Cognitive Warmup. Earlier this year, to the cheers of the short-sighted and the foolish, Microsoft claimed that about 30% of the company’s code is now written by AI. Fast forward to the past few months, Windows 11 updates have become a complete mess. The October update breaks the Windows Recovery Environment by preventing input from USB keyboard or mouse accessories. The August update caused the hard drive to become inaccessible or data to be lost. There were installation failures in the big 24H2 update. Also, websites failing to load, random app crashing or freezing behavior, and even Windows’ inability to understand the “Update and Shut Down” command have been noted as highlights. I’ve said this before, but perhaps a reiteration is in order – Humans > AI. fact.
algorithm
We talk about why everyone wants to fail, Satya Nadella’s comments on defining destiny and the importance of sovereignty for AI companies, and is the rabbit out of the suspect’s hat?
The “most likely to fail” moment of confusion
I would say two themes came up the most at Cerebral Valley Conference 2025 – there is a lot of unease among investors around all things AI, and second, Perplexity which is constantly trying to get Google’s attention and is also now in a legal dispute with Amazon was voted by attendees as most likely to fail. And keep in mind, the audience also included investors and founders. Originally, the question asked was – if you could short a $1B+ valuation startup, which one would it be? Perplexity won the gold medal, followed by OpenAI and the group of Cursor, Figure, Harvey, Mercer, Mistral and Thinking Machines that won the bronze medal. This may be an informal survey, but it captures the pulse of the matter. There is much more to the AI ​​bubble now. Perplexity can be cynical all they want (Business Insider reports that Perplexity spokesperson Jesse Dwyer replied in an email: “Oh my God, this sounds like a Judgmental Valley conference.”), but the reality is simple. The perception of being the winner or leading the way, but they have to grapple with the reality of the broader AI field – growing investment needs, increasing power requirements and in an ecosystem of perplexity, are competing in the same space as Google and OpenAI (which, mind you, are also facing headwinds) for search and AI browsers.
Estimates show that Perplexity processes 780 million queries monthly, while Google handles 13.7 billion queries daily. If Perplexity thinks search is easy (just becoming the second largest behind Google), it’s not. Microsoft incorporated ChatGPT into Bing in February 2023. in 2025, but the market share increased from about 3.4% to 4.31% by October 2025, and this is for a tech company that has a deeper reach and presence on millions of devices globally. The reality is that Perplexity is wasting cash, and fighting Amazon in court. It seems investors love the swagger but quietly worry about unit economics, bandwidth bills and whether AI search can sustain itself as a business model. Some may think this is the hottest story in AI search, but signs point to a simple truth – the greater the hype, the harder the landing. Don’t shoot the messenger.
Satya Nadella warns about extractive AI…
On a Friday night in that time zone (the peak working environment in Silicon Valley), Saturday morning for us, Microsoft’s CEO posted a thoughtful note about the dangers of “extractive” AI partnerships. Even mentioned Bill Gates’s lessons on defining platform and economic value. “Even in our somewhat zero-sum mentality industry, we can create partnerships that create value for all parties involved. Our partnership with OpenAI is a great example. Our investment helped them scale massively; their research fueled our own innovation. That’s what healthy platforms and partners do – they catalyze and compound progress,” he wrote.
Commendable sentiment, except for the part where Microsoft has invested $13 billion in OpenAI over the past few years, exclusive hosting of OpenAI’s commercial API products on Azure, supercomputing infrastructure, dominant position with multiple generations of frontier training, etc. Satya warns that every company must ensure that they have control of their destiny and sovereignty versus a press release with a tech/AI company or lose value through what may seem like a partnership, which is actually extraction in terms of value exchange in the long run. What Nadella is forgetting is simple – consumers don’t care about corporate ethics. What they care about is that Windows 11 is shoving AI in their face (more on this in a bit) while at the same time not helping usability.
