# Notes by Sandor Gyuris

Source: https://sandor.works/notes/what-are-we-willing-to-delegate-to-ai

What are we willing to delegate to AI?

As AI models become more capable and products make those capabilities easier to use, more of our everyday work becomes possible to hand over. That leaves us with a question about our own preferences: which tasks do we want to do, which do we want help with, and which would we happily delegate completely?
We already make these arrangements with people. Your partner might book a restaurant for your birthday because they know what you like. A colleague might take care of something before you ask because, over time, you’ve come to trust their judgment. We give them room to act and expect them to recognize when to bring us back in.
With memory, context, and access to our tools, AI could take on more of these responsibilities. Our willingness to delegate will depend on what the work means to us.
Different parts of the same task
We might happily hand over expense admin: matching receipts, filling in details, and submitting claims within company policy.
For a birthday dinner, we might want help finding a restaurant, delegate the booking and reminders, but choose the guests and write the birthday message ourselves.
One task can involve help, full delegation, and room to do things ourselves. Automating all of it can remove the parts we enjoy.
Effort doesn’t tell us what to hand over
We might gladly spend an hour writing a personal message while resenting ten minutes of expense admin. What matters is how much we value taking part, not just how long the work takes.
That changes from day to day. Cooking might help us unwind one evening and feel like a chore the next. Knowing our preferences shouldn’t mean assuming how involved we want to be.
Choosing how to work with AI
A useful AI product should let us choose our involvement at the level of each subtask. It needs to know when to offer help, when to act, and when to leave something with us. Completing the whole task is only useful if that’s what we wanted.
Taking part has value, even when AI could get us to the same result faster. We might want to work through an idea, make a personal choice, or simply enjoy doing something ourselves.
When you partner with AI, ask yourself: which parts do I want off my plate, and which would I miss if it did them for me?

---

Source: https://sandor.works/notes/ambition-is-a-design-skill

Ambition is a design skill

We talk a lot about the craft of design: systems, process, research, collaboration and execution. We spend less time discussing the scale of our ambition, even though it often determines the ceiling of the work.
Ask how to improve a flow and we will probably improve it. Ask what the product should become and the conversation opens up to new behaviors, systems and ways for people to get things done.
Ideas need room to develop before we reduce them to what feels achievable today. As making gets faster, our judgment about what deserves attention becomes even more valuable.
Ambition lets us imagine a better outcome. Agency gives us the drive to move towards it. Together, they can make almost any design problem more workable: we can question the frame, make a first version, learn from it and find another route when we meet a constraint.
We learn to think within the existing product
Working inside an organization teaches us to be realistic. We learn the roadmap, technical constraints, business priorities and which ideas are likely to gain support.
Eventually, we get so good at anticipating those limits that we start negotiating against ourselves. The idea gets smaller before anyone challenges it. The current product becomes the frame for every future improvement.
Ambition means asking what should exist before deciding what we can deliver. Reality should shape the route towards an idea, not set its size from the start.
Think one level above the request
A request for a new interface might reveal that the product is built around the wrong model. A usability issue might point to a missing capability. A chatbot might be an early step towards an experience woven throughout the product.
The most useful question is often one level above the one we were given:
What are we actually trying to change?
Which assumptions belong to an older version of the product?
What would the ideal experience feel like?
What could we make possible that people would not think to request?
If this works, what could the product become next?
These questions help a team see the larger opportunity and decide where its attention can make the greatest difference. That is one of the most valuable things design can provide.
Ambition needs judgment
A big idea still needs to be a good one. That is where judgment and a strong eye matter.
