More than tart takes, these come with 8 ounces of distilled vinegar! Sip slowly and savor this weekend.
The always controversial Ed Zitron goes there. Way out there to challenge your preconceptions. If you’ve ever awakened at 3am wondering where in blue blazes AI is going, agonizing over where it fits in your platform or services, who is it really benefiting other than OpenAI and Anthropic, why VCs are so stingy because your tech isn’t AI All Over, and why despite the hype billions are going down a rabbit hole seemingly with no bottom [TTA 2 June, 25 June] –this article (not premium paywalled) is for you.
“Cargo Culture” refers to the cargo cults of the South Pacific. They originated after native contact with Westerners, usually Americans who landed in ships or built airstrips, mostly during WWII. When the Americans left, the indigenous peoples tried to recreate the “magic” of goods coming off the ship or from airplanes by creating a religion and ritual around these ‘miracles’ without, of course, the corresponding methods of delivery. Your Editor also found an IT version of this called “cargo cult programming“defined in a Wikipedia entry as “computing slang to describe the inclusion of code that serves no purpose in a program, indicating a lack of understanding of the program structure by the programmer.“
Mr. Zitron’s “Cargo Culture” refers to Silicon Valley. The endlessly reinforced belief system of the interlocked software, venture capital, and private equity businesses was built on decades of mythology of endless growth and endless Christensenian “disruption”. They want to believe that there will always be hypergrowth tech companies, always another ‘disruptor’. And the money to fund them will be equally endless. (Forgotten are the failures like Theranos and the billions down the drain.)
The reality, as he painstakingly explains, is that Silicon Valley and tech are, right now, fresh out of ideas. Worse, the software industry is declining. There’s no Google Search or iPhone waiting to be plucked from the ether. There is no New Big Transformative Idea. Company growth is incremental at best, declining or failing at worst. But instead of adjusting, the mythology is ingrained, and to doubt is heretical.
Any dissension, whether corporate or media, means you are Voted Off The Island.
The AI Bubble fits neatly and virally into Silicon Valley’s Cargo Culture and need for a disruptor. It’s a viral product, it was different, it meets a need for automating processes. It makes search a lot quicker. But will LLMs and agentic AI actually change anything other than make it faster, or reduce the need to perform your own logic in building a case/article–at amazingly high expense? Other than the spend, what’s exponential about it? Everyone’s buying a lot of GPUs–heck, Oracle at Larry Ellison’s behest is betting the company on moving out of software, including that ever-so-sticky EHR–and becoming a data center constructor and lessor to Open AI and Meta lest Microsoft eat his lunch. After a certain point, the beliefs become circular and self-reinforcing.
Mr. Zitron argues the points at length. It’s minimum a 30-minute read and not an easy one. Stay with it. It challenges a lot of premises. It will make you think differently the next time you’re pitched on a company’s miraculous AI software–or if you are a developer, understanding your customer’s needs and likely confusion. It is 100% guaranteed to make you think about Silicon Valley, their modus operandi, and how it’s inflated the whole AI bubble.
Sergei Polevikov shines a light on a recent Nature Medicine study comparing specialized clinical LLMs against the generics–and finds that the generics win. The peer-reviewed study pitted specialized clinical tools like OpenEvidence and Wolters Kluwer’s UpToDate ExpertAI against general-purpose AI tools–ChatGPT-5.2, Gemini 3.1 Pro, and Claude Opus 4.6–using three different benchmarks*. The specialized tools lost every time. What was worse was OpenEvidence’s unprofessional overreaction on social media, namely X. Instead of responding directly to Nature Medicine as a comment, then publishing a peer-reviewed paper refuting the Nature Medicine team findings, citing their own research from independent studies, conducting fresh research, even better admitting that some of the research team’s findings were valid and appreciated, their ‘Twitterstorm’ only created more PR blowback. Social media amplified the damaging reach of one academic study. (Having managed a few ‘crisis communications’ as a marketer, the worst thing that you can do is bang back in anger, though cooling corporate tempers might require a Titanic-sized iceberg of ‘splainin’.)
Mr. Polevikov’s free excerpt of his lengthy article, OpenEvidence Goes Hippocratic AI, is published on LinkedIn, but why not kick in a few $$, support his independent journalism, and subscribe to AI Health Uncut (Substack)?
The Hippocratic AI reference is to that company’s ‘bare knucks’ approach to any criticism of their platforms.
*FTA Nature Medicine: (1) 500 MedQA questions testing medical knowledge, (2) 500 HealthBench items measuring alignment with clinicians and (3) the real clinical queries (RCQ) benchmark, built from 100 de-identified queries from physicians to a general-purpose language model in a live clinical environment.
Leave a Reply