Yeah, we are trying to build systems around this. It helps, but it's not always perfect. Of course, neither are people.

I feel the more structure and requirements the agent gets, the better the results

How we design when the code writes itself AI isn't just increasing the speed of building, it's changing how we work

90% of PRs by Claude Code @intercom

Cheap code impacts design:
- Agents own implementation
- Design owns edges, what to build & whether it's good
- Less polish, more systems & guardrails

Craft = judgment & taste at scale: users don't care if AI helped, only whether it feels right: buff.ly/dk38eYm

AI and the Next Economy The narrative from the AI labs is dazzling: build AGI, unlock astonishing productivity, and watch GDP surge. It’s a compelling story, especially if you’re the

When #AI makes people more productive, who benefits? Mostly big tech according to @timoreilly.bsky.social. We need to turn this into a revolution that benefits everyone, like the internet to avoid creating a loop of job cuts → less spending → less innovation.

→

Friction by Design - Carl Barenbrug What gets lost when everything is effortless? For years, good digital design has been framed as the removal of friction. Faster flows. Fewer steps. Instant …

#Friction by design: ilo.im/169u4y

Avoid unintentional friction, but use it to create meaningful pause for thought in UIs:
- before critical actions like a Delete confirmations
- for high stakes like payment to reduce errors
- to endless content with rate limits

Using AI as a Design Engineer How I use AI as a design engineer on a daily basis.

#AI speeds up trial-and-error, but it demands understanding of the outputs.

Never outsource judgment, deep understanding of the result wins as AI commoditizes code creation. AI makes code cheap, but without proper direction, this just means creating bugs and legacy issues faster.

Life Happens at 1x Speed Why I stopped consuming content at 2x speed, and the simple rule that changed how I listen to everything.

“If it’s not worth 1x, it’s not worth it” by @terriblesoftware.org buff.ly/hTrbn3r

I do also watch videos at 1.5x. Interesting argument: more content, less understanding. I guess it does depend on the content. Do you want to "scan" digital content or is it good content you want to consume?

The Intelligence Curse This series examines the incoming crisis of human irrelevance and provides a map towards a future where people remain the masters of their destiny.

AGI may soon make humans economically irrelevant, breaking the #socialContract. The "intelligence curse" explained: pyramid replacement → human irrelevance → power consolidation.

Can we break it by averting catastrophes, diffusing AI & democratizing?

AI has forever changed coding and you should embrace it, according to @antirezdotcom. LLMs can crush weeks of work in hours. Avoid the hype, the positive and the negative. Instead, test AI deeply to build faster and do more with your time.

→ ilo.im/169rs0

Humanizing AI Is a Trap LLMs humanize by design. Adding personality/emotion amplifies risk. Design real tools, not fake friends.

Humanizing AI is bad for UX:
- boosts chit-chat
- erodes trust
- cuts advice acceptance
- harms productivity

Treat #AI as a tool, not a friend.

Real cases in the 2025 studies include love affairs, suicides encouragement, and more 😱

→

More Efficiency, More Demand The future of software engineers and data scientists is bright

AI won't replace engineers, it will expand their work.

💡 Jevons Paradox: efficiency increases demand. Radiology grew despite AI because it enabled more scans & complexity. Same for coding: more software, more systems, more opportunities. Future is brighter, not scarcer.

☛

Great products do less, but better When feature bloat can hurt more than help your business goals.

Great products do fewer things, but extremely well. Set a feature budget & swap in higher‑impact features instead of endlessly adding new ones. Say no more often. When cutting features, offer better alternatives, migrations, or clear comms so short‑term pain doesn’t hurt the product:

How to Set Measurable Goals for Your Next Design Project Setting goals isn’t something foreign to most of us. Frameworks like OKR or SMART are common among teams of all types and sizes. But I…

Setting goals for your #design project: buff.ly/NLsfsaA

Set measurable design goals, not vague wishes to create shared expectations:
- Start from outcomes (behavior or business result), not outputs
- Pair each goal with a metric + target
- 1-3 primary metrics per project
- Use SMART goals / OKR

Immutable Infrastructure, Immutable Code - The Phoenix Architecture Why "Never Upgrade in Place" Now Applies to Software

This looks like a bright future: buff.ly/YF8njJH

If we can get our code to be built by #AI from spec like a container we could indeed work a lot better. Way too often have I experienced changes not being done, because of legacy code being in the way. Can this be changed?

Back to the starting line I loved this reflection by Andrej Karpathy about how he has never felt so far behind as a programmer. In particular, this line resonated strongly with me: “I have a sense that I could be 10X more…

#AI changes work for Junior & Senior Devs alike: buff.ly/ur7R03g

- natural language code changes are faster than manual ones
- debugging, picking APIs or writing tests is be done better & faster by AI

BUT AI can create bad code & debt very fast. Expertise is needed to steer AI & design solution

Stop asking users what they want — and start watching what they do. - Annotated People's opinions about themselves and the things they use rarely match real behaviour.

Old truth with some new perspectives: buff.ly/h4mSBN8 by @renderg.host

Issues:
- roadmap becomes reactive to requests
- implementation of scattered requests bloats UI
- shipping features that don’t move key metrics → confusing outputs with outcomes & missing opportunities for higher-impact

AI Can Write Your Code. It Can’t Do Your Job. The companies building AI are spending billions to acquire engineers, not replace them. Here’s why your job is safer than you think.

#AI is a coder, not an engineer: buff.ly/ArgCeDf by terriblesoftware.org

Your job is about judgement: clarifying fuzzy requirements, deciding trade-offs, managing technical debt, and making call-after-call about risk, scope & architecture.

