The Wand of Wonders
It is so hard and exhausting to keep up with everything that’s happening in the AI universe. It feels like the new models or their derivatives are released every week. Each model surpasses its competition by several percent points on a metric which nobody knows what it measures but for sure it is very important.
At the same speed, products are also dropped because they are deemed distractive to the overall direction of the company. A good example is OpenAI’s Atlas — ‘The browser with ChatGPT built in.’
Introduced to us in October 2025 with the following — “AI gives us a rare moment to rethink what it means to use the web. Last year, we added search in ChatGPT so you could instantly find timely information from across the internet — and it quickly became one of our most-used features. But your browser is where all of your work, tools, and context come together. A browser built with ChatGPT takes us closer to a true super-assistant that understands your world and helps you achieve your goals.”
It was one of these weapons in the fight for your desktop. Well, that was then, but we live in now. While it was positioned as “your browser is where all of your work, tools, and context come together” it didn’t come close enough to a ‘true super-assistant’ that truly understands you and your world and to help you to achieve your goals.
That’s why OpenAI is sunsetting (to use a word like this feels so romantic) the Atlas browser and introducing a desktop app, ChatGPT Work — ‘ChatGPT for your most ambitious work’. It combines two things — ChatGPT and Codex. Codex is the tool which is meant for software developers to generate code at light speed. ChatGPT — using plugins — can connect to various other products, from your email, files, Slack, calendar to name a few.
The list of what this new app can do goes on for pages and it is nicely summarized at the end of the announcement — “This is the first step towards a broader vision for ChatGPT — where intelligence goes beyond answering questions to helping everyone turn their biggest ideas into reality.”
Please note that the objective is no longer to replace people, but to turn their dreams into reality.
Launching ChatGPT Work is an answer to Anthropic’s Claude Cowork and Microsoft’s Copilot Cowork.
What else has OpenAI introduced to the market? I thought you would never ask.
First, we have a new keyboard. It is called Codex Micro and it allows you to directly control your agent workspace. One of the features is “Set the Brainpower — Turn the dial to adjust reasoning level in the moment. Stay fast for simple tasks, or crank it up when the work calls for heavier thinking.” Signs of the times. I have no idea what it means to “crank it up when the work calls for heavier thinking.”
But if you have an extra $70 lying around, you can get a ChatGPT Basketball which in the words of the marketing team at OpenAI is “a physical reminder that creativity doesn’t just live on our screens.” And I was questioning how it is possible that OpenAI spent $5.73 billion on sales and marketing.
With all these new products coming from OpenAI, Anthropic or Microsoft, we must see some positive impact in organizations which are using these tools. We must and we do!
The article from Fortune says it all — ‘Bank of America’s AI bet is starting to pay off, its CFO says’. As reported “… its Q2 earnings on Tuesday, which included double‑digit profit growth, a 59% efficiency ratio, and roughly 17% return on tangible common equity. CFO Alastair Borthwick increasingly frames those gains as evidence of AI‑enabled productivity rather than just cost‑cutting and rate tailwinds.” Mr. Borthwick also added that “New AI capabilities now allow more than 200,000 of our employees to work more effectively, and they’ve helped contribute to producing a 59% efficiency ratio, a roughly 360 basis point improvement from last year.”
However, when asked what exactly is the definition of the ‘AI efficiency ratio,’ the answer was ‘expenses divided by revenue.’ <Sigh>
To bring things into perspective. As Bank of America announced its rosy numbers, so did the other banks:
JPMorgan delivered its most profitable quarter ever. Nearly every division beat expectations, helping America’s biggest bank earn $21.2 billion in profit ‐ up 41% from a year ago.
Goldman Sachs turned last quarter’s market chaos into record results, for the third time running. Stock-trading revenue surged 72%, pushing the bank’s overall figure to an all-time high.
Citigroup’s stock-trading business posted record revenue, while four of its five core divisions beat analysts’ forecasts. The result: the bottom line climbed 45%, well above predictions.
Wells Fargo rounded out the strong showing. Higher fees from wealth management and investment banking helped the lender beat expectations, with overall profit rising 17%.
I guess they all run AI. Otherwise how could you explain these awesome numbers … Or maybe, just maybe, the revenue grew faster than the expenses.
I am not disputing that people in these companies are finding use cases where this new technology helps them in their daily tasks, but given the vague answer from the BoA’s CFO, it is still not significant enough to have a real impact on revenue or any efficiency ratio.
When money rolls in, efficiency is never the objective. Once the bank’s (or any other company’s) revenue takes a hit, that will be exactly when the CFOs start asking how we can do more with less. That will be the time when it will be measured and reported.
The recurrent pattern? In the meantime, OpenAI, together with others, will keep searching for the super-app, the magic Wand of Wonders, and hope that they start making money with the magic spell. Sadly, people working there are not getting help from their own tech.