nnx

The pricing is surprising, Data Produced at $0.04/GB seems reasonable (when compared to other clouds event streams) but Data Consumed being at the same $0.04/GB is rather steep...

This means actual usage is $0.08/GB in the simplest case (one consumer) but fan-out consumer strategies get very expensive very fast.

> Some days, I would seriously rather work at Wendy's.

Narrator: He would not.

is flickering fully fixed yet?

Not saying there's no negative impacts, but what are the _huge_ negative impacts that have materialized so far?

Coincidentally this is how pretraining works :)

Looks really interesting but the Go binding sadly uses cgo. Could the binding be done in pure Go? Or at least purego (the cgo alternative using Go assembly for FFI) ?

I ended up just using the web version, which is actually better than the native app (multiple tabs works!).

> My M5 Pro can generate 130 tok/s (4 streams) on Gemma 4 26B.

This seems high. At which quantization? Using LM Studio or something else?

Note: Darkbloom seems to run everything on Q8 MLX.

The only limit is yourself!

Are you `Ionstream` on OpenRouter?

If so, it would be great to provide more models through OpenRouter. This looks interesting but not enough to make me go through the trouble of setting up a separate account, funding it, etc.

in JS, signals and AbortController can replicate some of the functionality but it's far less ergonomic than Go.

https://github.com/ggoodman/context provides nice helpers that brings the DX a bit closer to Go.

Can you describe what is this slightly different approach and why it should work on all models?

This looks very interesting, but I wonder how's the rewrite approach gonna impact the long-term maintenance and porting changes _back_ from Tree Sitter.

As you mention WASM-readiness, did you consider using the official Tree Sitter WASM builds nicely packaged with wazero (pure Go WASM runtime) ?

It may help staying sync with upstream for the long term and, while probably a bit slower, has nice security and GC advantages too.

Why do you think it's "the wrong direction" ?

Hmm no, because in the case of purchasing alcohol the ID check is 1:1, in time and in space, it's ephemeral (unless the clerk has extreme photographic memory).

In the case of an online-based ID check, even with nice looking privacy terms, there is no guarantee that your ID won't be stored forever and/or re-analyzed many times cross-checking with other services, and worse leaked.

Really interesting, I’m currently using https://github.com/fastschema/qjs but would love a bit lower-level control like your reactor and Go library provide.

loop_once may have been called run_microtask if I understand the “loop” boundary correctly?

Is there a way to be more granular in execution? Like running a single “basic block” (until a jump) or until next function call?

It's sad to see Cloudflare slowly adding egress fees to their new services.

It was a core differentiator to never* have to worry about egress with them.

*: unless it's so large that it borders on abuse or require a larger plan

Is there a way to somehow authenticate with ssh key instead?

I do not really get why user-agent blocking measures are despised for browsers but celebrated for agents?

It’s a different UI, sure, but there should be no discrimination towards it as there should be no discrimination towards, say, Links terminal browser, or some exotic Firefox derivative.

Not sure about general-purposeness, but the architecture looks rather perfect for LLM inference?

Wonder why they do not focus their marketing on this.

This suggests adding a `QWEN.md` in the repo for agents instructions. Where are we with `AGENTS.md`? In a team repo it's getting ridiculous to have a duplicate markdown file for every agent out there.

It would be nice if OpenRouter supported batch mode too, sending a batch and letting OpenRouter find the best provider for the batch within given price and response time.

> It's an unikernel design that is compiled as a single EFI binary and does not use virtual address spaces for process isolation.

Are you able to quantify what kind of performance improvement to expect from such a design?

My understanding is that this can completely skip overhead of syscalls and TLB, but I have no idea of how much performance benefits can it result in practice.

Is that a special case for SW devs or they also aren't able to deduct other employees?

...unless DeepSeek releases R2 to crash the party further

Would it make sense to ReactRouter SSR though Hono?

If not, great opportunity for bvrrr.

> - start on device boot

Don't think it's possible on iOS, and I'm doubtful TikTok does this even if it were possible on Android.

> - GPS location (browser can access location too if given permission ofc)

Actually this AND all of the other fingerprinting methods you listed also require user permission managed by the mobile OS (iOS/Android).

In effect there is nothing a 'native' app can get without user knowledge/consent that a web app could not. If there is, it should be considered a bug for mobile OS providers to fix.

I'm unable to figure out how to solve current Daily Puzzle (Puzzle ID: 79369cc6) at https://arcprize.org/play

Either I'm really dumb or the test is getting into captcha-like territory where humans aren't really good at solving/deciphering the test anymore.

This looks interesting but I can't think of an use case. Can you share some examples where it could be useful?

- Space pirates?? - You know, pirates! but in space!!