Ilya's SSI skyrockets to $32B

PLUS: Ex-OpenAI staff opposes company's for-profit shift

Author

Rowan Cheung
April 14, 2025

Good morning, AI enthusiasts. Former OpenAI chief scientist Ilya Sutskever just reportedly raised another massive round for his AI startup, Safe Superintelligence Inc. (SSI), catapulting its valuation to $32B in mere months.

But with no product in sight—just a bold vision for superintelligence and Sutskever's reputation—is this the most audacious bet yet in the AI funding frenzy, or a glimpse of Silicon Valley's new normal?

In today’s AI rundown:

  • Ilya’s SSI raises $2B at $32B valuation

  • Ex-OpenAI staff push back on for-profit shift

  • Build an AI-powered lead outreach automation

  • AI surpasses experts in tuberculosis diagnosis

  • 4 new AI tools & 4 job opportunities

LATEST DEVELOPMENTS

SAFE SUPERINTELLIGENCE INC.

Image source: SSI / Stanford University

The Rundown: Safe Superintelligence Inc. (SSI), co-founded by former OpenAI chief scientist Ilya Sutskever, just reportedly raised $2 billion at a post-money valuation of $32 billion, becoming one of the highest-valued startups just months after launch.

The details:

  • The $2B round has been led by Greenoaks (with $500M), with Lightspeed Venture Partners and Andreessen Horowitz in participation, FT reported.

  • A separate report from Reuters noted that Alphabet and Nvidia are also backing SSI, though their investment amount remains undisclosed.

  • The AI startup has been laser-focused on building “superintelligence” that goes beyond human-level AGI while making sure “safety always remains ahead.”

  • Previously, Sutskever told investors that the company has “identified a different mountain to climb,” hinting at a unique approach to AI development.

Why it matters: Even without a concrete product plan, SSI continues to rise, with its valuation growing sixfold since September 2024. The surge, alongside news of Mira Murati’s Thinking Machines planning a raise, reflects growing investor interest in AI startups led by prominent researchers, especially those emerging from OpenAI’s orbit.

TOGETHER WITH SANA

The Rundown: Sana provides a powerful enterprise AI platform designed to let every team build no-code AI agents grounded in your organization's unique data.

Upgrade your workflow with AI that can:

  • Find the exact info you need

  • Answer and brainstorm any idea

  • Recap meeting actions and automate repetitive tasks

  • Solve complex, multi-step problems in seconds

OPENAI

Image source: Getty Images

The Rundown: Twelve former OpenAI employees, who served in technical and leadership roles between 2018 and 2024, just filed a proposed amicus brief supporting Elon Musk's lawsuit challenging the AI lab's shift away from its nonprofit origins.

The details:

  • The brief makes the case that if OpenAI’s non-profit wing cedes its controlling stake in business, it would “fundamentally violate its mission statement.”

  • It adds that OpenAI’s restructuring would also “breach the trust of employees, donors, and other stakeholders” who supported the lab for its mission.

  • Todor Markov, who is now at Anthropic, called Altman “a person of low integrity” who used the charter merely as a “smoke screen” to attract talent.

  • They all noted the court should recognize maintaining the nonprofit is essential to ensure AGI benefits humanity rather than serving narrow financial interests.”

Why it matters: If admitted to the court record, these testimonies from former insiders could strengthen Musk's case ahead of the spring 2026 trial. OpenAI, for its part, says the non-profit remains intact and calls the changes a restructuring of its existing for-profit arm into a public benefit corporation while preserving the original mission.

AI TRAINING

The Rundown: In this tutorial, you’ll learn how to use Lindy AI's new loop feature to create an automated system that researches prospects and drafts personalized emails at scale with zero manual effort.

Step-by-step:

  1. Set your Lindy AI agent context by adding a description like "You are an outreach agent that has access to spreadsheets, researches leads, and drafts personalized emails".

  2. Create a workflow starting with "Message Received" trigger and an AI Agent configured to process spreadsheets of leads.

  3. Add an "Enter Loop" node that processes leads in parallel, with "Search Perplexity" and "Draft Email" nodes inside the loop.

  4. Finalize with an "Exit Loop" node and a summary AI Agent, then test your workflow with a sample spreadsheet.

Pro tip: The Rundown University members get 15% off Lindy Pro plans here.

PRESENTED BY WRITER

The Rundown: Despite the buzz, agentic AI has struggled to deliver — until now. Writer’s end-to-end platform finally makes agentic workflows real, effective, and scalable for enterprise.

In this product showcase, you'll:

  • Experience real agentic workflows live

  • Hear success stories from beta testers

  • Learn strategies to align your IT and business teams

Watch on demand and see the first platform for end-to-end agentic AI in action.

AI RESEARCH

Image source: Ideogram/The Rundown

The Rundown: A new study led by Swiss researchers from Lausanne University Hospital demonstrated that AI can diagnose pulmonary tuberculosis with greater accuracy than human experts, exceeding WHO standards for non-sputum TB tests.

The details:

  • Presented at ESCMID Global 2025, the study introduced ULTR-AI, an AI system trained to read lung ultrasound images from smartphone-connected devices.

  • The system uses a combination of three different models to merge image interpretation and pattern detection and optimize diagnosis accuracy.

  • When tested on 504 patients (38% of whom had confirmed TB), it achieved 93% sensitivity and 81% specificity, beating human expert performance by 9%.

  • The AI can identify subtle patterns that humans often miss, including small pleural lesions invisible to the naked eye.

Why it matters: With TB cases rising and diagnostics scarce or unaffordable in low-resource settings, this AI system could revolutionize triage by providing faster, cheaper, and scalable testing. And since it runs in real time through a smartphone app, even minimally trained healthcare workers can use it effectively in remote locations.

QUICK HITS

  • 🧠 Seed-Thinking-v1.5 - ByteDance’s reasoning AI that beats Deepseek R1

  • 🤖 AI HQ - Writer’s end-to-end platform for building and supervising AI agents

  • 🗣️ Amazon Nova Sonic - Amazon's Speech-to-speech AI on Bedrock

  • 📊 Gemini in Sheets - Access Google’s Gemini AI models in Google Sheets

Meta’s unmodified, release version of Llama 4 Maverick appeared on LMArena, ranking below months-old models, including Gemini 1.5 Pro and Claude 3.5 Sonnet.

DeepMind CEO Demis Hassabis mentioned that the company plans to combine Gemini and Veo models into a unified omni model with better world understanding.

Netflix is reportedly working with OpenAI on a revamped search experience, allowing users to look up content using different new parameters, including their mood.

OpenAI beefed up its security with a new Verified Organization status, which will be required to unlock API access to its advanced models and capabilities.

OpenAI CEO Sam Altman said that the company plans to release an open-source model that would be "near the frontier.”

Elon Musk’s xAI started rolling out the memory feature to its Grok AI assistant, following a similar move from OpenAI last week.

COMMUNITY

We just did a full workshop on n8n, where we teach you how to create your own AI assistant that speeds up your workflow and helps you get more done, led by Dr. Alvaro Cintas, The Rundown’s AI professor.

Watch it here. Not a member? Join The Rundown University on a 14-day free trial.

That's it for today!

Before you go we’d love to know what you thought of today's newsletter to help us improve The Rundown experience for you.

Login or Subscribe to participate in polls.

See you soon,

Rowan, Joey, Zach, Alvaro, Jason, and Shubham —The Rundown’s editorial team

Reply

or to participate.