You develop your eye through sustained attention: studying products, technology, art, culture and human behavior; making things; comparing alternatives; and noticing why one direction feels stronger than another. What we choose to study and pursue shapes the ideas we can bring into the world.
Over time, we get better at recognizing whether an idea is genuinely new or simply unfamiliar. We see when complexity supports an experience and when it hides a weak concept. We can sense when a direction has its own logic before every detail is resolved.
Ambition expands the possibilities. Judgment helps us choose where to go.
That is why I want to keep developing my technical fluency while studying art direction and training my eye. Technical knowledge gives me access to more of the product. Judgment helps me use that access well.
Agentic products need ambitious designers
The future form of agentic products is still unresolved. Almost all begin with a chatbot. It is a useful starting point, but it does not tell us what these products should become.
How should an agent behave when it understands someone’s intent and context? When should it act, ask or stay out of the way? How should it adapt to different levels of experience? What would make it dependable enough to become part of someone’s work?
There are few established answers. We already shape the interface, behavior, system and direction behind these products. As the space develops, we will need to make unfamiliar possibilities clear enough for others to see, challenge and help build.
That takes the confidence to propose a direction before every detail is known.
Direction is part of the work
A compelling vision helps people make decisions. It explains where we are going, why it matters and how today’s work contributes. It gives product and engineering something tangible to respond to, and helps smaller decisions fit together.
Ambition becomes practical when we set a destination, then test and sharpen it as we find a credible route.
One of the most exciting shifts is that describing an experience can begin to make it real. English is not my first language, but I can describe a feeling: how an interaction should move, the confidence it should create or the control it should give someone. That description can become a working prototype. Articulating an experience is becoming part of making it.
We do not need the whole route before we begin. We need enough clarity to move together and enough flexibility to learn.
We have fewer excuses to think small
It is easier than ever to make an idea tangible, test a direction and show what a better experience could be. We can question the frame, explore what else the product could become and invite others into that possibility, especially when the brief points towards a smaller idea.
As making gets faster, where we direct our attention becomes a bigger differentiator. Which customer problem deserves our focus? Which part of the experience is worth rethinking? Which promising idea deserves time and collaboration? Distinctive products come from choosing those questions carefully and pursuing them with conviction.
This is a skill we can practice: look above the request, make the stronger experience visible, explain why it matters and bring others into the work. Ambition sets a higher bar; agency helps us start moving towards it. Constraints will shape what we can deliver, but they do not have to decide how far we are willing to explore.
Ambition is collective
The strongest ideas rarely arrive fully formed from one person. They develop through critique, experimentation and different perspectives meeting.
One person may see the product opportunity. Another understands what the technology enables. Someone else finds the interaction that makes the idea clear. A team can reach a more ambitious direction by giving those perspectives room to meet and keeping attention on the experience they want to create.
Share work before it feels finished. Critique the direction as carefully as the details. Help one another develop stronger judgment. Collective judgment lets a team act on an ambitious idea.
Are we thinking big enough?
Design is shaping more of the product: its interface, behavior, system and direction. People will increasingly look to us for a view on what should exist and where the product should go.
That responsibility gives us the chance to reconsider what our products can become.
Our tools will change and our roles will expand. We have fewer excuses for leaving a better experience unexplored. Keep asking what deserves our attention, make a stronger possibility visible and bring others along until it can become real. Ambition helps us see further; agency helps us begin.

