Post by @redsakana@infosec.exchange

2026-08-24 08:37 UTC

@misty@digipres.club I would expect that this will be "solved" with browser-side hacks one way or another, like shipping a ton of stupid shit was "solved" with HTTP/[23] multiplexing/server push. Dayjob customers running largish mainstream sites _sometimes_ have a few people who can think in terms of esoteric low-level stuff like memory having limits, but even those that do are effectively captive to decades of bad business and architectural decisions. Many run on a rotating cast of hourly-rates consultants who can't lift a finger without a ticket approved by multiple layers of management; their sites working at all is largely thanks to a lot of magic implemented by large CDN providers (Cloudflare, Fastly and so on).

Replies (1)

@redsakana@infosec.exchange 2026-08-24 10:08

As for apps, if you use an app now that's stable and efficient, there's fairly good likelihood of the maker being able to respond to memories shrinking. If you need to use some bloated BigCo shitpile that crashes all the time now, it's probably better not to think about it too much because it's only going to get a lot worse. You may have heard of stodgy BigCos being years behind the curve on hype trends, with a lot of big-budget blockchain projects still churning along in big banks and so on. Recently I keep hearing from BigCo acquintances of a massive incoming tsunami of absolute videcoding insanity, driven by a vicious circle of clueless boards and opportunist consultants: consultant proposes some completely unrealistic AI-productivity-enhancement to board, board enthusiastically pours down millions, thing turns out to be garbage, consultant tells board that you just didn't do this hard enough and proroposes something even more unrealistic plus getting rid of "recalcitrant" employees, and the cycle repeats. Even if OpenAI and Anthropic were to collapse tomorrow, that cycle is going to keep rolling for years.

Open ##5257864