Yuval Avidani
Author
Turns out that while everyone's chasing the tool that generates you a presentation from one prompt, a single developer named 1weiho quietly dropped something going in the exact opposite direction: open-slide, a "slide framework built for agents," where you describe the deck in words, your coding agent writes React, and the framework handles everything else.
So what actually is open-slide
Let's break it down. open-slide describes itself as a "slide framework for the agent era." The idea is simple: you don't open PowerPoint and you don't drag text boxes around. Instead, you describe the presentation in natural language, your coding agent (Claude Code, Codex or Cursor) writes the code in React, and open-slide handles all the boring-but-critical parts: the canvas, scaling, navigation, hot reload (auto-refreshing the screen the moment you change code), presenter mode, and export.
An important clarification so I don't sell you a false dream: this isn't "write a prompt, get a deck" Gamma-style. This is infrastructure, a framework running under the hood, that a coding agent operates on your behalf by writing code. You need a Node dev environment to run it. In other words: this is an engine for developers and people working with a coding agent, not an app for everyone.
Why you'd even need this
Picture this: anyone who's ever tried to build a "pretty" slide in HTML knows the pain. The slide looks perfect on your screen, then you open it on your phone and everything shifts, or you project it and the text runs off the screen. That's the classic obstacle: how do you make the same slide look identical on every screen.
open-slide's solution is elegant in its simplicity. Every slide is built inside a fixed 1920x1080 canvas, and the framework just shrinks or scales that box uniformly to fit whatever screen it's on. Think of it like a picture in a frame: you draw once at a fixed size, and the whole frame shrinks or grows together. The result, according to the official docs, is that one slide "looks identical on phone, on a 4K screen, and on a projector." Sounds technical, but in my eyes this is exactly the part that steals hours from all of us: not having to redraw the slide for every screen size.
How it actually works
One slide is one React component. The framework renders (draws) one component at a time, and gives you navigation, hot reload and presentation mode for free. The stack behind the scenes is React plus Vite (a tool that runs and builds the code fast).
The thing that speaks to me personally: open-slide ships with "agent skills," built-in capabilities like create-slide/ and slide-authoring/, that any Claude Code, Codex or Cursor session can call. Meaning the agent doesn't have to guess how to build a slide, the project already teaches it. And that's just the start of the feature list: there's an in-browser inspector for live editing with persistent comments, an asset manager with logo search via svgl, a presenter mode with speaker notes, timers and dual display, static export to HTML and PDF, and a drag-and-drop slide manager with folder organization.
Getting started for real
Let's talk practically. How do you start. It all begins with a single script that scaffolds the project structure. Notice a small but important detail: there are two separate packages here. @open-slide/core is the "engine" that runs everything, and @open-slide/cli is the separate tool that scaffolds a new project for you.
As of the current version, @open-slide/core stands at 1.13.2, published on June 27, 2026. And here's a number that tells a story: the project has had 40 releases total. This isn't software sitting still, this is software that's been moving forward fast.
The twist: a one-person project
And here's the part that made me stop. open-slide isn't a product from a company with a marketing department. It's an open-source project by a single developer, the user 1weiho (David Ho), who's already racked up roughly 5,604 stars and 389 forks on GitHub. And here's the crazy part: the first package was published on npm on April 19, 2026, and the repo was opened on April 26. That means we're talking about a project that's only two months old.
The repo is written in TypeScript, under the MIT license (an open, permissive license that lets you use, modify and distribute it), and maintained by one person. In my eyes these numbers, two months and already thousands of stars, aren't just a vanity metric. They're evidence of real hunger in the market: a lot of people are looking for a way to turn slides into code that an agent can write well, instead of more generic AI images.
Bottom line
In my view, this is exactly the engine behind the idea of "slides as code." You describe, the agent writes, and the infrastructure makes sure everything looks professional on every screen. This is a thesis I've held for a while now: a slide is code, not an image.
But let's be fair about the limits. This isn't a tool for everyone. You need to be comfortable with a dev environment, you need to work with a coding agent, and this is a two-month-old, one-person project, so don't build your funding-round deck on it without a backup. The stars and versions will keep moving, and this is still a very early chapter in the story.
So the real question that stays with me: is the future of presentations really code that an agent writes for us, or is this just another stop on the way to something we haven't seen yet?