Random Llama
Random Llama
ProductsSolutionsBlogCase StudiesContact
Get a Quote
Weekly Newsletter

Get AI & productivity insights weekly

Privacy-first tools, workflow tips, and early product access. No spam — unsubscribe anytime.

Random Llama Software

Texas-built weird tools and custom web platforms—fast shipping, no creepy tracking, no enterprise bloat.

Links
  • Home
  • Products
  • Case Studies
  • Blog
  • Solutions
  • Credentials
  • Contact
Services
  • Custom CMS
  • Booking Engines
  • Mobile Apps
  • AI Integration
  • Website Maintenance
Connect
  • Privacy Policy
  • Terms of Service
  • Cookie Policy

© 2026 Random Llama Software, LLC. All rights reserved. Privacy Policy

Back to Blog
ai-newsopenaiarmsecurityai-infrastructure

OpenAI Kills Sora, Arm Ships Its First Chip Ever, and LiteLLM Gets Hacked

Robert HattalaMarch 25, 2026

OpenAI Pulled the Plug on Sora

Six months after launch, OpenAI is killing its Sora video app. The standalone app hit #1 on the App Store back in September. Now it is gone.

The real fallout is the Disney deal. Disney pledged $1 billion and exclusive character rights to OpenAI in December. That deal is dead now. Turns out "high compute costs" is corporate for "this thing is bleeding money and generating deepfake lawsuits."

Hollywood was already nervous about AI video tools enabling nonconsensual imagery. OpenAI just proved them right by shipping it anyway, then pulling the plug when it got expensive and legally messy.

Arm Built a Chip for the First Time in 35 Years

Arm announced the AGI CPU. Not "AGI" like artificial general intelligence. It stands for something else but the marketing team knew exactly what they were doing.

The actual hardware is wild. Up to 136 Neoverse V3 cores on TSMC 3nm. 300 watts. 800 GB/s memory bandwidth. Arm claims 2x performance per rack versus the latest x86 platforms.

Meta is the launch partner. They will run these alongside their custom MTIA accelerators. OpenAI, Cerebras, Cloudflare, and SAP also signed up. This is a big deal because Arm has never sold finished silicon before. They have always licensed IP to other chipmakers.

If you are building AI infrastructure, pay attention. The CPU side of AI workloads just got a serious new option.

LiteLLM Got Supply Chain Attacked

A threat group called TeamPCP backdoored LiteLLM versions 1.82.7 and 1.82.8 on PyPI. The package gets 97 million monthly downloads. It sits in 36% of cloud environments.

The attack chain was clever. They compromised a Trivy GitHub Action first, then used that to steal PyPI publishing credentials from LiteLLM's CI/CD pipeline. The malicious versions were live for about three hours before PyPI quarantined them.

The payload was a .pth file that runs automatically on every Python process startup. If you installed those versions, it was harvesting credentials silently.

Check your lockfiles. Pin your dependencies. And for the love of all things holy, pin your CI/CD actions to specific commit hashes, not version tags.

Quick Hits

Google is rolling out Personal Intelligence to all free US users. Gemini can now pull from your Gmail, Photos, and YouTube data. Arcade launched ToolBench, a benchmark ranking how open enterprise apps are to AI agents. GitHub and Figma scored highest. Slack and Workday scored lowest.

Related posts

Anthropic Beat OpenAI on Revenue and the AI Race Just Shifted

Anthropic just passed OpenAI on enterprise revenue, Claude cracked a UK AI security benchmark nobody else could, and China blocked Metas 2B AI buyout.

May 9, 2026

GPT-5.4 Beats Humans While AI Compute Crisis Hits Hard

GPT-5.4 just beat the human baseline on real desktop work, Anthropic grabbed all of SpaceX's Memphis compute, and Wall Street is staring down ten new AI agents.

May 8, 2026

AI Got Practical: GPT-5.5, Finance Agents, and Apple's Pivot

OpenAI swaps in GPT-5.5 as default, Anthropic ships ten finance agents, Apple opens to third-party AI, and ElevenLabs crosses 500M ARR. Big moves.

May 7, 2026

Need custom software or maintenance?

We build privacy-first apps, booking engines, and full-stack platforms — and keep them running.

Browse SolutionsGet in Touch
All posts