---

Source: https://sandor.works/notes/the-software-atelier

The software atelier

Before industrial production, a coat was expensive partly because making it required so much human effort. Wool had to be sheared, cleaned, spun, woven, cut, and sewn. A well-made coat was valuable because producing one took time and specialist skill.
Mechanization changed the economics. Mills, standardized patterns, and industrial cutting made durable clothing cheaper and more widely available. As basic quality became easier to achieve, more of the premium moved into the cut, material, silhouette, and the designer’s ability to turn those choices into something people wanted to wear.
From reliable process to distinctive experience
The restaurant industry makes the same shift easy to see. A fast-food restaurant coordinates ingredients, equipment, labor, timing, and distribution so that a familiar meal can be produced quickly and consistently. Its value comes from delivering the expected result at scale.
A great restaurant depends on the same operational discipline. Food must be safe, ingredients must arrive on time, and every dish must reach the table at the right temperature. A Michelin-starred kitchen is usually more disciplined about these fundamentals, not less.
The chef adds another layer of value by deciding which ingredients deserve attention, what belongs on the menu, how each dish should be balanced, and how the meal should unfold. Food, service, pacing, and atmosphere work together as one experience. Function remains essential, but it accounts for only part of what people value.
Software is reaching the same point
For decades, software was valuable partly because producing it was difficult. Turning an idea into a dependable product required engineers, designers, and product managers working together for months. A tool that worked reliably had already achieved something expensive.
AI is reducing the effort required to reach that baseline. An agent can generate an authentication flow, database, dashboard, or settings page in a fraction of the time these features once took. Complex products remain difficult to build well, but functional software is becoming faster and cheaper to produce.
Reliability, security, accessibility, and performance still matter, although customers increasingly expect them from every serious product. More of the challenge now sits in choosing which problems deserve attention, which users to serve, and which version of an idea is worth developing.
Abundance needs a stronger editor
When production was expensive, cost acted as a filter. AI weakens that filter because teams can add another feature, create another variation, or respond to another request with much less effort. This creates more room for experimentation and makes unfocused products easier to produce.
Features can compete for attention, introduce conflicting behaviors, and make a product harder to understand. As production gets faster, quality depends more heavily on the team’s ability to decide what belongs.
This is a useful way to think about the software atelier: a team organized around clear product judgment. AI speeds up execution, while people remain responsible for direction, quality, and coherence. They decide where automation helps, where human review still matters, when a strong default is better than another setting, and what the product should leave out.
Design becomes more editorial
On a team like this, design begins before requirements are settled. Designers work with product and engineering to shape who the product serves, what outcome matters, and what the simplest coherent version could be.
The role becomes more editorial, with greater emphasis on selection, sequencing, clarification, and removal. These decisions still need to be grounded in customer needs, business context, and a strong understanding of the medium.
Output also reveals less about a team’s quality when prototypes and production code can be generated quickly. Stronger signals include the clarity of the product, the reasoning behind its decisions, and the team’s ability to learn without filling the roadmap with every possible response.
Restraint has business value
Great restaurants do not offer every possible dish, and fashion houses do not produce every possible garment. Software teams often treat breadth as an obvious advantage, even though every capability adds explanation, maintenance, and support.
A smaller set of well-integrated features can be easier to adopt and trust than a long list of loosely connected ones. The value of restraint becomes clearer when it is connected to outcomes such as adoption, trust, retention, and operating cost.
The foundations still matter
A respected restaurant cannot survive an unsafe kitchen, and a beautiful coat has little value if its seams come apart. Software follows the same rule. Reliability, accessibility, performance, privacy, and security remain fundamental.
AI changes where teams spend their effort. Less time can go into producing familiar patterns, while more attention goes into understanding the problem, testing alternatives, and improving the whole experience.
Industrial production made good clothing widely available, and standardized kitchens did the same for food. AI is increasing the productive capacity of software teams in a similar way. The advantage will come from using that capacity to make clearer choices, build more coherent products, and apply human judgment where it matters most.

---

Source: https://sandor.works/notes/you-cannot-one-shot-design

You cannot one-shot design

Every new AI model arrives with another round of comparisons. Can it produce a polished interface in one prompt? Does it need a second pass? How quickly can it turn a rough idea into something that looks ready to ship?
The progress is genuinely impressive, yet the conversation keeps collapsing into the same tired conclusion that design is dead.
That conclusion depends on a narrow understanding of design. If design simply means producing polished screens, then the role is clearly under pressure. The practice reaches much further because design shapes how things work, how people experience them and what happens when they interact with them.
A model can one-shot an interface. It cannot one-shot the decisions that make it right.

A polished screen can still be unresolved
We tend to judge these models through what we can immediately see, including typography, colour, spacing, components and animation. Those details matter, although they reveal only the visible layer of a much larger set of choices.
The work still has to answer questions such as:
Who should this experience serve?
What deserves the most attention?
How should the experience behave?
Which compromises are acceptable?
What can be removed without losing meaning?
A generated interface can appear complete while leaving every one of these questions unanswered. It may resemble a good product without reflecting the context and judgment that would make it the right product.
Speed makes judgment more valuable
AI can produce more directions in a few minutes than a team could previously explore in days. This gives us greater freedom to experiment, compare ideas and move beyond the first plausible solution.
An abundance of options also gives us more to evaluate. Someone still has to recognise which ideas deserve attention, understand their consequences and decide how far to push the work. Selection, editing and refinement become more valuable as generation becomes easier.
Design has always developed through making, observing and refining. AI accelerates that process and provides more material to work with, while the quality of the result continues to depend on the decisions made along the way.
We may finally see design more clearly
Many people still associate design with its visible outputs because screens and prototypes provide the clearest evidence of the work. As machines become increasingly capable of producing those outputs, the deeper responsibilities will become easier to recognise.
Setting direction, understanding behaviour, navigating trade-offs and refining ideas until an experience feels coherent will remain essential. AI will transform how design is produced, and it may also help more people understand what design has always been.
You cannot one-shot design because every output still needs to be judged, refined and connected to the whole.

