Sora Is Dead. OpenAI's Video Generator Shuts Down.

Good morning, welcome to Builder's Briefing for March twenty-sixth, twenty twenty-six. I'm Alex, joined as always by Sam, and we have a packed show today — a major AI product is officially dead, Apple is making moves into business software, and there's a founder story that honestly might be the best advice you hear all week.

Yeah, there's a real theme running through today's news. It's all about dependencies and what happens when you build on sand. Let's get into it.

So the big story — Sora is dead. OpenAI's text-to-video model, the one that absolutely broke the internet when it debuted in early twenty twenty-four, is officially shutting down. Over six hundred points on Hacker News, the community is split between 'I told you so' and genuinely mourning what could have been.

This one stings for a lot of builders. If you integrated Sora's API into content pipelines or creative workflows, you need a migration plan like yesterday. Runway, Kling, Veo — they're the obvious alternatives, but none of them are drop-in replacements.

Right, and what's wild is the pattern here. Sora never found its business model. Too expensive to run at scale, too unpredictable for professional work, and open-source video models kept catching up. It's the classic gap between impressive demo and reliable production tool — and that gap is where products go to die.

The real takeaway for anyone building on generative media APIs: treat any single-provider dependency as something that could vanish overnight. Abstract your model layer. The teams that built behind a gateway or an abstraction are in way better shape today than the ones who went deep on Sora-specific features.

And speaking of abstraction layers, the timing on this is almost poetic — LiteLLM is trending hard on GitHub right now. Over fourteen hundred engagement. It's the proxy server from BerriAI that lets you call Bedrock, Azure, OpenAI, Vertex, Anthropic, and dozens more through a single OpenAI-compatible interface.

That's not a coincidence. The market is literally telling you what it needs. If you're not routing through a gateway yet, today is the day to start. Model churn is only accelerating.

On the model side, Google Research published TurboQuant — a new approach to aggressive quantization that maintains quality at much smaller model sizes. If you're deploying on-device or at the edge, this is directly relevant.

And pair that with Ente launching a local LLM app — everything stays on-device, fully private. Ente's known for encrypted photo storage, so privacy is their DNA. For anyone in healthcare, legal, finance — this is a reference implementation worth studying. Two hundred fifty-eight points on Hacker News, so clearly people want this.

A couple of Claude ecosystem tools worth flagging too. Letta AI released something called claude-subconscious — it gives Claude Code a persistent memory layer across sessions, so you stop re-explaining context every time you start a new chat.

Oh, that's huge for daily driver users. And there's also claude-squad, which lets you run and manage multiple AI coding agents in parallel — Claude Code, Codex, OpenCode, Amp, all at once. If you've been scripting that orchestration by hand, someone finally built the proper tool.

Alright, dev tools. This one caught my eye — Video.js version ten beta just dropped. The original creator came back after sixteen years and rewrote it from scratch. It's eighty-eight percent smaller.

Eighty-eight percent! That's not a refactor, that's a demolition and rebuild. If you're embedding video in web apps and you either stuck with the old bloated Video.js or switched away because of it, version ten is absolutely worth evaluating. The web video player space just got competitive again.

Also, Microsoft's TypeScript-Go port keeps moving — that's the native Go port of the TypeScript compiler. When this ships, expect dramatically faster type-checking and compilation. If you maintain large TypeScript codebases, track that staging repo now. Link in the briefing.

And a quick shoutout to email dot MD — it converts Markdown to email-safe responsive HTML. Two hundred sixty-one Hacker News points, which tells you how many developers have suffered through email rendering across clients. Simple tool, real pain solved.

Let's talk security for a second. Strix is an open-source tool using AI agents to find and fix vulnerabilities in your app — basically automated penetration testing. Over a thousand engagement on GitHub.

That's interesting because it's hitting the exact compromise most teams need — you want to move fast, you're probably skipping security reviews, but you also don't want to get breached. If Strix actually delivers on the promise, it could become a standard part of CI pipelines.

Two infrastructure stories worth connecting. IEEE Spectrum is reporting on data centers shifting from AC to DC power distribution, and Arm announced their AGI CPU architecture — that's Arm General Infrastructure, not artificial general intelligence, before anyone gets too excited.

Ha, yeah, that naming is going to confuse people for at least a year. But both stories point the same direction — better performance per watt, more density, lower costs. If you're deploying on ARM-based cloud instances like Graviton or Ampere, the ARM server ecosystem just keeps getting more competitive with x86.

Now here's the one that might shake up a lot of builders — Apple just launched Apple Business, an all-in-one platform targeting SMBs. Payments, invoicing, scheduling, the works. Nearly six hundred points on Hacker News.

If you're building SMB tools, Apple just became your competitor, and Apple's distribution advantage in that space is massive. On the flip side, if you're building inside Apple's ecosystem, there might be integration opportunities. But either way, you need a differentiation story that goes beyond just existing.

Alright, I have to talk about this founder story because it's so good. A founder took an actual job as a pest control technician before building vertical SaaS for the industry. Two hundred sixty points, a hundred fourteen comments — the community loved it.

This is the template, honestly. Domain immersion before code. If you can't explain your customer's daily pain in their language, you're just guessing. It's the cheat code for vertical SaaS and I wish more founders did it.

Quick hits — PLANKA is a self-hostable open-source Kanban board if you want a Trello alternative. There's a curated list of free public APIs for your next side project, link in the briefing. Meta was found liable in a landmark verdict for knowingly harming children. And antimatter has been physically transported for the first time, which is just wild.

Also shoutout to the 'Slowing the Fuck Down' post — a developer's case for sustainable pace. Given how fast everything is moving, maybe worth a read this weekend.

So the through-line today is crystal clear: abstract your dependencies or get burned. Sora's shutdown is the latest proof that building on a single AI provider is a liability. LiteLLM trending the same day is the market screaming the answer at you.

Route through a gateway. Build your abstraction layers. And if you're starting something new, go deep on the domain first — be the pest control technician before you write a line of code.

That's the briefing for March twenty-sixth. All links and details are in the show notes. We'll be back tomorrow with more — until then, build smart and keep those dependencies loose.

See you tomorrow, everyone.