Tag: AI

• Bubbles (October 14, 2026)

  I’m not saying they’re right or wrong, because I don’t know anything that’s going to happen, but the unshakeable confidence of people looking ahead to the AI bubble popping any day now, brining suffering and woe to all their enemies and righteous joy to themselves, reminds of millenarians who are ever-sure that Christ’s return is imminent and won’t they be sorry. Or, perhaps, the rumors that AI researchers are taking up smoking en masse since they figure that their work will either cure all cancers or kill all humans well before they suffer any ill effects.

• Bicameral Google, still (September 29, 2026)

  In When did Google get so f-ing weird?, Sancho Panza experiences the very same phenomenon I wrote about last December, and called “Bicameral Google”. This being that, for many Google search queries, you still do indeed get the results you wanted—but only after Google’s AI butts in and interprets your search query as if you had addressed it to a chat bot, which you did not. (Except, clearly, you did.)

• Reinforcement porning from human feedback (RPHF) (September 28, 2026)

  Publicly accessible LLM-based chatbots receive a ceaseless stream of requests from around the world for custom pornography, which in itself shouldn’t surprise anyone. But Joseph Cox at 404 Media has interviewed contractors who are now tasked by companies like Microsoft to judge whether the bots are generating lewd or suggestive images of acceptable accuracy and quality:

• Australia has no time for this (September 25, 2026)

  Australia discovered than an OpenAI agent gained unauthorized access to a government-run health-statistics website earlier this year. Instead of responding with fear and awe directed at the agent itself, as we’ve seen so much of lately, the government has instead launched an investigation aimed at the agent’s corporate owner. I appreciate this.

• Best practices for LLMs and open-source contributions (June 26, 2026)

  The Software Freedom Conservancy has published Recommendations

• Revisiting Toombs (June 15, 2026)

  During some recent tinkering I revisited this Fogknife post from 2021, which I had illustrated with the following XKCD comic :

• Bicameral Google is unpleasant (December 16, 2025)

  Google's 'AI Overview' might be an OK chatbot, but it's always an uninvited busybody in my search results.

• Could I get a hat wobble? (November 29, 2025)

  And a Flarhgunnstow?

• Down in the mörk (November 28, 2025)

  I spent last weekend at PAX Unplugged in Philadelphia. It had been a long time since I’d last attended a tabletop gaming conference of any kind with my game-loving partner, and I pushed to have us attend this, both because-of and despite how it slotted perfectly between two other interstate trips that filled out most of November for us. We traded away a chance to catch our breath at home for several days of exhausting overstimulation that reminded me why I love non-digital games at least as much as video games, and caught me up on recent trends and innovations I’d been missing. It was a terrible idea, and I regret nothing.

• My commonplace safe space (July 30, 2025)

  Five months after my last post announcing the revival of my notes blog under a new title, I re-revive it with a new goal: a commonplace book, kept by and for me, but readable by the public. This is what can differentiate notes here from posts on Fogknife. To this end I’ve added a tags page, and will start tagging my ’Twas This posts as of right now.