Learn these “above-the-code” skills to stay valuable.

Prompt to Design Interfaces: Why Vague Prompts Fail and How to Fix Them Create better AI-prototyping designs by using precise visual keywords, references, analysis, as well as mock data and code snippets.

On #prompting #AI buff.ly/66MJ8kZ

Strategies to improve AI prototyping:

- Narrowly scope task
- Reference known styles or systems
- Attach visual references
- Provide realistic / mock data
- Attach code or specs

AI can execute & remix patterns well when given clear constraints, examples & data.

So your AI wants a personality Emerging personality patterns that drive differentiation in AI products

#AI personalities are an interesting aspect of AI product design: buff.ly/wN1pgQ9

Building blocks
- Purpose, worldview, values
- Archetype: mentor, librarian, coach, concierge
- Context adaptation: change tone & verbosity based on context
- Identity: Name, avatar, microcopy & interaction patterns

Estimates – a necessary evil? Product Owner: Hey, how long do you believe Feature F will take? Developer: Idk. We haven’t even started working on it and it’s bound to stir up some old issues.

Estimates: ift.tt/tAxhF49 necessary but misused.

See them as best guesses, not as solid prediction & promises. Instead of tying a release to a date, focus on the work and ship when you are ready. Focus on smaller batches & frequent delivery for improve handling of tech debt & dependencies.

Why Federated Design Systems Keep Failing Why do federated design systems keep failing? I watched it happen twice at Spotify. Here's what actually happened, where they broke down, and why centralised models usually work better.

Why you shouldn't build a federated #designSystyem: buff.ly/5ObCfnA by @shaunbent.co.uk

DS are already challenging. Don't make them harder than they need to be.

I also feel this: ... thinking tends to emanate from leadership, often abstracted away from details of running a design system.

Context is King Why your tech stack matters less than the context you give your AI

#Context beats #techStack for your #ai #agent: buff.ly/Kz4osFc

Don't change your stack to what is supposedly working best with AI. Any modern stack works with AI. Running into issues? Just provide a little bit more context, explain when to use what token, or provide access to component API docs.

This article buff.ly/CEhRZQW argues that the best way to deal with #LLMs is the same way as dealing with people:

- clarity
- explicit knowledge
- tight feedback loops
- shared standards
- vigilant error-catching

This means we have the tools to deal with #AI, but may need to adjust our expectations

Accessible by Design: The Role of the 'lang' Attribute - HTMHell A collection of bad practices in HTML, copied from real websites.

Today I learned of the importance of the lang attribute ilo.im/168yc1 thanks @matuzo.at

- tells browsers, screen readers & translation tools page language
- if missing: assistive tech guesses (system settings) → mispronounced, garbled speech, bad translations & incorrect typographic details

Double Click: What Does It Mean To Be A Designer In The Age Of AI? | Figma Blog As workflows shift what happens to job titles—and the ways of working and sense of identity that comes with them?

Insightful piece by @figma.com buff.ly/omcdFqZ on #designers in a world with #AI.

- curiosity, framing problems, & crafting experiences are still important
- Jobs are “bundles of tasks” requiring multiple skills
- more folks can “do design,” that doesn’t eliminate the need for deep expertise

Generating Utility Classes from Design Tokens using Style Dictionary Generate Utility Class CSS from design tokens using Style Dictionary for seamless design-to-development workflows and collaboration and snowflakes.

#AI brings utility #CSS back into the spot light, so here is how you can turn your #designTokens into utility classes: buff.ly/XD5onwu by @sturobson.com

Why? Utility classes are a middle ground between rigid components and ad‑hoc CSS: you can move fast in markup without hard-coding raw values.

The Math of Why You Can't Focus at Work Interruptions, recovery time, and task size: three numbers that determine if you'll get real work done. Interactive visualizations show the math behind bad days.

Read @can's buff.ly/FQloy9T It should be the only argument needed for remote work.

With notifications turned off I get a lot of focus time in my home office. Once I need a break I check slack, etc. Never got this in the office. Even people entering the room without talking get you out of focus.

Inclusive Design Principles Contributors: Henny Swan, Ian Pouncey, Heydon Pickering, Léonie Watson

#InclusiveDesign isn’t just #a11y compliance, it’s designing for real people in real situations: permanent, temporary, and situational disabilities.

Principles like giving control, choice, and prioritizing content make products better for everyone:

Public design systems are worth it It’s incredibly valuable to make a design system available to all–no matter what the bean-counters say.

I share @pjonori's concern about #designSystems going private: buff.ly/n6nP01A.

Public design systems are great:
- Forces higher standards for your system
- Shows applicants what the company does
- External scrutiny adds feedback, improving the system
- Offers inspiration to other creators.

Design systems don’t kill creativity – our biases do - zeroheight Donnie D'Amato looks at the cognitive biases that go into the way we frame design systems, and how to make sure they don't creep into our work

Agree with buff.ly/z9Nj12w @zeroheight.com / @donnie.damato.design

Design systems empower creativity by handling predictable, repetitive components & patterns, enabling teams to focus on meaningful, novel work.
Systems are enablers coexisting with continual innovation, not blockers.

3 practical ways LLMs can support design systems teams today - zeroheight

Managers love #AI, but it is often not ready and gets in the way of your day-to-day. @Zeroheight.com suggest some ways AI can be useful for #designSystem folks today buff.ly/AJJenPN

- turn docs into conversational UIs
- automate some boring maintenance tasks
- enhance tracking, feedback & insights