---

Source: https://sandor.works/notes/taste-needs-friction

Taste needs friction

I saw someone recently on a Reel say that he missed Blockbuster because it made him finish bad movies.
Once you had chosen a film, paid for it and taken it home, you usually watched it to the end. If you didn’t like it, you had to sit with that feeling. You began to notice where the story lost you, which scenes dragged and why the ending didn’t work.
That kind of attention is how you refine your judgment: you learn to recognise what works, understand why it works, and decide what you would do differently.
Taste doesn’t just develop by consuming only things you already like. You also need work that bores, confuses or challenges you. A bad film can teach you about pacing. An awkward interface can reveal the cost of making people think too much. A weak argument can show you where logic gives way to persuasion.
What matters is understanding your reaction instead of escaping it. Why didn’t this work? Where did I lose interest? What would I have done differently?
Today, escaping is easy. Recommendation systems learn what holds our attention and give us more of it. Yet research on algorithmic curation suggests that a system can accurately predict what we’ll choose next while narrowing what we might learn to enjoy later.
AI takes this further. It can answer a question immediately, solve a problem on demand, or generate an interface around a specific need. This is often useful, but it also means we spend less time struggling with questions, adapting to constraints, and forming judgments of our own.
If you want to make good work, you have to remain open to things that don’t immediately satisfy you. You have to handle the friction long enough to learn from it.
Blockbuster made that friction unavoidable. Now we have to choose it for ourselves.

---

Source: https://sandor.works/notes/design-is-chasing-a-feeling

Design is chasing a feeling

For too long, we’ve treated efficiency as the ultimate goal of design.
Form follows function became a rule instead of a reminder. But when every layer of an experience is stripped down to its bare minimum to be the most efficient, something vital gets lost. Form and function shouldn’t compete. They should complete each other.
Design has always been an emotional pursuit. The question is never just “what does it do?” but “what should it feel like to use?”
That chase between logic and intuition, between function and feeling, is where design comes alive.
“One of the reasons the tech industry is so creatively bankrupt is that people just focus on attributes you can measure.
I was very lucky at Apple, because Steve showed me there was a way to articulate things like sensibility, intuition, taste, which all start with a feeling. But first I had to make friends with this frustrating part of myself, which was the inability to articulate what I feel.”
— Jony Ive
To design well, you have to make peace with the frustrating part of the process: the moments you can’t yet explain why something feels off, only that it does.
1. Chasing emotion
The best designs make people feel something. Think of the calm clarity of Apple Notes, the playful warmth of Duolingo, or the quiet trust of a banking app that feels reliable and honest.
Good design turns function into emotion. People might forget the interface, but they remember how it made them feel.
2. Chasing intuition
Designers often make decisions that cannot be proven with data. A layout feels heavy. A motion feels abrupt. A color feels too loud. You keep adjusting until it feels right.
That is intuition at work, a kind of muscle memory built through experience. It is what turns something that works into something that feels alive.
3. Chasing empathy
Design is not just about your own feeling. It is about understanding someone else’s. You think about the traveler rebooking a flight at midnight, or the parent using a baby monitor app in the dark.
Empathy means chasing their feeling, noticing their frustration or relief, and shaping the experience to meet them there.
4. Chasing flow
When designers are in flow, the work feels effortless. Ideas connect easily, details fall into place, and everything moves with purpose.
You can feel that same flow in products like Figma or Notion, where the interface stays out of the way and lets your thoughts move freely. Flow is not just how it feels to design. It is how it feels to use.
5. Chasing harmony
Harmony is when every detail, from typography to motion to tone, supports the same emotional note. Spotify’s dark gradients feel immersive. Airbnb’s rounded shapes and warm palette feel human and safe.
That balance cannot be automated. It is what makes an experience feel cohesive rather than just consistent.
6. Chasing hope
It’s the belief that every problem has a path through it, and that each iteration brings us closer to a more generous outcome. You can feel that hope in products that make complexity feel manageable, where thoughtful details reassure you that you’re moving in the right direction. Hope becomes the invisible thread connecting the work to its purpose. Design is pushing what’s possible.
7. Chasing meaning
In the end, design is about connection. It is why we are drawn to products that reflect who we are, like a minimalist note-taking app, a joyful camera interface, or a tool that just feels right.
Meaning gives emotion depth. It is when design stops just working and starts resonating.
Design is chasing a feeling in what we make, how we make it, and how we care for it over time.
The real craft of design is not just building things that function. It is nurturing things that feel alive, human, and cared for. Metrics can tell you what people do, but only feeling can tell you why it matters.

