PWV Portfolio Feed

<3 Berlin

`undo` is one of the coolest commands in gitbutler. 

pretty much **any operation** can be undone. 
and yes, there is a `redo` and yes, you can run them multiple times

i think this piece does already introspect my negative reaction to llm-generated portfolio pieces: a lot of llm-assisted work is competent but impersonal and generic, and it's hard to determine anything about the person behind it, to have legibility of their effort, intent, skill, personality, etc.

Per-function compute, defined in code.

Region (run close to your users or data), memory up to 4 GB, and CPU up to 2 vCPU, all configurable per function. Validate it in a deploy preview before it hits production.

Background functions, without the magic suffix.

Long-running tasks (up to 15 min) used to require a -background filename. Now you just set background: true in config. No naming rules, easier for agents to set on your behalf.

Fetchable module support is now native.

Code written for Deno, Bun, Hono, or Nitro runs on Netlify with minimal changes. And a single module can handle web requests and platform events together: a fetch handler plus named event handlers on the same export.

One function can subscribe to multiple events. Multiple functions can subscribe to the same one.

You structure code around what belongs together, not around filename conventions the platform forces on you.

Event-triggered functions, reimagined.

Before, subscribing to a platform event meant naming your file a specific way that humans forgot and agents couldn't discover.

Now any function subscribes by exporting a named handler: userSignup, deploySucceeded, formSubmitted. Fully typed.

In 2019 we launched Netlify Functions. In 2023, Functions 2.0 rebuilt the whole experience around Developer Experience.

Today we're taking it to the next frontier: Agent Experience.

Here's what shipped. 🧵

fwiw, I don't think that my specific piece said that using llms for writing is bad/etc, but specifically that people who use LLMs for personal writing have been getting results that sound generic and don't achieve their purpose, and using LLMs to write mass spam-like emails is ineffective and spammy

What’s the difference between your mcp and our official mcp?

combo of following me back at the same time as a critique is a new one, props for being direct.

i don't agree with all of your counterpoint but do agree that  there are folks for whom llms are genuinely enabling. like with all writing it's very hard to disclaim all unintended extrapolations

GLM-5.2 Fast is now live on Command Code!

The high-throughput build of GLM-5.2, same frontier coding, now at 120–250 tokens/sec for snappier long-horizon runs.

1M context. Tool use. Reasoning.

Available in the $1 Go plan with $10 usage credits and above!!

Try it out now with /model.

#ZedEditor

- Diff scrolling, multi-cursor editing, and markdown preview search are all snappier too.

Thanks lgeiger and Rani367!

Lots of performance work this release, with a heavy focus on Git:

- Faster HEAD resolution, uncommit, remote listing, and default-branch lookups, plus fewer git processes spawned as you work.

New commands landing this release:

- `editor: select {inside,around} delimiters` expand a selection through the nearest brackets, braces, parentheses, or quotes.
- `workspace: reset pane sizes`

Thanks subeax and 42piratas!