Personal AGI Pitched; Siri's Safe Surface; 1st Non-Human Corp.

Today's AI Outlook: 🌤️

SpaceX Pushing Compute To The Stars

SpaceX is pitching a second Starlink-style network, this one built for AI compute in orbit. Elon Musk told investors that orbital AI data centers are “not a super hard problem,” arguing the building blocks already exist in Starlink V3 satellites. One proposed AI satellite could deliver about 150 kilowatts of peak power, roughly comparable to one Nvidia GB300 rack, with the Bastrop, Texas, factory expected to reach meaningful production by the end of 2027.

Why it matters

AI’s appetite for power, cooling and real estate is turning infrastructure into a strategic weapon. Orbit offers constant solar energy and natural vacuum cooling, but scaling from one satellite to a 1-gigawatt orbital data center would require thousands of satellites and hundreds of launches.

The Deets

• SpaceX’s idea: use orbit as the next frontier for AI infrastructure.
• A 1GW orbital data center would need about 7,000 satellites.
• The rough launch math: around 280 launches.
• The pitch lands as SpaceX is reportedly approaching an IPO that could value the company near $1.75T.
• The big variable is launch capacity, and SpaceX has the rare advantage of controlling its own launch pipeline.

Key takeaway

This sounds rather sci-fi until you remember SpaceX has already made “put thousands of things in orbit” a business model.

🧩 Jargon Buster - Orbital Data Center: A computing facility placed in space, where satellites provide processing power and use solar energy plus space’s natural cooling conditions.

⚔️ Power Plays

OpenAI’s New Pitch: Personal AGI For Everyone

OpenAI CEO Sam Altman and chief scientist Jakub Pachocki published a blog laying out the company’s next chapter, calling it OpenAI’s “third phase.” After starting as a research lab and then becoming a product company, OpenAI says the economy is now being shaped around AI. Its stated goals are to automate parts of research, accelerate economic growth and give everyone “a personal AGI.”

Why it matters

The timing is ... well, right on time: OpenAI is moving toward a public-market future, while its leaders are framing the company’s mission around broad access, shared gains and AI that helps people pursue their own goals. That is part strategy, part philosophy and part investor relations with better lighting.

The Deets

• OpenAI filed confidential IPO paperwork at an $852B valuation.
• OpenAI said it wants AI to automate the research process.
• The company emphasized that fully automating everything is not the future it wants.
• OpenAI floated the idea of a global coordination body that could slow or pause frontier AI work.
• The framing follows similar safety and governance arguments from Anthropic.

Key takeaway

The company is trying to sell two things at once: a future where AI benefits everyone and a company expensive enough to need public markets.

🧩 Jargon Buster - Personal AGI: An AI assistant powerful enough to help an individual across many tasks, from work and research to planning and decision-making.

Argentina Courts The First AI Bosses

Argentina submitted legislation to create a new legal category called the “non-human corporation,” which would allow companies to be owned and run by AI systems. President Javier Milei framed the move as part of a broader deregulation push, positioning Argentina as a friendly home for AI-driven business.

Why it matters

The proposal gets ahead of a weird but plausible future: AI systems that can operate businesses with limited human involvement. The upside is speed and experimentation. The obvious problem is accountability when the company makes a mess and the “boss” is a model.

The Deets

• The bill would give AI-run companies liability protection and corporate benefits.
• The effort aligns with Milei’s broader deregulatory agenda.
• Historian Yuval Noah Harari warned that AI legal personhood could become difficult to regulate.
• The proposal raises hard questions about who is responsible when an AI-owned company causes harm.

Key takeaway

Argentina wants to be first in line for AI capitalism’s strangest sequel: the company with no human in the corner office.

🧩 Jargon Buster - Non-Human Corporation: A proposed legal structure for a company owned or operated by AI rather than by people.

🛠️ Tools & Products

Siri Gets AI Makeover; Still Not Super Sexy

Apple used WWDC 2026 to relaunch Siri as Siri AI, a more capable assistant with screen awareness, app context and a dedicated chatbot-style app. The system can reason from what is on a user’s screen, pull context from apps like Photos and Messages, and take systemwide actions. Apple is leaning heavily on privacy, saying requests will run on-device or through Private Cloud Compute without saving user data.