Rabbit’s “We Can’t Pay You, But Can Trust Foresight” Era
Do you remember the time when everyone thought AI wearables would be a big deal? Rabbit, the company behind the ill-fated R1, is reportedly months behind on pay. And also, teasing a new piece of AI hardware. A hardware teaser suggests they’ve decided the best way to solve the financial crisis is by trying to build another expensive gadget. In the AI ​​winter-with-heatstroke of 2025, none of this is surprising.
ready
Mozilla has launched something fresh and non-gimmick in the AI-drenched browser wars: AI Window in Firefox. But should you care?
Rather than pushing a new generation of “AI everywhere” or “AI browsers” like Microsoft’s, or models sitting behind telemetry data like Google’s approach, Mozilla insists they want to be respectful. This leads us to an AI window, which is essentially a browser window with an AI chatbot in the sidebar, and something that’s being built with user feedback. This will sit next to the standard window (or current window) and the private window. The company says, “In Firefox, you’ll never be locked into an ecosystem or have AI forced into your browsing experience. You decide when, how, or not to use it.” For now the AI ​​Window feature requires you to join the waiting list.
This is an approach that can easily work. Mozilla talks about model flexibility, meaning that the option of AI models that you can access through a chatbot is coming at some point in the future. At present, no specific information has been shared. Vision is an AI window that can summarize pages, rewrite text, translate, brainstorm – but only when implemented, and only using the model you choose.
If you want to balance AI in the browser without sacrificing privacy or control, yes (for example, I wouldn’t recommend a generic AI browser if you want to keep your internet banking passwords safe), along with an AI-free experience, Firefox has probably got the balance right. It’s the closest we’ve seen to AI that actually respects the user – which probably explains why no other AI company…let alone.
Thinking
“The team (and I) take in a lot of feedback. We balance what we see in our product feedback system and what we hear directly. They don’t always match, but both are important. I read the comments and focus on things like reliability, performance, ease of use, and more.”
– Pavan Davuluri, President, Windows + Devices.
Windows 11 is becoming the world’s most extensive (and expensive) beta test, with almost no one from its millions-strong user base actually signing up for it. These comments from the Windows boss come after he posted something on X a few days ago after which he received so much constructive feedback that he had to stop replying to that post completely. The reality is that Microsoft has changed things so much and so often with Windows 11 that it is no longer recognizable.
Context: It all started a few days ago, when Davuluri posted, “Windows is evolving into an agentic OS, connecting devices, cloud, and AI to unlock intelligent productivity and secure work anywhere.” As I said, no replies came back.
One user wrote, “This is growing into a product that is driving people towards Mac and Linux.” “Stop this nonsense. Nobody wants this. You live in a Twitter bubble where AI will create lots of wealth and if you don’t adopt it now you will be destroyed. But your users are not in this bubble. They don’t care about this nonsense,” said another, apparently saying nothing. According to another opinion, “Microsoft is going to collapse because of these moves by investors who are abandoning customers.” These are some of the good ones I have prepared for this family audience.
This push toward AI obsession is not new. For over a year now, Windows has had AI features available to unsuspecting users. Most of them just get in the way, feel half-baked, and slow down the process and the system. Windows users are clearly tired. The response is obvious – stop loading AI features that people neither asked for nor benefit from. Suddenly, the dream of the “AI PC” (oh yes, Microsoft forced PC makers to rebrand their entire portfolio around an AI theme) feels like a forced march rather than an upgrade cycle.
A reality check: Microsoft probably isn’t wrong that AI will be central to the next decade of personal computing. Equally, it can also happen. Nobody knows, why throw caution to the wind at this point? Shipping half-baked features, entangling users in AI workflows, and treating the OS as a billboard for AI and cloud upsells is eroding trust faster than Microsoft anticipated. Or maybe it does, who knows. Windows 11 doesn’t need more AI – it needs stability, performance, a seamless user experience, and a return to the simplicity of Windows 7. The real problem is Windows doesn’t have AI. This is the assumption that every feature needs to be AI-shaped, even when the user doesn’t explicitly want it to.
