Anger is growing, and so is use
Instagram's head Adam Mosseri said on the Colin and Samir podcast that there is "more anxiety and more frustration and more anger about AI," and not only among young people and creatives. "It's not only getting worse, it's going to continue to get worse." I agree with his direction here.
https://www.theverge.com/ai-artificial-intelligence/1021644/adam-mosseri-instagram-ai-negativity
At the same time, the Reuters Institute's new survey shows weekly use of generative AI rising from 34% to 51% in a year, across the six countries with earlier data. Daily use is rising quickly too. Only 12% are comfortable with news made entirely by AI. The train is rolling and nothing will stop it.
And a small step for the angry ones: OpenAI, Google, Microsoft and Anthropic will back one shared signal in a website's robots.txt that says whether its content may be used for AI training. The idea came out of the IETF, the body that sets internet standards. That's a good idea.
https://www.theverge.com/tech/1023763/openai-google-microsoft-ai-scraping-standard-opt-out
Design tools learn to design
Figma explained how it trains its own model for precise edits on the canvas. Compared with its previous agent, users rate the new one 33% higher and undo its work 60% less often. Canva followed with Brush, its own model that looks at a design, understands its structure and edits it, from one line of copy to a whole deck, with everything staying editable. I don't think this will save Figma. They are still doomed.
https://www.figma.com/blog/training-a-figma-fluent-editing-model/
https://www.producthunt.com/products/canva
Nielsen Norman Group's Raluca Budiu wants to drop the "user" from "user experience." With AI agents acting on our behalf, the field now covers people, the agents working for them and the handoffs between the two, so she suggests "experience design." Ok, interesting idea and something worth thinking about. I am afraid the future will have less user and more bot, so we better start planning for that.
https://www.nngroup.com/articles/ux-vs-experience-design/
Karl Koch argues that AI products need a real undo, because a tidy rewrite can quietly change the meaning. He suggests showing the AI's version next to the original, with "Use this version" and "Keep original," and keeping the old version after you accept. Sounds valid to me, a lot of tools already do that too.
https://karlkoch.me/writing/on-giving-people-an-undo-button-in-ai-products
Martin Fowler wrote about "LLM voice": prose that seems machine-made makes him lose interest. He strongly recommends that authors don't let an LLM write a draft, "even if it's edited by a human," because writing is how he forms his thinking. Interesting opinion. But I am afraid the train is rolling.
https://martinfowler.com/bliki/LLM-Voice.html
Agents, still learning
Surge AI tested AI agents on 50 real tax jobs with 1,789 files, built and reviewed by tax experts. The best one, Claude Sonnet 5.5, passed 10.8% of them. Agents mostly failed at the issues clients never raised. That is surprisingly bad.
https://surgehq.ai/blog/dayjob-tax
Ars Technica asked Google's AI Mode who won the 2020 US presidential election. Instead of an answer, it pointed to the Federal Election Commission's official report. Same for 2024, and for 2012 it only offered a link. Every other election back to 2000, it answered directly. What's going on here?
https://arstechnica.com/ai/2026/10/google-ai-gets-cagey-when-asked-who-won-the-2020-elections/
Andrej Karpathy built HackerTube, a YouTube-style site with 6,454 videos recommended in Hacker News comments since 2007, with Claude Code in about two hours, without writing code himself. Isn't this a bit of a marketing stunt? It's a fancy cover for YouTube, which still hosts and delivers the videos.
https://karpathy.bearblog.dev/hacker-tube/
Also worth a look
Recipes, rewritten
NYT Cooking is testing an AI that suggests substitutions, adjusts equipment and scales recipes. In one example, it tells a cook to add baby spinach later than kale, because it wilts faster. Now I am afraid to try something out of a cookbook. Well, maybe we will end up being surprised and there will be AI fusion restaurants soon where we can be surprised by the newest food innovations.
The AI wingman
In a Match Group survey of 3,000 British singles, 39% said they had used AI in their dating lives, the Guardian reports: pep talks, checking messages, judging potential partners. Wild use of AI. If both people on a first-date chat use it, who is actually talking to whom? Somehow reminds me of the hiring dilemma.
https://www.theguardian.com/lifeandstyle/2026/oct/31/ai-your-wingman-chatbots-transforming-dating
Where are you on the train: on board, or standing on the tracks? Leave a comment here or reply on Bluesky.