← Feed

Posts

• View post

  Resource integrity for HTTP. Rather than having HTTP redirect to HTTPS with a content-free dummy page in the request body, the body of the 301 response is the same content being served over HTTPS, but the HTTP request has an additional Integrity header that allows: 1. verifying that the payload is not forged (signed by the same party that has the TLS certificate) 2. a subsequent, ETag-based conditional request to the HTTPS endpoint (Only for perambulatory browsing—not auth'd.)

• View post

  The plural of "mouse" is "mouses".

• View post

  Even when people don't set out to have LLMs write entire blog posts for them, I run across instances where a person has copied and pasted choice passages/snippets from LLMs (with all of their obvious tells) into something that the person has clearly otherwise written themselves, and I just think, "Why?" Even if you *do* want ChatGPT or whatever to straight up give you a blog post to post, it takes like, what, 20 minutes (less!) to redo it with a read-over and original-...

• View post

  A protocol (read: norm) whereby the value of the content attribute for a <meta name="generator"> element is a URL, and for any given page following this protocol, you can load the appropriate page designer/CMS interface/whatever for the page that you're on because it's a simple matter of appending a <script> element whose src attribute is the URL in step 1.

• View post

  Trying out something new.

• View post

  It feels to me like a big oversight for anything that implements its own flavor of followable profiles not to also support generic RSS and Atom feeds. (Note I'm not referring here to server software *generating* RSS/Atom feeds of its users' profiles—though that's important, too. I'm referring to the ability of any user to paste in an ordinary RSS/Atom feed and be able to "follow" it, too, alongside everything else in their timeline. For exampl...

• View post

  Shouldn't the focus on LLMs and the incorporation of them into workflows have by now started to look towards "lowering" (as in compilers/autocoding) the workflows that involve/depend on them into ones that instead use pseudo-natural language processing systems (like Inform7) to provide conversational interfaces, instead of wasting all that expensive compute on what often turn out to be rolls of the dice trying to get the LLM to do the right thing?

• View post

  Reminder in response to yet another billing incident involving insanely high surprise charges (<https://discuss.ai.google.dev/t/unexpected-54k-billing-spike-in-13-hours-firebase-browser-key-without-api-restrictions-used-for-gemini-requests/140262>): The Purpose of a Billing System Is What It Does See also: mrkurt commenting on why Fly.io also went with the standard playbook of having no limits on spending and refused to let you set a self-service hard cap so you wouldn't...

• View post

  A total conversion mod for Discourse to make it look and work like Discord (to further confuse the difference between the two).

• View post

  Friends don't let friends use the term "web engine".