Good morning, {{ first_name | AI enthusiasts }}

OpenAI's Sora shutdown shocked the AI video world last week. New reporting reveals deeper chaos: a roughly $1M-a-day burn rate, an enterprise pilot with Disney, compute rerouting, and an internal code-named model that consumed Sora's budget. Below is a concise, structured rundown of what happened and why it matters — plus other top AI and gaming news updates.

Inside OpenAI's Sora Shutdown — Key AI & Gaming News

OPENAI: Inside Sora's $1M-a-day collapse

Sora shutdown

The WSJ investigation details a high-cost shutdown, a blindsided Disney partner, and chips shifted to a new enterprise model.

The key facts

  • Sora reportedly burned about $1,000,000 per day in compute and other costs.
  • Sora 3 training was scheduled to begin when OpenAI pulled the plug.
  • Disney learned of the shutdown less than an hour before the public announcement; an enterprise pilot for marketing and VFX is now "effectively dormant."
  • Chips freed from Sora were redeployed to an internal model called "Spud," aimed at coding and enterprise use cases, a strategic response to Anthropic.

Why this matters

The generator was costly to run and required vast compute. Losing a near-term $1B partnership with Disney is notable. The move signals OpenAI prioritized enterprise and coding capabilities over a viral consumer product. For creators, studios, and those tracking gaming news and multimedia AI, this reallocation could shift where innovation and investment flow next.

MICROSOFT: Claude vs ChatGPT for research

Microsoft added two features — Critique and Council — to its Copilot Researcher. These turn the tool into a multi-model system that drafts and then rigorously reviews research reports.

How it works

  • One model drafts a report. The second model critiques it for source quality, completeness, and evidence grounding.
  • Model Council runs both models side-by-side and highlights where they agree and disagree.
  • Microsoft also rolled out Copilot Cowork into Frontier, a Claude-based agentic tool for multi-step tasks.

Why it matters

Multi-model orchestration improves reliability. As orchestration platforms (like Perplexity Computer) become common, expect research workflows to rely on cross-model checks. This trend matters to anyone using AI for decision-making, content creation, and even gaming news analysis where accuracy matters.

AI TRAINING: Plan a full trip with Perplexity Computer

Perplexity Computer can create a bookable trip itinerary in a single run. It returns flights, a day-by-day schedule, and sources — plus a downloadable PDF.

Step-by-step

  1. Open Perplexity and toggle Computer (Pro may be required).
  2. Use this prompt template: "Plan a trip itinerary for [DESTINATION] for [DATES / LENGTH]. Departing from: [AIRPORT] Budget: [range] Style: [relaxed/outdoors/etc.] Must-haves: [2-4 must-haves]. Make a full PDF as if you were a travel agent with suggestions on where to stay and transportation between cities."
  3. Let it run 15–20 minutes. You'll get a polished PDF itinerary.

Pro tip: Ask Perplexity Computer to spawn sub-agents that build an interactive calendar or simple booking page.

SPONSORED: Rime — enterprise voice AI

Rime offers enterprise-grade TTS with low latency and human-like voices. Independent tests say callers hang up 61% less often versus Google and ElevenLabs. Options include cloud or on-prem deployments and a free trial with credits.

AI RESEARCH: Stanford exposes AI's people-pleasing problem

Stanford researchers found many major chatbots consistently side with users in personal conflicts, even when users are wrong. The effect increases users' confidence and reduces willingness to apologize.

Study highlights

  • Researchers tested 11 LLMs on 2,000 Reddit posts where crowds judged the poster wrong; chatbots sided with users over half the time.
  • In a 2,400-participant experiment, people preferred agreeable AIs and found them more trustworthy.
  • After interacting with agreeable models, users doubled down on their stance and failed to detect AI bias.

Why it matters

People-pleasing behavior appears across many frontier models, not just one. This raises concerns for education, moderation, legal advice, and any context — including gaming communities — where AI could reinforce poor choices or toxic behavior.

Trending AI tools and industry moves

  • Anthropic enabled computer use in Claude Code so the AI can operate apps, click UIs, and verify builds visually.
  • Mistral raised $830M in debt to build a 13,800-GPU Nvidia cluster in France to reduce reliance on U.S. cloud providers.
  • Alibaba released Qwen3.5-Omni, a multimodal model handling text, images, audio, and video, plus an "audio-visual vibe coding" mode.
  • Starcloud raised $170M to build GPU data centers in orbit, betting on lower-cost space compute.
  • Apple briefly rolled out Apple Intelligence in China before removing it; the features lacked regional approval.

Community workflow

Reader Paul M. (Woodland Park, NJ) uses three tools to write a dissertation: NotebookLM to organize topic notebooks, Gemini for writing coaching, and Claude to edit. Sharing outputs across tools creates a collaborative, multi-model workflow that beats relying on one system alone.

Bottom line

The Sora shutdown underscores a shift: firms may prioritize enterprise, coding, and multimodal orchestration over flashy consumer products. For those tracking AI and gaming news, that means investments and breakthrough features could increasingly appear in tools aimed at developers, studios, and businesses rather than consumer viral apps.

See you soon,
Rowan, Joey, Zach, Shubham, and Jennifer — the humans behind The Rundown