### Why now?

Galileo AI just launched their public beta and raised their seed round led by Khosla Ventures, it's an interesting tool that I think lots of people could leverage to communicate ideas and learn a bit about product design.A few notable features:

1. **Text to UI:** the standard format we've come to know through other LLM interfaces, I think this un-opinionated interface is good for people who generally have an inkling of an idea but bad for getting down to specifics as a co-pilot for a designer or PM. It's nice that it gives you multiple outputs to riff on, from what I can tell it tries to generally keep them distinct enough that it's an easy choice which one to proceed with.

2. **Image to UI**: Use images or wireframes to generate hi-fi designs. I think this could be neat for exploring different design patterns. as simple as uploading a post-it note and adding some context, it seems.

## Testing the tool

My first prompt with Galileo was focused on seeing how far I could stretch the tool, and how far it would propagate things like style requests. Its output was three fairly distinct designs. Each one has glaring faults/fallacies immediately obvious to me but could be good starting points for working off of. 

```cv
gallery
{"layout":"hscroll","images":[{"src":"https://firebasestorage.googleapis.com/v0/b/maitake-project.appspot.com/o/pages%2FdZYy1hUn8Ud6iIoIHOvn2T6UGXI3%2F1kUhVRwrCTr7ONfS1Rsy%2F6f003093-ade9-4b4c-85fd-6b7a91bc9eaf.png?alt=media&token=12420d50-99f8-4a95-b7c3-31bc537e6e14","id":"6f003093-ade9-4b4c-85fd-6b7a91bc9eaf","width":2126,"height":1266,"filename":"Screenshot 2024-02-07 at 8.41.20 AM.png","type":"image/png","caption":"My first prompt with Galileo","border":false}]}
```

Interestingly Galileo then allows you to:

1. "Choose" which design to riff on by selecting "Edit" indicating to the tool to focus on this single frame. Interestingly in my next prompt I mentioned I liked a feature from one frame and wanted to apply it to the others, but it chose to focus on the first frame and modify that instead. 

2. Export the design into figma (copy+paste). It generates a pretty gnarly design for you, with everything auto-layouted or absolutely positioned, making quickly changing the design in figma a bit of a nightmare.

3. Edit the theme of your designs quickly, this is a nice touch that allows you to quickly see the difference between tones (playful, very round v. sharp, more masculine etc.)

```cv
gallery
{"layout":"hscroll","images":[{"src":"https://firebasestorage.googleapis.com/v0/b/maitake-project.appspot.com/o/pages%2FdZYy1hUn8Ud6iIoIHOvn2T6UGXI3%2F1kUhVRwrCTr7ONfS1Rsy%2F11a4188c-e565-4742-82a8-ba44478e288b.png?alt=media&token=2dbe6509-f12e-4a28-95a8-1435b549ae17","id":"11a4188c-e565-4742-82a8-ba44478e288b","width":2050,"height":1204,"filename":"Screenshot 2024-02-07 at 8.42.34 AM.png","type":"image/png","caption":"Galileo's UI ater generating some options.","border":false},{"src":"https://firebasestorage.googleapis.com/v0/b/maitake-project.appspot.com/o/pages%2FdZYy1hUn8Ud6iIoIHOvn2T6UGXI3%2F1kUhVRwrCTr7ONfS1Rsy%2Ff5ad1d3d-3fd2-4deb-aafa-8abd0698f824.png?alt=media&token=505250a0-c4c7-4eb8-a702-dfaf6fc0e2bf","id":"f5ad1d3d-3fd2-4deb-aafa-8abd0698f824","width":1414,"height":1506,"filename":"Screenshot 2024-02-07 at 8.44.33 AM.png","type":"image/png","caption":"A Galileo frame exported into Figma.","border":false},{"src":"https://firebasestorage.googleapis.com/v0/b/maitake-project.appspot.com/o/pages%2FdZYy1hUn8Ud6iIoIHOvn2T6UGXI3%2F1kUhVRwrCTr7ONfS1Rsy%2Fe5d130f9-531d-4dbb-bad3-043e761205b2.png?alt=media&token=85a75eb5-c6b2-4454-887a-5b382b7311a4","id":"e5d130f9-531d-4dbb-bad3-043e761205b2","width":608,"height":1036,"filename":"Screenshot 2024-02-07 at 8.45.58 AM.png","type":"image/png","caption":"Galileo's theme selection UI.","border":false}]}
```

## Who is it for?

So, after testing the tool I have a few thoughts about how something like this can and should be used. 

