The AI news pile this weekend was a whole lot more than the usual hype. Google blew up its own search box. OpenAI started lining up to ring the bell. Anthropic shipped another model before anybody had time to finish testing the last one. Here's what actually matters and what I think about it.
Google Killed The Search Box And Handed It To Gemini
Google officially flipped the switch this week. The default search experience is now a conversational interface powered by Gemini 3.5 Flash. The classic ten blue links are not gone, but they are no longer what shows up first.
DuckDuckGo saw US app installs jump more than 18% week over week, with some days spiking above 30%. People are running for the exits.
This matters because Google search has been the front door of the internet for twenty five years. Switching the door to an AI agent rewrites how every business, every publisher, and every regular person finds anything online. Publishers who depended on clicks are about to have a real bad quarter.
My take: this was inevitable but the rollout feels too fast. Google clearly decided it was better to cannibalize itself than let somebody else do it. Fair enough. But betting the whole search business on Gemini being right most of the time is a heck of a gamble. I give it six months before the lawsuits start over hallucinated answers.
OpenAI Files For An IPO Worth A Pile Of Money
Reuters and the Wall Street Journal both reported OpenAI is confidentially filing IPO paperwork in the next few weeks, aiming for a September 2026 listing. The company has already raised 186 billion dollars in private funding.
That is not a typo. 186 billion. Before going public.
This matters because OpenAI going public turns the entire AI race into a quarterly earnings show. Right now Sam Altman can spend whatever he wants on compute and answer to nobody. After September he has to look Wall Street in the eye every ninety days and explain why he is lighting cash on fire.
My take: I do not see how the numbers work at any sane valuation. They are burning more cash than most countries. Either the public markets price it like a regular software business and the stock craters, or it gets priced like a religion and we get a bubble that makes 1999 look quaint. I am leaning toward bubble.
Anthropic Shipped Opus 4.8 Forty One Days After 4.7
Anthropic dropped Claude Opus 4.8 this week. Same price as the previous version. Forty one days after Opus 4.7.
That release cadence is absurd. Most software companies cannot ship a patch in forty one days, much less a frontier model.
This matters because the model release treadmill is now the actual competition. Whoever can iterate fastest on training and post training wins, and Anthropic just put up a number that nobody else has matched. Holding the price flat while improving the model is also a quiet shot at OpenAI, who keep nudging prices up.
My take: I switched our internal stuff to Opus 4.7 about two weeks ago and now I have to do it again. That is a champagne problem. Keep them coming.
Huawei Says The Transistor Race Is Over
Huawei unveiled what they call the Tau Scaling Law. The pitch is that we have hit physical limits on shrinking transistors, so the next decade of chip progress comes from cutting signal transmission times through the chip instead.
This is a real engineering statement, not a marketing slide. Moore's Law has been wheezing for years and somebody had to say it out loud.
This matters because if Huawei is right, the chip game changes from a fab race to a packaging and interconnect race. That is a different competition with different winners. TSMC has the fabs but Huawei is betting the next moat is somewhere else.
My take: I think they are mostly right about the physics and mostly self serving about the timing. Of course the company that cannot get the newest fab equipment wants to argue that fab equipment does not matter anymore. Still, the underlying point about signal transmission being the new bottleneck is solid and US chip designers should be paying attention.
That is the weekend. Three model and product moves that reshape the industry and one engineering claim that might reshape it later. Talk to y'all tomorrow.