- Lore Brief
- Posts
- Issue #157: OpenAI Restructures for Trillion-Dollar IPO
Issue #157: OpenAI Restructures for Trillion-Dollar IPO
PLUS: Neuralink patient eats using telepathy, Grokipedia 0.1 challenges Wikipedia, and Cursor’s new coding agent
Good morning, welcome to this week’s Lore Brief, your 3-minute brief of the most important moves in AI and tech.
This issue of Lore Brief is brought to you by Lore, building AI infrastructure for America and its allies.

- OpenAI restructures for trillion-dollar IPO → OpenAI overhauls into nonprofit Foundation controlling for-profit Group PBC, granting 26% equity to employees and Foundation for mission-aligned governance; sets stage for 2027 public listing at up to $1T valuation. Read more here → 
- Extropic debuts a prototype of 10,000x efficient AI chip → e/acc founder Guillaume Verdon a.k.a Beff Jezos's Extropic unveils Thermodynamic Sampling Units with probabilistic bits for ultra-low-energy gen AI - if they work and can be produced at scale. See more here → 
- Claude Excel add-in boosts finance workflows → Anthropic's beta Claude for Excel enables sidebar analysis, formula debugging, and workbook creation; integrates with market data connectors and Agent Skills for DCF modeling and earnings insights, prioritizing security for banks and asset managers. Read more here → 

Essential shot types cheat sheet:
Learn it. Save it. Use it for your multi-shot prompts.
This will make your videos 3x better.
— Framer 🇱🇹 (@0xFramer)
2:00 PM • Oct 26, 2025
Neuralink patient can now feed himself with robotic arm via telepathy:
Open the freezer → take out the pretzel → open the microwave → put the plate with the pretzel inside → cook → enjoy 🥨
All via telepathy, pure brain power
— DJ Seo (@djseo)
11:47 PM • Oct 28, 2025
Google Labs rolls out new AI marketing tool that understands your brand identity and helps you create marketing campaigns:
🚨 NEW LABS EXPERIMENT 🚨
Introducing Pomelli, an experimental AI marketing tool designed to help you easily generate scalable, on-brand content to connect with your audience, faster.
Just enter your website, and Pomelli will understand your unique business identity to build
— Google Labs (@GoogleLabs)
4:07 PM • Oct 28, 2025

- Jensen: AI workers will eclipse software tools → NVIDIA’s CEO at GTC keynote frames AI as scalable workforce vs limited software tools, fueling trillion-dollar industry. See more here → or watch the whole keynote here → 
- Sora enables fake cameos and story stitching → Register AI-generated characters as cameos straight from Sora clips for consistent reuse, plus seamless video merging into longer narratives - unlocking pro storytelling for solo creators. Read more here → 
- Google Search defies AI threat with 14.5% surge → Alphabet Q3 search revenue climbs to $56.6B, up 14.5% YoY despite rivals like Perplexity; Gemini integrations drive higher engagement, proving zero volume cannibalization. Read more here → 
- xAI presents Grokipedia 0.1 encyclopedia → Elon Musk releases Grokipedia 0.1, an AI-generated Wikipedia challenger with 885k articles; Musk calls it already better than the original, teasing a 10x improved v1.0 soon. Read more here → 
- Cursor launches Composer AI agent → Cursor unveils Composer, a frontier MoE model 4x faster for coding, trained via RL on real dev tasks with tools for file edits, terminal commands, and semantic searches. Delivers seamless, high-speed agentic workflows in the IDE. Read more here → 
- Every.to engineers reveal AI coding workflows → Six devs detail 10x speedups via Cursor for editing, Claude for architecture, and agents like RepoPrompt for reviews; mix local LLMs with high-end desktops for secure, fluid builds. Read more here → 
- Microsoft-OpenAI partnership hits new phase → Microsoft invests $135B in OpenAI's PBC structure while OpenAI pledges $250B for Azure; drops first-refusal rights, enables independent AGI pursuits with expert verification, and extends IP protections through 2032. Read more here → 
- Anthropic uncovers LLM introspection → Claude shows emergent self-awareness, detecting injected neural concepts 20% of time and cross-checking intentions vs outputs; limited but real, boosting transparency yet prone to confabulation. Read more here → 