### Is this for designers?

This is for designers in the same way Dribbble is for designers—this is a cool tool to quickly get inspired and explore different looks and feels for features. *It's not a tool for designing, yet*. I think we, in the next one to two years, will see a tool like this that *is* for designing.

### Is this for product managers and engineers?

This could be a great tool for communicating your ideas visually, or testing out a theory, as a PM. I would caution anyone in product from using this as a tool to spin up a design based on your PRD or as a shortcut to design heuristics/pattern-matching. This isn't a tool that will design with these things in mind. It will design you a page for your product in the same way apartment complexes are made today—cleanly, repetitiously, and without consideration for the user.

## Closing thoughts

I think this is a cool proof of concept tool for what a design copilot could feel like outside of a tool like Figma. I'd like to see them push the feature set further. I'd also like to see competition from tools like Figma (I'm sure we will see this at Config '24, and I'm sure it will cost extra), TLDraw, Sketch, etc. We need healthy competition and a solid amount of thought put into how we might interact with AI while designing and solving problems, rather than just being handed un-opinionated LLM interfaces. But this is coming, and it's coming quickly. I'll be looking at more of these tools as they crop up so I can understand how to adapt my workflow to them.

Testing Galileo: A designer's perspective

### Basis

Uber has large teams vying for their user’s attention, and all operate on a greedy basis— Rides, Eats, Delivery, and more all with their own marketing, promos, and merchandising that serve to create profit for the company. The company only makes money when someone engages it and uses one of its services, or through their subscription-based service. With all of these teams and goals vying for attention, it becomes important that notifications don’t lead someone to disengage, and are a very important part of what drives revenue.

### History

In the outset of when the Uber team overrode my friend and co-worker Casey Edgeton’s rule of not using notifications for marketing (March 2020), Uber’s notification system and experiments had some issues around scheduling—namely there were quality issues with the content (dead links, expired codes being sent) and timing issues (notifications stacking within minutes of each other, possibly after hours). The marketing team owned these notifications and were manually checking messages being sent. Today’s system came about to resolve these issues as different orgs scaled their contact with users.

### System Design

Uber’s teams engage with all users through an internal queuing and scheduling system that intakes requests for pushes from all teams. The system is designed to answer this question: *What is the best “schedule” with which to send the pushes currently buffered for a user?*

The system takes in pushes that have associated meta-data, if relevant, across the following constructs:

* Push expiration time (e.g., must send the push before a promotion expires)

* Push send window (e.g., only morning hours)

* Daily frequency cap (e.g., at most 2 pushes per day)

* Minimum time difference between push notifications (e.g., 8 hours between pushes)

* Restaurant open hours

Uber’s system considers possible combinations of messages across the day, with different schedules for N pushes fitting into S slots. Notably, it doesn’t combine messages into a single push. It takes all the constraints above and uses a linear program solver in favor of greedier systems like Multi-Armed bandits, in that it can “notice” that a push notif is expiring soon and prioritizes getting it sent out quickly, even if other pushes in the queue appear more likely to get the user to engage in the near-term. If the size of the queue reaches a larger value than the frequency cap, the system constructs a schedule with the most valuable pushes queued for delivery and drops the remaining ones.

**Considerations**

1. The system pairs a push and time and evaluates with a success criteria that the user makes an order within 24hrs of receiving that push at its time.

2. The conditions that their model accounts for are:

   1. *Time features*: hour of day, and day of week

   2. *Push features*: category, content identifier, and deeplink

   3. *User features*: order history by day of week and mealtime, and engagement history with push notifications by day of week and mealtime

Overall, really interesting implications of how we might queue from competing teams or health apps, along with some parameters that can help us keep our system running smoothly day-in, day-out.

