Sketches Through the Fog
Possible paths for where the judgment goes when AI handles the production.
I wrote on the blog this week about the difference between accelerating design work and automating it. We’re getting pretty good at the first. The second is still hazy. AI compresses how fast PRDs, flows, and prototypes appear, but each one still demands a lot of back-and-forth to push the output above mediocre. Real automation, the kind I sketched in the essay, needs specialist agent teams stitched together by a human who can shape what they produce. The destination is visible. The path is still hidden in fog.
Two of the pieces I linked this week extend the argument from opposite ends, and reading them together gives me a clearer picture of where the next move could go.
Tara Tan, an investor at Strange Ventures, audited more than a dozen AI design tools and landed on a finding that operationalizes my essay’s premise. As she puts it: “The competitive moat in this market is not generative quality, which is commoditizing fast. The moat is the design system graph.” Her example is Uber’s Ian Guisard, who didn’t stop being a design systems lead when uSpec automated his spec-writing. His job moved from producing documentation to encoding expertise into the system itself: writing the skills the agent runs on, defining the validation rules that decide what “correct” looks like across Uber’s seven implementation stacks. Guisard’s taste still matters. It lives in the system now.
Chad Johnson, writing in his newsletter, approaches the same shift from the discipline side. He watched a PM ship a v0 prototype that was “maybe 80% of the way there” and noticed the gap that mattered sat upstream of the visual polish: nobody had checked the assumptions baked into the user flow, and nobody had asked whether the feature was worth building at all. As Johnson puts it: “They’d built a beautiful answer to a question nobody had confirmed was worth asking.” His prescription is that designers become stewards. Not gatekeepers, not arbiters, but people responsible for the quality of thinking happening across the org, including the parts they’re not in the room for.
Tan suggests where the judgment could go. Johnson suggests when. Both prescriptions assume the same thing my essay assumes: the human work doesn’t disappear; it has to show up at a different altitude.
But “judgment” is too vague a word to leave there. Raj Nandan Sharma narrows it. In a world where competent first drafts are cheap, the scarce skill is refusal: knowing what to throw out, and why. Then he warns that refusal alone is a trap. As he puts it: “taste without authorship, stake, or construction can become a narrow and eventually fragile role.” Selecting from machine outputs has a ceiling. The judgment that holds up over time combines refusal with authorship: owning what gets built, and carrying the consequence when it’s wrong.
Pablo Stanley, who designs at Vercel, draws the line on the personal side after a weekend of making pixel art by hand: “The parts that feed my soul, I protected. The parts that would’ve killed the project with friction, I automated.” That’s the call every project now asks of you, and if you don’t make it on purpose, it gets made for you.
None of us has the path through the fog yet. I’m currently working on a preschool site as a side project, practicing the answer in miniature: visual language by hand, plumbing handed to the model. The judgment goes wherever I decide it has to live.
Acceleration Is Not Automation
I’ve been wandering the wilderness to understand where the software design profession is going. Via the blog and my newsletter, I’ve been exploring the possibilities by reading, commenting, and writing. Many other designers are in the same boat, with Erika Flowers’s Zero Vector design methodology being the most defined. Kudos to her for being one of the first—if not the first—to plant the flag.
Directionally Flowers is right. But for me, working in a team and on B2B software, it feels too simplistic and ignores the realities of working with customers and counterparts in product management and engineering. (That’s her whole point: one person to do it all, no handoff.)
The destination is within view. But it’s hazy and distant. The path to get there is unclear, like driving through soupy fog when your headlights reflecting off the mist are all you can see.
At its core, the UX design process remains unchanged, mirroring the scientific method—observe, question, hypothesize, experiment, test, analyze—and aligning closely with IDEO/Stanford d.school’s design thinking framework and the Design Council’s Double Diamond. Even unconsciously, designers follow this cycle through research, ideation, testing, and iteration, whether via hexagonal diagrams of empathy-driven stages or divergent/convergent diamonds.
So the question about where design is going is less about the overall process—because it stays the same, just compressed—and more about who is doing what with what. In other words, on a daily basis, what are designers doing and what tools are they using.
What I’m Consuming
Introducing Claude Design. Anthropic Labs shipped a research preview that turns prompts into prototypes, decks, mockups, and marketing collateral, powered by Opus 4.7 and available on Pro, Max, Team, and Enterprise. During onboarding, Claude reads your codebase and design files to build a design system, then applies it to every project after that. I haven’t tested it yet, but reactions are split: Figma’s stock dropped 5–7% on launch (Mike Krieger had quietly resigned from the Figma board three days earlier), The Register ran “who needs designers?”, and HN commenters worry about UI homogenization. Designer reactions feel more measured, focused on rapid prototyping wins. (Anthropic)
Sam Altman May Control Our Future—Can He Be Trusted? (Apple News+ subscriber link) Ronan Farrow and Andrew Marantz held more than a dozen conversations with Sam Altman and reviewed the seventy pages of internal Slack messages and HR documents that Ilya Sutskever sent to OpenAI’s board in 2023. One memo opened with the heading “Sam exhibits a consistent pattern of...” with the first item: “Lying.” The piece reconstructs the firing, the Blip, and the case (never publicly aired in full until now) that Altman is not the person Sutskever, Helen Toner, and Tasha McCauley believed should have his finger on the button. (Ronan Farrow and Andrew Marantz / The New Yorker)
I Mapped the Opus 4.7 Release to Your Role, Goals, and Real Workflows. Karo Zieminski’s launch-day breakdown of Anthropic’s Opus 4.7: the $5/$25 sticker price is unchanged, but a new tokenizer makes the same text cost up to 35% more on code and structured data, and three API parameters were quietly removed. Two pieces stand out for designers. Vision quality jumped from 1.15MP to 3.75MP, which means Claude can finally read a full Figma frame or dashboard screenshot instead of guessing. And the new literal-instruction default means your Skill files have to spell out tone and pattern with examples, instead of relying on hints. (Karo Zieminski / Product with Attitude)
Come at the king, you best not miss. Marcin Wichary traces my favorite Finder view—the column view—from NeXT to Mac OS X to the iPod and on to iPhone, where every Settings screen still uses it. Then he points at iOS Google Maps, where the design team replaced the standard Y/X relationship with a confusing Y/Z that just looks like Y. His principle: if you’re reinventing something well-established, the reasoning and the execution both have to be really, really solid. Apparently this didn’t happen here. (Marcin Wichary / Unsung)
Endgame for the Open Web. Anil Dash makes the case that 2026 is the year that decides whether the open web survives. The hectobillionaires running Big AI have started a final assault on every layer that made the open web possible: publishers hammered by AI scrapers without consent, robots.txt effectively dead, open APIs locked down, Wikipedia under siege, podcasts moving to closed platforms, open source projects flooded with slop submissions. His call is to stop carrying on with business as usual and fight like the threat is existential, because it is. (Anil Dash)
Prepping for the endgame of the open web. Jay, writing at The History of the Web, picks up where Dash leaves off and offers a sober second opinion. The open web has been attacked before, and survived: WordPress, Movable Type, RSS, the small web, and Wikipedia all came out of earlier collapses. The technology is resilient by design. The strategy doesn’t change: keep building openly, resist the technologies that aim to do harm, and find refuge in the smaller communities that have never gone away. (Jay / The History of the Web)