do that.
I'm mad because it is opt-out instead of opt-in, and it came in a "We’re updating our Terms of Service" email instead of an email titled "Alarm! Opt out now quickly or your stuff is fed to AI".
What are good alternatives to Vercel for free static hosting with good CI integration nowadays?
I really just want to deploy some JS, and get preview builds for branches/PRs.
A screenshot of an email from Vercel introducing AI model training, with a default that requires a manual opt out for free plans. Full text: Optional AI model training You may choose whether Vercel can: Use your code and Vercel agent chats to improve Vercel models Share your code and Vercel agent chats with AI model providers Vercel will not share personal data, account details, or sensitive information like environment variables, and any information of this nature will be removed from data you make available for AI training. Defaults by plan Hobby and Trial Pro: Opted in to AI model training by default, can opt out at any time Pro: Opted out of AI model training by default, can opt in at any time You can manage your preferences in Team Settings → Data Preferences.
I can't remember when I last noped out of something that fast.
Those skills were created via TanStack Intent, which was an overall great experience.
In the next release, full TanStack Intent support will be shipped, but for now they're already available via `npx skills`.
Something I brewed up over the weekend:
```
npx skills@latest add reduxjs/redux-toolkit
```
Have fun with those skills :)
Since git supports gitignore in subfolders, you can also drop a .gitignore with the content "*" inside your dot folder.
Wouldn't that do the same?
Apollo GraphQL is joining the @aaif.io as a Silver member.
The infrastructure for agentic AI should be open. We’re in. https://apollographql.pulse.ly/y5tbwajp9h
#MCP #AgenticAI #AAIF #GraphQL #OpenStandards
JavaScript—making regular expressions easier to read:
const RE_API_SIGNATURE = new RegExp(
String.raw`
^
(?<prefix>
new \x20
|
get \x20
)?
(?<name>
[
A-Z a-z 0-9 _
.
\[ \]
]+
)
`.replaceAll(/\s+/g, ''),
'v'
);
exploringjs.com/js/book/ch_r...
Incredibly hyped to announce Schema Benchmarks, a project I've been steadily working on in collaboration with @fabianhiller.com!
It aims to provide a central and transparent resource for many of the metrics that matter for schemas, including bundle size and runtime performance 😄
Hi, I think there's a lot wrong with me, but in this case I'm just caring for users with potentially weird edge cases too much 😅
bsky.app/profile/tom....
Phew. Does that make any sense? ^^
interfaces) too soon, it would be inlined based on the augmentation in place at compile time of the library, instead of at the compile time of the "final consuming app".
So adding this self-referencing loop here prevents that inlining until the last moment.
It makes the type lazy.
It prevents a type from being evaluated when it is emitted into .d.ts files by the compiler.
We have a few interfaces that can be augmented in userland, and if the library build process - or the build process of a userland monorepo utility library - compiles this type (that depends on augmented
TypeScript type of the day
```
export type LazyType<T> = T & { [K in "" as never]: LazyType<never> };
```
TypeScript doesn't have "Render Types"
(unlike Flow, more opinionated toward JSX)
This plugin proposes to implement render types constraints as an ESLint plugin
Cool idea 👌
This could improve type-safety and prevent many React composition mistakes
That sounds great! Thank you!
The functionality itself is great. It's just a bit much ^^
I've put a lot of work into that apollo-client skill.
It contains our latest recommendations on best practices - so if you want to tune out 10 years of outdated blog articles that went into AI training data, add this skill :)
All that said, today is a day where I just have a headache.
But that's probably the best state of mind to detect these issues and report them ;)
So I constantly scan the bottom left of my browser to find out which buttons are functionality and which buttons are links that lead me off-page.
A screenshot of a styled tooltip with very concrete information.
A screenshot of an alt tag with the text "Inspect dependency tree". The button is an off-page link.
One example of what's causing me problems:
Some icons have a tooltip that immediately appears.
For other tooltips, there's a really ambiguous title (would you have guessed this takes me to a 3rd party side?) that takes a few seconds to pop up.
Not sure where to leave that feedback, but the devs are active here, so it goes here :)
Could you consider to also add a "less distracting mode" next to "light/dark" modes?
I like the tabs in npm, they help me concentrate on one thing. I don't need all information at once.
I really applaud the work on @npmx.dev, but I just tried the page, and I find it to be extremely overstimulating.
It's a *lot* of information on the page at once, and there's no clear distinction between what is "npmx" and "package readme".
Hidden buttons and functionality are everywhere.
Rolldown - Lazy Barrel Optimization
🚀 Lazy Barrel Optimization in Rolldown
Barrel files, files that re-export from other modules, are common in modern JavaScript projects but a nightmare for your build time.
Rolldown now has Lazy Barrel Optimization to fix this.
Here's how it works 🧵
I've created a testing library based on the `Profiler` component that turns rendering into a stream of "render/commit snapshots" (including DOM snapshots) that can be iterated on in tests.
But that's more targeting library authors and their unit tests - I don't have a good solution for E2E tests.
Just one - testing consistency, not just eventual consistency.
Most UI tests only test that UI will reach a target state, but anything could happen in-between. Literally, the whole page could go black for a second in-between.
