The people who built Starview's tools started writing HTML in the late 1990s — on Geocities, hand-editing tags in Notepad, FTP-ing files to servers one at a time. We went through Flash, which was going to replace the web. We went through Web 2.0, which was going to make everything social and frictionless. We went through the mobile revolution, which required throwing out assumptions and starting over. The app economy. The cloud. The API era.
Each wave brought real tools that genuinely changed how things worked. Each wave also arrived surrounded by an enormous amount of noise — vendors, consultants, and conference speakers claiming that everything from before was now obsolete, and that whoever didn't adopt immediately would be left behind.
Most of that was wrong. The tools changed. The underlying question didn't: does it solve the actual problem? Does it work reliably? Can someone maintain it a year from now?
The Rush to AI Has a Paper Trail
AI is genuinely useful. More so than Flash ever was. But the race to bolt it onto everything — to be "AI-powered" before anyone asks what that means — is producing failures that are specific, documented, and worth knowing about.
A lawyer submitted six fake cases to a federal court. In 2023, attorney Steven Schwartz used ChatGPT to research precedents for a lawsuit against Avianca airline. The AI generated case citations that sounded completely real. None of them existed. He submitted the brief to the Southern District of New York. When opposing counsel couldn't locate the cases — including Varghese v. China Southern Airlines and Shaboon v. EgyptAir — Judge P. Kevin Castel investigated. Schwartz and his firm were sanctioned $5,000. The AI had produced the citations with complete confidence and zero accuracy.
CNET published AI-written financial advice that was factually wrong. Starting in late 2022, CNET quietly ran dozens of AI-generated articles on financial topics without disclosing their origin. An investigation by The Markup found errors throughout — including incorrect interest rate calculations. More than 40 articles required corrections, issued months after publication. By then the articles had been read widely, and some errors had been picked up by other sites.
Air Canada was held liable for its chatbot's bad advice. Passenger Jake Moffatt asked Air Canada's AI chatbot whether he could claim a bereavement fare after traveling to his grandmother's funeral. The chatbot said yes. The airline's actual policy said no — the discount had to be requested before travel. Air Canada's legal argument: the chatbot was "a separate legal entity" responsible for its own statements. A British Columbia tribunal rejected that in February 2024 and ordered Air Canada to honor the fare. The AI gave wrong information confidently. The company paid for it.
Why This Keeps Happening
AI tools produce text that sounds authoritative regardless of whether it's correct. That is not a bug that will get patched — it's structural. A language model predicts what text should come next based on patterns. It doesn't know when it's wrong. It doesn't flag uncertainty the way a careful person would. It will cite a case that doesn't exist in the same confident tone it uses to cite one that does.
This matters for businesses using AI to draft contracts, product descriptions, customer-facing policy, or anything where a confident wrong answer has consequences.
What the Right Balance Actually Looks Like
We use AI tools in our work. They make certain tasks significantly faster, and pretending otherwise would be dishonest. But every output gets reviewed by someone capable of catching an error — and we don't use AI for work where a wrong answer has consequences we can't catch before it ships.
That's not a philosophical position. It's just what working carefully looks like.
Flash was a great tool when you needed animation in a browser and had no other options. It was a terrible foundation for anything that needed to be maintained, secured, or accessed on a phone. Knowing that took time and a few burned projects to learn.
We're still early in the AI era. The burned projects are accumulating. The lesson is the same one it always is: a new tool doesn't replace judgment — it just gives you a faster way to make mistakes if you don't have any.