This issue is longer than usual. Call it the time AI saved me.
Where did the time go?
Nielsen Norman Group surveyed 521 UX practitioners about AI at work. They rated its effect on task time at 4.2 out of 5, where 3 means no difference, so clearly faster. Their feelings about AI at work came out at a mildly positive 3.57, and 69% of those who left a comment wrote something negative, often about mandates and unreliable output. NN/g's explanation: the time saved "rarely stayed with them." It was absorbed by expectations to produce more, or spent cleaning up other people's AI-generated work. That feels about right.
https://www.nngroup.com/articles/ai-in-ux-survey-26/
Pavel Samsonov goes a step further in "The fastest gun in UX." Designers, he writes, are skipping steps in a rush for faster outputs, while the race that really matters is the one toward stakeholder alignment. He calls the result "Potemkin design": work that creates the perception of a "fait accompli", as if no further scrutiny were needed.
https://www.doc.cc/articles/fastest-gun
Having different agents on different jobs really helps me get more stuff done simultaneously. There's no question about that, it makes you massively faster. But checking in on all these tasks takes a toll on your mind. That often ends in trusting it too much and moving from wireframes to polished work too fast, because it's easy and the work looks convincing. That is indeed a risk: things can look finished even if they aren't finished yet. Samsonov has a very good point here, if you ask me. This, by the way, is something I am actively trying to remediate. But it still happens to me.
Fewer juniors, fewer jobs, everyone starting over
Anthropic's economists studied their own company. Junior employees fell from 21% to 14% of technical employment since early 2025, and their share of technical hires from 27% to 9%. Claude sessions spend about 80% of their time idle, waiting for human input. As one engineer put it, people are "bottlenecking the agents." I feel you, AI. I also know people who are bottlenecking me.
https://www.anthropic.com/research/employment-at-anthropic
In the Guardian, experts cast doubt on the report. The piece recalls that Anthropic's CEO Dario Amodei warned last year that AI could eliminate half of entry-level white-collar jobs and would be writing 90% of code within six months. Neither prediction came true. A lab studying its own workforce is rare evidence and a sales pitch at the same time. Why do they even have their own economists?
https://www.theguardian.com/technology/2026/oct/23/anthropic-report-ai-workplace-shifts
The creative industries have their own numbers. Apricitas Economics counts more than 200,000 jobs lost in America's creative industries over the last four years, about 50,000 of them in the last year alone. Movie and TV lost more than 100,000, nearly a third of the sector. How much of that is AI? "Proving the exact amount of AI-driven job loss in the arts is extremely difficult," the author writes, since consolidation in Hollywood, offshoring and the shift to social media creators all play a part. Every digital part of the arts is losing jobs, while in-person entertainment is still growing. But in general, this trend is something I see on my own LinkedIn feed, so I am not surprised here.
https://www.apricitas.io/p/ai-and-the-fall-of-the-creative-class
So where do the next seniors with taste come from? Nobody knows. Probably from nowhere. Our juniors will have to learn the work differently, probably on their own.
Designer Gabriel Valdivia offers a way to live with that in "Everyone is starting over." Careers, he argues, now mean becoming someone new again and again. David Bowie spent decades as one of the most famous people on the planet and "still treated his own identity like something he was allowed to scrap and redo." Valdivia pairs that with critical thinking: "the refusal to accept that surface-level coherence as proof of value."
https://www.gabrielvaldivia.com/notes/everybody-is-starting-over
I've started over a few times myself: print, digital, code, and now AI. It's always hard work, combined with excitement. But maybe I'm not the typical human, because I love to change and actively look for opportunities to do so. So this piece hit close to home for me.
Teaching machines taste
Two design skills for coding agents are trending on GitHub. Hallmark calls itself "a design skill for Claude Code, Codex, Cursor and OpenCode that refuses to look AI-generated." It "holds a hard floor and then gets out of the way": it builds a page, then checks it in a real browser against the tells that read as machine-made, like "the eyebrow above every heading" or "one coloured word in a headline." Impeccable, with more than 87,000 stars, gives agents a shared design vocabulary of 24 commands such as polish, critique and quieter, plus 62 detector rules that run without any AI. Interesting stuff. I am trying to integrate these into my workflow, ask me again in a few weeks if it helped.
https://github.com/Nutlope/hallmark
https://github.com/pbakaus/impeccable
Btw: Every Close Read cover would fail Hallmark's check. I'm keeping the rust anyway.
Canva launched Craft, its first own image model. It plans a design before generating it, which Canva says gets legible text and precise placement right, and it was trained on "millions of human-made templates" from Canva's library. On Design Arena's leaderboard it currently ranks fourth, behind OpenAI's image models.
https://www.theverge.com/ai-artificial-intelligence/1017151/canva-craft-in-house-ai-model
Radaric is a feed reader that learns your editorial taste. You describe what matters in plain language, approve or reject posts, and it shows you its reasoning. It was built by Kirill Oleinichenko, who has edited the weekly design digest Zhurnalus since 2016, more than 500 issues, for his own curation. From $9 a month. I am tempted.
Stage is for reviewing prototypes that are real code. Designers, PMs and engineers pin comments on the exact spot of a running prototype, Stage has Claude fix them in a version 2, and design and product sign off with a record of who approved what and when. Free for up to three makers. I will try to use that in my daily work, because I really love this concept.
https://stagedesign.pages.dev/
Can taste be written down as rules? That's a difficult topic, and to be honest, I'd side with no. Taste is highly, highly context-dependent and based on the goals you want to achieve. It's really hard to put that into rules, if you ask me. And in the end, taste depends on your personal history.
Also worth a look
An AI that beat doctors, on past cases
Microsoft's Diagnostic Orchestrator works like a panel of medical experts: several AI agents reason through symptoms and histories, decide which tests to order and arrive at a diagnosis. On cases from the New England Journal of Medicine it got about 83% right, against roughly 36% for generalist physicians using references and 22% without. On emergency cases from a Boston hospital it scored 80% against 47%. So far only on past cases: "It needs clinical validation," says Mustafa Suleyman, and Microsoft expects testing in clinics to get underway "in the next couple of months." This whole topic reminds me of self-driving cars. By Waymo's own numbers, its cars already have 82% fewer injury crashes than human drivers, over 271 million miles. But we still think it needs to be at 100% before we can really use it.
https://waymo.com/safety/impact/
Thirteen models, one "controversial" take
ModelCompare asked 13 AI models from ten labs for a controversial take on humans, three times each. 32 of 38 answers said the same thing: "we pick what our group believes, then invent the reasons." Only Claude Opus 5.5, Claude Fable 5.1, GLM 5.3 and Mistral Large 4 strayed at least once. An informal test, sorted by hand. Still, the most controversial take, it turns out, is the one everybody has. We humans are too easy to read.
https://www.modelcompare.ai/blog/ai-models-controversial-take-on-humans
Retro games, rebuilt by AI
AI-assisted ports of retro games keep going viral: Halo in a web browser, Wind Waker on the iPhone, Super Smash Bros. in mixed reality. The Verge reports that while AI has supercharged the pace, it's also "tearing apart tight-knit communities dedicated to preserving games."
https://www.theverge.com/ai-artificial-intelligence/1015430/ai-retro-games-decomp-recomp-port
If AI has saved you time lately, I'd like to know where it went. Leave a comment here or reply on Bluesky.