---

Source: https://sandor.works/notes/beauty-the-new-business-tool

Beauty, the new business tool

Recently I came across a 1927 article in The Atlantic about how the industrial revolution let people churn out functional products fast, including cars, toasters, lamps, and furniture, but with little care for how they felt to use.
After years of mass production, people started craving beauty again. Once function was good enough, aesthetics and care became the next big differentiator. People no longer wanted just what worked; they wanted what felt good to use and to own. Beauty became a huge business advantage.
It feels like we’re entering a similar moment with AI today. Everyone can now create 7/10 experiences quickly: things that are functional, usable, and even clever. But the real opportunity lies beyond that.
Design feels really undervalued right now. The role will certainly evolve, but designing deeply human, high-quality experiences will soon be the true differentiator: the modern equivalent of craftsmanship after the industrial revolution.
Pretty soon, people will long for products that are intentional and beautiful, the kind that feel genuinely considered and human. We’re about to see a new wave of design where craft, emotion, and taste matter more than ever.

---

Source: https://sandor.works/notes/taste-in-the-world-of-ai

Taste in the world of AI

In product design, there’s always been a quiet force shaping everything from the curve of an icon to the tone of a push notification. We don’t talk about it much. We don’t A/B test it. But we feel it. It’s called taste.
Now that AI is everywhere in the design process, from writing copy and generating flows to suggesting layouts, taste is more important than ever. It’s what keeps the work human. It’s what separates products that feel assembled from ones that feel intentional.
AI can now generate infinite output. But not everything that works should ship. Not everything that performs should live. Taste tells you what’s worth keeping. It helps you see what matters before anyone else can put words to it.
That’s why, in the age of AI, the value designers bring is shifting. Ivan, founder of Notion, puts it simply: what remains is a mix of capability, agency, and taste.
Capability is your skillset. Designing interfaces, writing copy, building prototypes. AI is getting good at helping here.
Agency is your drive. The ability to push through ambiguity, make decisions, and ship work.
Taste is what determines which direction you push. It’s your compass. It tells you where to go, what feels right, and what to leave behind. Without it, the work may be functional, but it won’t feel alive.
Taste is the quiet instinct that says this is off when a margin feels wrong. It’s the part of you that changes a word because the tone doesn’t quite fit or removes a step in a flow that technically works but just feels tired.
In AI-first design, your job isn’t just to prompt and pick. It’s to guide. To shape. To curate. The best experiences aren’t just fast or functional. They’re coherent. Expressive. Alive. That coherence doesn’t come from the model; it comes from you.
Taste becomes your filtering system. It’s how you decide when to intervene and when to let the machine run. It’s an internal compass for quality. It tells you when to keep pushing and when to say this is done. It lets you spot inconsistency before it becomes friction. And in a world where anyone can ship something that looks decent, the difference is how it feels.
Because software makes you feel a certain way when it’s working well. And even if the customer doesn’t notice each detail, they can feel their collective presence.
But taste isn’t just about refinement. It’s about conviction. About being willing to take risks, to try something that hasn’t been done before. Look at Apple’s new iOS 26 glass aesthetic. It’s not just visual polish. It’s a statement that the platform can evolve. That the interface can feel fluid, alive, and grounded in something new. No prompt would have surfaced it. No metric would have demanded it. It came from a team with a point of view.
Taste is what gives a product personality. And often, that personality is what makes people care. Taste is also holistic. It’s not just how something looks, but how it feels, sounds, moves, and behaves. It shows up in micro-details like spacing and motion, but also in macro decisions like whether a feature should even exist. Visual craft, experience flow, tone of voice, structure, and performance all carry taste.
It’s not innate. Taste is cultivated. You develop it by studying great work. By making things, over and over again. By building with people who set a high bar. By holding yourself to that same bar and refusing to compromise.
Taste compounds. The more you care, the more your instincts sharpen. The more you practice, the more naturally it shows up in the work. But it only sticks if leadership shows it matters. Taste starts at the top. Not by hiring great designers, but by modeling great decisions. By sweating the details in reviews. By refusing to let speed or scale water down the craft. If quality isn’t protected, it disappears fast and it’s hard to get back.
AI is changing how we design. Faster prototyping, faster iteration, faster production. But speed without sensitivity leads to sameness. A thousand outputs, all technically correct, none emotionally resonant. The best designers will use AI to amplify their taste, not replace it. To explore more deeply. To polish faster. To obsess over details with more time, not less.
In the age of AI-driven design, taste isn’t just polish. It’s strategy. It’s how you make something worth remembering. Worth choosing. Worth loving.
Because when everything is possible, taste is what makes it personal. It’s what turns raw potential into real products. Into work that resonates. Into software that feels not just made but meant.

---