Uber's Notification Queuing System

In product design, having mentors from different disciplines can be incredibly beneficial. It's like adding various flavors to a dish; each one brings something new and exciting to the table. These mentors, with their diverse backgrounds and experiences, offer fresh perspectives that can lead to more innovative and well-rounded designs. This approach to mentorship is all about combining different ideas and techniques, much like an artist blending colors to create a unique painting. It's about opening up to a world of possibilities and seeing design through a multitude of lenses.

### **Broadening Perspectives**

My journey to product design through the lens of biomedical engineering has profoundly shaped my belief in the value of diverse mentorship. This unique pathway underscored the importance of incorporating a wide range of perspectives into design thinking.

Mentors from different disciplines can be invaluable in this regard. For example, a mentor with a psychology background can shed light on user behavior, enriching my understanding of how people interact with products. In biomedical engineering, designing with constraints like ergonomics and bio-reactivity really forced a look into user-centricity. An engineer might provide practical insights into materials and functionality, while an artist could open doors to more creative and aesthetic elements.

This fusion of diverse insights encourages me to think beyond conventional boundaries, integrating novel concepts and varied problem-solving strategies. As a result, the designs I work on are not just innovative and user-friendly, but they also bear the unique imprint of a multidisciplinary approach, harmonizing functionality with human-centric design.

### **Personal Growth and Versatility**

My journey in product design has been significantly enriched by the mentorship of three key figures: another engineer-turned-designer, a brilliant user psychologist, and a product manager experienced in overseeing large-scale product launches. Each has uniquely contributed to my personal and professional growth.

The engineer-turned-designer mentor has been particularly relatable, having navigated a similar path from engineering to design. Their guidance helped me bridge the gap between technical functionality and aesthetic appeal in my designs. We shared a common language and perspective, making their advice deeply resonant and practical.

The user psychologist opened my eyes to the nuances of human interaction with products. Their insights into user behavior, needs, and motivations have been invaluable in shaping user-centric designs. This mentorship honed my ability to empathize with users, a skill crucial for any successful designer.

Lastly, the product manager mentor brought a wealth of experience in the business and strategic side of product launches. Their perspective was instrumental in understanding market needs, managing resources, and navigating the complexities of bringing a product to market. This mentorship expanded my view beyond design, considering the broader impact and lifecycle of a product.

Together, these mentors have not only broadened my skill set but also instilled a versatility in my approach to design. Their diverse expertise has equipped me to tackle challenges from multiple angles, making me a more well-rounded and adaptable designer.

On Mentors

### **Summary of system**

Duolingo's notification system is a complex and personalized system. It uses a variety of pre-written notifications for practice reminders and personalizes them based on factors such as the language being studied and user behavior. The system employs a custom AI to select the best notification for each user each day, aiming to help learners stay motivated and on track with their language learning. The notifications are based on a complex set of heuristics that involve modifying multi-armed bandits, a form of reinforcement learning, to optimize user engagement while also promoting learning[5](https://vicki.substack.com/p/duo-the-push-and-the-bandits). This shows that Duolingo's notification system is designed to not only remind users to practice daily but also to encourage long-term engagement and learning. The system employs a custom AI to select the best notification for each user each day, aiming to optimize user engagement and promote learning. Additionally, the app leverages social influence, competition, and teamwork to encourage user engagement.

