Cache

Picture opening wearecoded.com in the morning. Your phone doesn't send a request all the way to our server every time you scroll the page - it's already kept the images, the fonts, part of the text close by, because it asked for them five minutes ago. It doesn't fetch them again from far away, it grabs them from nearby. That's the cache. It's not a trick, it's not a deception. Just saving a trip.

The problem comes when something changes and the phone is still looking at the old copy. I've changed a product's price, uploaded a new photo, fixed a bug, and the person on the other end still sees the version from before. Not because something broke deep down somewhere, but because their copy is outdated and they don't know they're holding something old.

The cache doesn't hide information from you. It only hides the time you'd have lost asking for it again.

So "clear the cache" isn't a magic phrase tech people say to sound important. It literally tells the device: throw out the old copies, go ask again from far away, grab whatever's there now. In most cases it really does fix things. We've just stopped looking at something old.

It's everywhere - in the browser, in apps, even inside the sites themselves behind the scenes. Every layer keeps its own copy of the most frequently requested things, so it doesn't have to send a request all the way to the end of the chain every time. The more layers, the faster things feel, but also the more places where an old copy can hide unnoticed.

There is a cache inside language models too: the KV cache is the memory of text already processed, so the model doesn't compute it again. On October 26, 2026 Cloudflare explained that a bug in serving GLM-5.3 on Workers AI returned data from other customers' requests in 155 responses, between August 27 and September 9, and only for prompts longer than 681,280 tokens. According to Cloudflare, a small number of customers were affected and there is no evidence of deliberate exploitation. The model isn't to blame, the memory between requests is. A cache saves time, but when it is shared, a bug in it shows one person someone else's data.

Here's the detail

A drawer with the things you use most stays within reach, because nobody wants to dig through the cabinet every time. The cache works on the same principle: close, fast, convenient. Except the drawer doesn't tidy itself - and sometimes the digital copy doesn't either, until something explicitly tells it to check again.

So when something "won't update", my first question isn't whether it's broken. My first question is where it's looking from - the drawer, or the real cabinet.

The visual is generated code-art, no third-party images.