Oh yeah the last point is very very important

The solution to the first scenario is that we should also install a way to disable the system, but then the police is notified and it will have to be justified in court as to why you disabled the system. Otherwise automatic fine.

The web is destined to be The app platform for the future. Even Steve jobs saw it. Apple can only delay that for so long.

And it comes from a place of love, not hate.

The React team is made of very smart people. But smart programmers, especially when grouped together, can be dangerous IME.

This is what I mean when I say that React's complexity is rising at an alarming rate.

So now you have:
- "use client", non hook, non async functions
- "use client", non hook, async functions
- "use client", hook, non async functions
- "use server", non hook, non async functions
- "use server", non hook, async functions
- "use server", hook, non async functions

React 19 introduced React Server Components. In particular the split between "use client" and "use server" basically created a whole different context and set of rules for each of the previous 3 colors of functions.

React 16.8 introduced green functions: hooks. Hooks have their own set of rules for how they work and when they can be used.

JS has red/blue functions (journal.stuffwithstuff.com/2015/02/01/w...).

Everybody understands that.

It's the mix of "unnecessary magic" AND security vulnerabilities that pisses us off. Coupled with a vague sense that RSC's complexity was created for Vercel and Next.JS. That React is becoming increasingly complex for little to no gain while ignoring actual issues.

We disclosed two new RSC vulnerabilities:
- Denial of Service (High): CVE-2025-55184
- Source Code Exposure (Medium): CVE-2025-55183

Patches are available now, please update immediately.

react.dev/blog/2025/12...

It seems as though I need to go through one test at a time. It just cannot reliably come up with good tests based on the implementation. I've just realized there's a whole batch of other cases we need to test.

Third round: it used parameterized tests (which is good) until now. But then for one of the tests it included multiple parameters in the same test (adding assertions, then changing parameters, more assertions, etc.). Why? 🤔

After listing all the combination of cases it should test, it managed to make it to 50%.

I'll try to repeat the missing 50%. 🤞

I've added a non trivial GraphQL query that I want to test. There are a lot of variables and edge cases involved (essential complexity).

It "thought" for like 5/10 minutes just to give me the most basic tests ever. Uber surface level. It's not even testing 5% of the query.

query planner was trained by Mike Tyson himself

damn! the alt text reads like an erotic novel

And once again UI/UX "experts" are looking at all the usability and accessibility issues of liquid glass and nodding in unison.