AI Leadership Weekly

Top Stories

Thinking Machines now out of stealth
The former CTO of OpenAI, Mira Murati, has finally unveiled her stealth startup Thinking Machines. She is also joined by OpenAI cofounder John Schulman.

In a tweet, Murati said that Thinking Machines will focus on helping adapt AI systems to meet user needs, develop the foundations for "more capable AI systems", as well as foster a culture of open science.

OpenAI moving to prevent Musk takeover
The OpenAI board is considering instituting special voting powers to prevent a takeover by Elon Musk.

This is in the wake of news last week, where Musk offered to buy OpenAI for almost $100 billion, a move which wasn't considered a serious bid. It is clear, though, that the board and company want nothing to do with Musk, and are also discussing a "poison pill" option.

New York Times now using AI tools
The New York Times (NYT) has recently announced its new AI tool, Echo.

Echo is an internal tool that will assist with some writing, SEO headlines, and some code. They insist, though, that all stories will still be written by a human, but that the AI could assist to tighten up language or suggest SEO headlines. They also say that AI will help with reader accessibility, with AI voices reading news articles, and translations.

In Brief

• AI Pin creator, Humane, closing down and selling to HP

• Meta’s Llamacon set for April 29th

• OpenAI founger raising $30 billion+ valuation for new startup

Market Trends

AI can fix bugs, but can’t find them
OpenAI's own research has shown that, while LLMs (including their own) can squash many software bugs, they fail at identifying why those bugs occurred in the first place and continue to add more mistakes. Thes are interesting revelations, given recent comments from many tech CEOs, including Sam Atlman, that current-gen LLMs can replace lower level software developers. According to the researchers, “Results indicate that the real-world freelance work in our benchmark remains challenging for frontier language models."

AI agents vulnerable to simple manipulation
Researchers from Columbia University and the University of Maryland have found that AI agents with internet access can easily be manipulated into performing risky and damaging actions. The actions include revealing provate information, downloading malicious files, sending emails, etc. It isn't difficult to see how these exploits could majorly affect computer networks.

DeepSeek downloads suspended in South Korea
DeepSeek, the Chinese startup that recently released DeepSeek-R1, has halted downloads of its chatbot app in South Korea. The move came on Saturday when South Korea's Personal Information Protection Commission raised privacy concerns, although those who already have the app downloaded to their phones or PCs aren't affected, although are advised to either delete the app or to not enter any private information. DeepSeek is working with authorities to address these concerns.

Tools and Resources

Mistral Saba
An LLM designed for Middle Eastern and South Asian languages and cultures.

Assistant-UI
An open source library to add a conversational chat AI to your app.

SEO AI Agent
An AI agent to help automate your entire SEO workflow.

Recommended Reading

How to prompt o-series models
OpenAI has written up a series of best practices for prompting and interacting with its o-series of models.

Hit reply to let us know which of these stories you found the most important or surprising! And, if you’ve stumbled across an interesting link/tweet/news story of your own, send it our way at [email protected] It might just end up in the next issue!

Thanks for reading. Stay tuned for the next AI Leadership Weekly!

Brought to you by Data Wave your AI and Data Team as a Subscription.
Work with seasoned technology leaders who have taken Startups to IPO and led large transformation programmes.