The Baker's Over-Fed Yeast: On the Swollen Promise of Image Modern Formats

There’s a piece of advice handed down in our craft with the certainty of a natural law: Convert your images to modern formats. AVIF or WebP, we are told, is the unambiguous answer. It is the clearest path to a lighter page, a faster load, a better Core Web Vital. We treat this conversion like a simple act of alchemy, turning the leaden JPEG into golden bytes. But I’ve been kneading this dough for a while now, and I’ve come to suspect we’re often over-proofing the loaf. In our zeal for the modern, we risk creating a different kind of bloat—not of bytes, but of process, expectation, and brittle complexity.

The Leaven of Complexity

The promise is seductive because it is, in isolated tests, absolutely true. A well-encoded AVIF can be staggeringly smaller than its JPEG ancestor at comparable quality. The trouble begins when we move from the laboratory to the bakery, where we must feed a thousand different ovens. We are not serving a single, static image to a single, modern browser. We are building a system that must degrade gracefully, that must account for the visitor on an older device, the CMS editor uploading a PNG, the CDN that must store and serve multiple variants. The ‘modern format’ stops being a simple image and becomes a node in a sprawling pipeline of fallbacks, feature detection, and storage duplication.

This pipeline itself has a weight. It is the cognitive load on every developer who must now configure, not just a compressor, but a whole suite of converters and `` element syntax. It is the infrastructural cost of storing originals, WebPs, and AVIFs. It is the silent failure when a CMS’s auto-conversion botches an image with subtle transparency or fine text, and no human eye catches it because they only checked the ‘modern’ version in Chrome. The bytes we save on the wire can be invisibly spent tenfold in the complexity of our systems and the fragility of our workflows.

Worse, I fear this mandate can become a substitute for judgment. I’ve seen teams obsess over converting every tiny, 4KB favicon-style logo to AVIF while leaving a 2MB hero image sourced from marketing poorly cropped and unoptimized at its fundamental dimensions. The modern format becomes a talisman, a ritual we perform to feel we have done the work of performance, while the low-hanging fruit of sensible sizing, appropriate scale, and intelligent lazy loading goes ignored. It’s like using the finest, most expensive yeast to leaven dough made with stale, low-grade flour.

This is not a Luddite’s plea to abandon AVIF or WebP. They are remarkable tools. This is a plea for perspective. The modern format is the final, delicate fold in the dough, not the foundation of the bread. The foundation is still the choice of image, its purpose, its dimensions, its compression quality (even if that compression is old-fashioned JPEG). It is the `` tag’s humble `width` and `height` attributes doing the quiet, heroic work of layout stability. Before we reach for the advanced codec, we must ask the simpler, harder questions: Does this image need to be here at all? Is it the right size? Is it properly focused?

Without that foundation, our modern formats are just empty air, a loaf that looks impressive but collapses under the first touch of reality. The goal was never to serve an AVIF. The goal was to serve a user, quickly and gracefully. Let’s not let the yeast of the new make us forget the weight of the grain.

Notes & further reading

A few pages I came back to while writing this: