Deno

HTTP servers, PORFOR can now host an HTTP server efficiently.

Comparing to Node, Bun, and Deno, an unrealistic Hello World server uses 10 times less memory and beats all 3M performance for throughput and latency.

But I think the operating word there is unrealistic.

But that's still really cool.

It defaults to node true, dino true, bun true, and your own value is merged into that default.

So stable Node.js releases switch without asking because PMPM verifies them against the Node.js release team signatures.

Deno, Bun, Node.js pre-releases, and package bins you enabled yourself all ask do you trust this project once per project and per version.

So that's pretty cool.

But PMPM still reads from the MPM registry and still have the same problem.

And the only alternatives, there's another package registry by the team at Deno called JSR.

And they tried to solve some of these problems from the beginning, but it just didn't, it didn't gain in popularity.

And MPM as a registry has so much momentum and gravity that it still remains the default.

Moving on to some web platform news.

So Deno has come out with not really a library, but more of a tool.

And if you're familiar with Cloudflare Workers and Durable Objects, this is called Cell D, and it allows you to do self-hosted distributed durable objects.

So essentially, take your workers and durable objects and self-host them in your own containers.

So TC39 meets, I don't know, maybe I think six times a year, and representatives from different companies and organizations come together, and they propose additions or amendments to the language to try to help either improve it, disambiguate it for implementers, right? So these, these are typically, like, your engines in the browser.

They'll run in Node.js or Deno or Bun.

And basically, the goal is to try to create improvements in the language that are all consistent across these different engines.

That way your code can run anywhere in a consistent way.

You get to be both the victim and the problem with this one.

Eight, Iran-linked operators are running a backdoor called Dindor through the trusted Deno runtime, so it blends into legitimate developer tooling, seen as a later stage payload against US software targets.

Watch for unexpected Deno execution and encoded script launches on your developer and server hosts.

Nine, a backdoor called Sleepwalker poses as DPAPI.dll, sideloads into the ESET management agent and sits completely inert with no outbound traffic until a crafted packet wakes it up.

Your connection watching tooling probably won't catch it, so hunt for a rogue DPAPI.dll sitting next to the agent executable.

Its contribution is architectural.

Instead of the prevailing LLM-centric pipeline that projects vision into a language model space before decoding actions VLA, Turbola encodes vision and language independently, exchanges information through a lightweight bidirectional interaction module adapted from grounding Deno, and predicts continuous action chunks with a compact decoder VLA.

It pairs DenoV-3 vision backbones with BERT for language, ships Apache 2.0 with checkpoints already on Hugging Face, and continues a thread of tiny, real-time policies that keep proving you do not need a large language model in the control loop to get competitive manipulation.

At 31.2 MS of inference latency, it is the rare efficiency claim that translates directly into a robot that can actually react.

The wider ecosystem.

But yeah.

Uh, Deno says, "Another mock masterclass last night.

Myself and everyone on screen will be locked in for the fifth world title win." Yeah, I hear you.

It was a, yo, the match was really cool.

If he gets another title win, I will not be doing Destination Dynamite, and that is just the truth.

I'm putting in my resignation for the show.

Uh, thank you though, Deno.

Uh, there's any more up?