### **Basis of Notifications**

Duolingo uses notifications to engage users on a daily cadence for an opted into language learning practice (also more recently mathematics). In their bandit system, each round corresponds to any notification that is sent to any user, and each arm corresponds to a template that can be chosen. These notifications are based on a modified multi-armed bandit algorithm that pulls from a series of hand-written templates which may contain slots for user-specific and personalized information. The use of a MAB for selecting the right template based on user behavior is common in recommender systems like this one, but Duolingo found success in modifying their algorithm to accommodate two observed differences in their system:

1. Fresh templates, or templates that haven’t been recently seen by the user, have a higher engagement rate (otherwise the MAB would converge on the single best-performing template in a vacuum)

2. Some templates are only appropriate when users are in specific states, with certain conditions met, such as risking the end of their streak (the gamification construct), or that the user has indicated it’s using Duolingo to prep for a trip to a country

### **Duolingo’s modifications**

Duolingo modifies the pure Multi-Armed Bandit approach based on two related but distinct applications: the recovering bandits problem, the rested bandits problem, and the sleeping bandits problem. Recovering bandit says that the expected reward/outcome of an arm is based on how many rounds have passed since the arm was previously chosen. Rested bandit says that the expected reward/outcome of an arm is based on how many times the arm has been chosen previously. While the authors feel both problems are relevant, they weight the recency (recovering bandit) as more important than the number of occurrences (rested bandit), since timeframes extend over months and years for a Duolingo user. Sleeping Bandit accounts for the second observation above: that some arms might be ineligible during some rounds, meaning that some arms are conditional.

An interesting difference between forward’s engagement and Duolingo’s: many of their daily lesson completions, the purpose of their notification system, happens organically (although it’s impossible to know the true causality, or distinguish between organic engagement and notification-based, since someone might see the notification but not click through. This introduces noise). Whereas Forward’s engagement is majority inorganic, meaning we push someone and they engage with their app to accomplish the action.

Some of their learnings around experimentation and sample size leading to odd/degenerate behavior feel very relevant though: For large test groups, the authors generally see 90% perform within +/- 3.4%,  normally distributed. For smaller sample arms, they saw only 50% of arms  perform in that range, with 38% having scores +/-5% for their relative difference in performance, leading to that strange behavior in their bandit algorithm. Essentially, since our test groups are going to be small we should be cautious of false positive/false negative performance of a specific type of notification or template. The authors used Bayesian estimation toward a prior to accommodate

**Takeaways:**

1. Freshness and variance of content matter quite a bit when trying to achieve a goal, especially a homogenous goal. This means that if we need a user to log weight or something repetitive, leveraging different ways of communicating that prompt could increase success.

2. Personalization based on user behavior will increase success—for example, if someone habitually engages at specific times of day over others, tailoring our system to anchor that behavior will increase engagement.

3. Pooling templates for engagement centrally and injecting customized content into them is a successful model for engaging users across different use cases, leaning on psychological principles rather than just the content itself as a reason to engage.

4. Adapting a holdout set where an arm is chosen at random, as well as a softmax system, will accommodate adding new templates with no penalty to the system at large.

Looking at Duolingo's Notification System in 2024

This year I’ve spent a lot of time thinking about how I work. I think the tendency for people is to do this when they come into one of a few different situations, or multiple. But more on that later. I’ve been thinking about how I work because I want to do better work quicker with less headache. That’s a lot of problems to solve and you could imagine saying, “slow down, take that one at a time!” But that’s the premise of the challenge, right? Find tools or strategies or ways to achieve all these improvements together, intuitively. I spent the first half of the year thinking about different strategies that make sense to my way of working, then spent the second half of the year searching through different tools to enable these goals. I’ll talk about both in turn, but want to discuss what I learned from a meta sense during this process.

## Strategies for Work

Here I’ll list out the strategies I’ve had success with over the past year. A lot of these are specific to how I think about problems, and I wouldn’t prescribe this as a one-size-fits-all toolkit for everyone. But they could help you next time you feel stuck.

### Asking questions until you can’t help but understand something

This is my favorite method to get involved in product strategy and get opinionated about my design when the project is in a super ambiguous state. The process looks like taking the problem being solved, and writing out all my questions about that problem, trying to be exhaustive. I don’t prescribe an order or hierarchy or importance to these questions, but I make sure to get them all down. From there, I go one by one, taking each question in isolation, and I make my best guess forming an opinion or hunch about what the answer might be. I’ll end up revealing other questions and issues that might get between me and creating solutions for the problem at hand—which is great! Revealing these second-order questions is valuable, I’ve been able to do rounds of self-feedback by being honest with myself if my solutions or hunches don’t make sense. This tool is amazing for helping get all my stakeholders aligned—if our mental models are off by degrees or off by a lot, we’ll figure that out quickly once they start going through my hunches. It’d be strange to think that any of these hunches will hold up when building product, but early stances others can react to has been hugely valuable for getting everyone together early and going wide in our thought process.

