Every major computing primitive followed the same path:

Text became a primitive when we had standards (UTF-8), instant rendering, near-zero cost Images became a primitive when we had standards (JPEG/PNG), CDNs, cheap storage APIs became a primitive when we had standards (REST), infrastructure (AWS), tooling

Video is next. The technology is getting there (Sora, Veo, Runway), but the infrastructure layer doesn't exist yet.

I'm building that layer.

What makes something a primitive: - Fast enough (instant or near-instant) - Cheap enough (marginal cost near zero) - Standard interfaces (abstraction layer) - Composable (programmatic manipulation) - Great DX (easy to integrate)

Current status: - Multi-provider abstraction (not locked to single API) - Intelligent routing (optimize for cost/latency/quality) - Developer-first (npm install, 5 lines of code) ```python & Node JS

pip install vloex npm install vloex

import vloex video = vloex.generate("Your content") # Auto-routes to best provider, handles failover

Test case: Auto-generate video release notes from GitHub commits Live example: https://vloex.com check live example section This is Part 1 (generation).

Vision includes composition, templating, manipulation, serving - full primitive stack.

The bet: Video generation hits "fast enough + cheap enough" in 2-3 years. When it does, every app will use video. They'll need infrastructure.

This is that infrastructure. YC's current RFS: "Treat video as a computing primitive, not an output"

The question: Am I building at the right time, or 2 years too early?

Docs: https://api.vloex.com/docs

Code: https://github.com/vloex/vloex-python

Honest feedback welcome.

PS: I have few slots for beta users and design partners that are interested in video space so do let me know or connect with me at sats@vloex.com

DeepMind: CodeMender: an AI agent for code security

Self hosted LLM cost monitoring

Claude Code uses WebFetch vs. WebSearch (observations, schemas, prompts)

Rust-Written Redox OS Enables Multi-Threading by Default

WebGPU and the Price of Compiling WGSL

Pretraining Large Language Models with NVFP4

My Misadventures in Gentle Parenting

Audacity 4: a glimpse of a new, more modern UI for the free audio editor

Show HN: Making video a software primitive (starting with generation)

The Next Computing Revolution: Bringing Processing Inside Memory

The century-long search for the Loch Ness Monster

Apple Puts Hardware Chief John Ternus in the CEO Succession Spotlight

AI Video Overview

A rubber duck, even on acid, won't take your job

OpenAI tells developers ChatGPT is ready to be their gatekeeper

GPT-2 implementation in Modular MAX

Quake Bot Archive

The Machine Won. Can Humans band together before it's too late? [video]

The Little Book of Deep Learning (2024)

Build Trust Through Empathy

Fresh approaches needed to tackle political 'age of rage', US study suggests

Mutation Testing

Valorant's 128-Tick Servers

Immutable Value

U.S. Banks Missed Warning Signs on Tricolor. Now, Their Losses Are Adding Up

Discovery of Unstable Singularities in Navier Stokes (DeepMind)

The AI clankers yearn for cultural cachet

Paramount Buys the Free Press, Ushering in a New Era at CBS News

Beyond horizons: world-record long-distance photograph 443 km, Finestrelles, Pyr

Briefly Noted: Summer Reading

Show HN: Ranked Vim Motions — move, select, and delete as fast as you can