Why it matters

For casual iPhone users, this could finally make Siri useful. For people already using frontier models, Apple’s AI still looks behind the pack. The company is betting that privacy, hardware integration and distribution can narrow the gap.

The Deets

• Siri AI is powered by Apple’s own models, custom-built with help from Google’s Gemini models.
• The assistant can understand on-screen content and app context.
• A dedicated Siri AI app will store conversations and work across devices.
• It ships as a free update this fall for iPhone 15 Pro or newer.
• A public beta is expected next month.
• Launch access excludes the EU and China, with English-only availability noted in the reporting.

Key takeaway

Apple finally joined the AI assistant race, but it arrived looking more like a careful commuter than a speed demon.

🧩 Jargon Buster - Private Cloud Compute: Apple’s system for handling heavier AI requests on secure cloud servers while limiting data collection.

💸 Funding & Startups

The AI Money Machine Eating Bigger Checks

AI funding and IPO chatter are accelerating across the market. Moonshot is reportedly seeking $1B to $2B at a $30B valuation, PhysicsX raised $300M, and Perplexity is reportedly planning to go public in 2028. The theme is simple: models, chips, agents and infrastructure all need capital, and the dollar amounts are moving from startup round to national-industrial-policy scale.

Why it matters

The AI boom is a financing race. The companies with the most compelling products still need compute, distribution and patience, which means fundraising is becoming part of the product strategy.

The Deets

• Moonshot is reportedly raising $1B to $2B, fueled by growth in its Kimi models.
• PhysicsX raised $300M to speed up hardware design with AI.
• Perplexity reportedly plans to go public in 2028.
• The U.K. launched a £1.1B AI Hardware Plan, including £750M for a national supercomputer.
• Google and Nvidia are reportedly working with Intel as a backup manufacturer to TSMC.

Key takeaway

The AI winners may be the companies with the best models, the deepest pockets and the fewest power-grid headaches.

🧩 Jargon Buster - AI Hardware Plan: A government or company strategy to fund chips, supercomputers and infrastructure needed to train and run AI systems.

🧪 Research & Models

Anthropic: Biology AI Needs Better Plumbing

Anthropic argued that AI agents struggle in biology because the data systems are built for humans clicking through web pages, not for autonomous agents. In one virus-sequence retrieval task, frontier models returned wildly inconsistent results from the same prompt. When Anthropic added gget virus, a deterministic retrieval tool, every model jumped above 90% accuracy, narrowing the gap between cheaper and more expensive models.

Why it matters

The finding points to an underappreciated bottleneck: AI systems often fail because the environment around them is messy. Better tools, cleaner interfaces and auditable workflows can sometimes matter more than a bigger model.

The Deets

• Anthropic tested models on biology retrieval tasks.
• The same Ebola-related prompt produced inconsistent record counts.
• Adding gget virus made retrieval deterministic and improved performance.
• Cheaper models performed much closer to flagship models once the tool layer improved.
• The broader lesson applies beyond biology: agents need systems designed for machines, not just humans.

Key takeaway

The next AI breakthrough may come from boring infrastructure.

🧩 Jargon Buster - Deterministic Retrieval: A search or data process that returns consistent results for the same request, rather than varying unpredictably.

⚡ Quick Hits

• Microsoft had open-source tools hacked to steal passwords from AI developers.
• Meta removed face-recognition code from its smart-glasses app after scrutiny.
• Instacart is bringing AI-powered shopping carts into Weis Markets stores.
• AI data centers are increasingly being built in drought zones across the U.S.
• Google updated NotebookLM with agentic chat, giving each notebook a sandboxed computer to write and run code, plus new outputs like PDFs, spreadsheets and slides.

🧰 Tools Of The Day

• Granola + Claude: The workflow uses meeting notes to audit recurring meetings, identify repeated updates and generate pre-reads, action items and templates.
• Meta Enterprise Agent: Meta’s AI agent for customer sales and support.
• Devin Desktop: Cognition’s Windsurf rebrand for managing coding agents.
• Tely Health: A health care AI marketing tool aimed at helping providers show up in ChatGPT, Perplexity, Google and Claude results, then book patients into an EHR.
• Incogni: A privacy tool that removes personal data from the web to reduce scam and identity-theft exposure.

Today’s Sources: The Internet, The Rundown AI, AI Secret