A week in Generative AI: OpenAI, Firefly & Robots
News for the week ending 2nd November 2025
No huge game changing new product announcements this week, which makes a welcome change of pace but we did have big news around the restructuring of OpenAI, the release of Firefly Image 5 by Adobe, and NVIDIA hit another financial milestone, being the first company to reach a $5tn valuation just 3 months after it became the first company to reach a $4tn valuation.
In Ethics News, Amazon’s CEO says that AI isn’t responsible for its latest round of mass layoffs, Trump bought $80bn of Nuclear Reactors to power AI infrastructure, and Character.AI is ending its chatbot experience for kids.
There’s also a great article from John Gruber over at Daring Fireball about the recent release of ChatGPT Atlas.
OpenAI completes conversion to for-profit business after lengthy legal saga
OpenAI this week finalised a big restructuring and resetting of the terms associated with Microsoft’s investment in the frontier AI company. TLDR is OpenAI is no longer a non-profit but a public benefit corporation, a type of for-profit entity that expresses commitment to bettering society.
Under the terms of the new ownership structure, Microsoft owns a 27% stake in OpenAI, which has been valued at $500bn. 26% of OpenAI is owned by the non-profit OpenAI Foundation, and 30% is owned by OpenAI employees. The remaining c.15% belongs to other investors.
Below are some interesting features of the new ownership structure and Microsoft partnership:
Microsoft holds exclusive IP rights on all OpenAI models and software products until Artificial General Intelligence (AGI) is declared and verified by an independent expert panel, or through 2023, whichever is first.
Microsoft’s IP rights now explicitly exclude any OpenAI consumer hardware products, paving the way for the development of mobile devices and robotics platforms.
The non-profit OpenAI Foundation will grant $25bn toward health, curing diseases, and protecting against cybersecurity risks as their first big investment.
OpenAI is now able to release open weight models that meet requisite capability criteria.
In the short term, these changes won’t really impact anything outside of OpenAI itself, but does pave the way for an IPO, the raising of more funds, and the development of hardware products. We’ll probably start seeing the start of this in 2026.
Adobe Firefly Image 5 brings support for layers, will let creators make custom models
Adobe released the latest version of Firefly Image this week, with lots of great improvements, the biggest being the ability to generate 4 megapixel images, up from just 1 megapixel with the previous generation.
Artists can also train their own image model using their existing art and Firefly Image 5 allows creators to edit the different layers of the image with prompts. There are lots of other great features that give people much more control over the images they are generating and it feels like we’re now getting to a level of maturity with image generation technology that will allow it to be used more widely across the creative industries.
NVIDIA becomes world’s first $5tn company amid stock market and AI boom
This has become a bit of a recurring theme this year - every couple of months I share that NVIDIA has become the first company to reach the next trillion dollar milestone. Let’s recap:
In May 2023, NVIDIA reached a $1tn valuation
In February 2024, NVIDIA reached a $2tn valuation (9 months later)
In June 2024, NVIDIA reached a $3tn valuation (4 months later)
In July 2025, NVIDIA became the first $4tn company (13 months later)
Now, in October 2025, NVDIA has become the first $5tn company (3 months later)
The next most valuable companies are Apple and Microsoft who are both valued around $4tn, with the three combined accounting for 20% of the entire S&P 500.
The Problem with this Humanoid Robot
On one hand this is a video from MKBHD about the 1X Neo humanoid robot that can be yours next year if you live in the US and want to spend $500 per month, or $20k. On the other hand this is a video about the huge gap there still is between the reality and the promise of AI/robotics.
Unfortunately right now a lot of AI/robotics products are being sold on the promise, not the reality. Part of this is just riding the AI hype wave, and part of it is needing to sell the product to gather the data to deliver on the original promise of the product. This uses early adopters/enthusiasts as beta testers to provide the data, and I’m not sure I’m ok with this. Seems like false advertising if the product delivered doesn’t deliver on the promise from day one.
Web 4.0
‘A good moment in time for us’: Firefox head on AI browsers and what’s next for the web
PayPal partners with OpenAI to let users pay for their shopping within ChatGPT
AI Ethics News
Amazon CEO Now Says AI Is Not Responsible for Recent Layoffs
OpenAI says over a million people talk to ChatGPT about suicide weekly
Trump Just Bought $80 Billion Worth of Nuclear Reactors to Keep the AI Bubble Cooking
‘Game of Thrones’ Author George R.R. Martin’s OpenAI Lawsuit Takes a Step Forward
Teenage boys using ‘personalised’ AI for therapy and romance, survey finds
AI can help authors beat writer’s block, says Bloomsbury chief
New AI-powered anti-scam tool wins praise from UK fraud minister
Long Reads
Daring Fireball - Thoughts, Observations, and Links Regarding ChatGPT Atlas
Emergent Behaviour - Some confirmations from OpenAI
“The future is already here, it’s just not evenly distributed.“
William Gibson