Twitxy Saved

Kenton Varda

Verified account Cloudflare

@KentonVarda

Tech lead for @Cloudflare Workers, sandstorm.io, capnproto.org, lanparty.house. Married to @qiqing.
Austin, TX Joined November 2008
299 Following 22K Followers 7K Posts

Latest posts 5

Pinned post

Kenton Varda Verified account @KentonVarda
Today we are releasing Cloudflare OS, a chatbot with connectors, just like every other tech company is doing.

Except actually, it's different. This is a remake of Sandstorm[.]io, my startup from 10 years ago, except this time built on Cloudflare Workers (the platform I've spent the last 9 years building) and deeply leveraging AI. This is more or less the culmination of my secret 10-year master plan.

This is a full-on personal app vibe coding platform, in which the sandbox is so secure that you can pretty much go wild -- the AI cannot introduce a significant security bug. We believe a company's security team can feel comfortable giving non-technical users permission to vibe code and then sleep soundly at night.

How is that possible? It's the Sandstorm security model, revisited. A "Gadget" is the same thing as a Sandstorm "Grain": a fine-grained app instance. For example, if you have a document editor app, each document runs as a separate instance of the app, in a separate sandbox (one "Gadget").

This means two things, both of which I think are Big Deals:
1. The platform can manage all access control, by controlling who can access the Gadget at all. There is no way the Gadget can accidentally leak itself to an attacker -- even an attacker who has access to other Gadgets based on the same app.
2. Since everyone is running their own copy of the code, everyone can freely *modify* their copy of the code.

Think about #2 a bit more.

What if, when you wanted a new feature in the software you are using, you could just prompt your agent to add it?

This doesn't work in the cloud Software-as-a-Service model, because you are not running your own copy of the app.

Sandstorm tried to change that 10 years ago, but the world wasn't ready, because not enough people had the skills or patience to actually modify their software. AI has changed that. Now you just ask the agent -- the same agent that you are using to help you interact with the Gadget can also modify the code of the Gadget.

And it is so fun.

github.com/cloudflare/clo…
github.com GitHub - cloudflare/cloudflare-os: Agent workspace built on Cloudflare Workers for creating... Agent workspace built on Cloudflare Workers for creating documents, building apps, and running agents with your company’s context and systems. - cloudflare/cloudflare-os
236 replies 696 reposts 5.9K likes 874K views View on X
Kenton Varda Verified account @KentonVarda
And now, a year later, every major harness has implemented it... and they all kept the name.
Dane Knecht 🦭 Verified account @dok2001 Dec 18, 2025
It's great to see everyone adopting Code Mode! Like most good things @Cloudflare, it started with a blog post by @KentonVarda. He wrote it before the code, and we never got around to figuring out a better name.
View quoted post on X
17 replies 24 reposts 528 likes 21K views View on X
Kenton Varda Verified account @KentonVarda
In the late '90s I was a huge 3dfx fanboy. For a few years, they were so far ahead of everyone, making all other video cards look like jokes. I had a Voodoo 1, 2, and 3, and was excited for the Voodoo 5 with its promised full-screen anti-aliasing (FSAA). (I was in high school / starting college, a big gamer, and doing some OpenGL graphics programming as a hobby project.)

The day the first GeForce was announced, though, I immediately recognized that 3dfx was toast. Like, immediately. Why?

3dfx had been entirely focused on maximizing fill rate. That is, given a set of textured triangles that are *already* specified in 2D screen space, the hardware would rasterize them as pixels. It was up to software to get the triangles into screen space in the first place, which mostly involves applying some matrix math to every vertex. But 3dfx figured there are way more pixel than vertices, and focusing on more pixels was the way to go to improve quality. With enough pixels they figured they could literally render every frame at 4x resolution and then down-sample to make the edges smoother (FSAA). The Voodoo 5 was a monstrous brute force design whose top-end version had four copies of the chip in order to render those four pixels at the same time.

The GeForce, on the other hand, went ahead and moved vertices into hardware. The result? It could render an order of magnitude more triangles, which meant surfaces could be smoother. It was immediately obvious that this had a far bigger impact than FSAA, both from the screenshots and from what I'd been learning about graphics programming.

A friend of mine (who still attends my LAN parties today) was staying with us at the time. I remember going over to his computer while he was out, loading up the screenshots from the announcement, and leaving a little note in a text editor about how I'd switched allegiance.

Of course, games did not get the benefit automatically. Engines had to be rewritten to move their vertex handling into hardware, whereas 3dfx's brute force and FSAA would immediately apply to existing games. This is why 3dfx's bet was the easy one to make, and Nvidia's so much harder.

But Nvidia has continued making these bets, over and over, ever since. They have repeatedly built for markets that didn't exist yet, investing in stuff that would take many years to pay off. CUDA when nobody was asking for it. Ray tracing when nobody was asking for it. Almost every generation of Nvidia hardware has something qualitatively new.

Jensen calls these "zero-billion-dollar markets".

Doing this requires leadership. The natural state of a big company is to focus on whatever incremental improvements their biggest customers are asking for, because those are the easiest decisions to justify. It works great for a while, then fails. That's the Innovator's Dilemma. And even executives who have read the book find it hard to internalize. Even executives hand-picked by the CEO who has internalized it often cannot themselves internalize it. It's weirdly hard to find leadership that gets it, and no amount of bottom-up innovation can make up for it.

And that's why I sold my Nvidia stock this morning.