### Going deep on wireframes

This isn’t a new concept at all, but I’ve been using wireframes as sacrificial lambs to conversations more and more lately. Giving others something visual to point to can shape how people think about their problems quickly and inform my own design work by getting specific feedback early and often. Waiting until something has entered mid-fidelity to get it in the hands of leads is a death knell to progress when building product at my company. I’ve been training leads that they can engage at lower and lower levels of fidelity over the course of this year, and it’s been incredibly helpful in getting to meaningful discussions about product at a fast cadence.

### Designing from ideal first

There were times in the past where I would immediately start design work with constraints in mind. Whether those were time, technical, or feasibility constraints, they loomed over explorations for early projects in my mind. I thought I was being crafty—skipping a step, designing something I knew my engineering teams could build quickly and achieve what we needed. I was selling the product short—I wasn’t able to expand my mind to solutions that would require an army to build, or more time than I had to design the high fidelity version of, but those were the designs I needed to see, needed to show my stakeholders to set vision. These days, I start from my ideal, from an opinionated place that lands the concept in the best way possible from my skills. As you improve your craft, the ideal comes faster, the clean-up takes less time, and you have time to set the vision and have a goal in mind. Even if creating that ideal is a gargantuan task, I could at least point to it with any constrained designs I do from that point on. Those constrained designs end up better for it, every time.

## New Views

This one is interesting to me. I’ve recently been tinkering with new tools and hobbies almost weekly. I’m writing this on a writing app I started using two days ago, to see how I work when I use it. I’m going to write this list to be tool-agnostic, I think it’s better to find tools that suit your own taste, rather than relying on a recommendation from me. I wonder if it’s my curiosity about what other people are building, or me exploring how I can play around with my own work stream to feel better about my work and achieve that lofty goal I set out for at the start of the year. Either way, here are a few of the tools I’ve been using to help me do what I consider to be better work.

### A film camera, like a [Nikon FM3A](https://en.wikipedia.org/wiki/Nikon_FM3A)

This has been one of the biggest things that changed my view on the world in the past decade. I’ve always had a camera of some kind, a digital micro-4/3s camera was the one before this. But then I got the inclination to dive into the world of film this year, and it’s changed the way I view the world around me and the work I do on any project, whether at work or not. I started shooting much more, and more intentionally. Each shot costs me about a dollar on 35mm film, making it count matters. In the same way, I’m trying to be more intentional in a lot of different aspects of how I’m spending my time. It’s been super gratifying to see myself become happier, healthier, and feel more balanced through being more intentional, even if it sometimes feels like I’m trying too hard.

### A new design tool, like [Play](https://createwithplay.com/)

Design tools are awesome, and as a designer it feels meta to look at a tool created for someone like me by someone like me. Design choices made here and there—you can tell what is in service of retaining designers, what’s in service of growing to new ones. This new tool is built with a lot of love baked in. It’s the first tool I’ve used to successfully create an intuitive mobile app designer on mobile. You can design a mobile app straight from your mobile phone. And do it well. This is not an easy thing to achieve, and to see that has inspired me to think about the intersection of novelty and intuition. How can you create a brand new experience that changes the way people think about all their other experiences, while feeling familiar?

### A new writing tool, like [iA Writer](https://ia.net/writer)

Who needs anything other than Google Docs? That’s what I asked myself when contemplating the thought of paying for a writing tool in 2023. But curiosity got the better of me. I’ve been thinking a lot about content ownership of information and artifacts—whole companies have started with a user-first ideology around content ownership. I’m using one of those writing tools right now. It exports into .txt, and it looks great. It’s a focus-oriented tool, to reduce clutter and cut through to getting pen to paper. I’ve been able to write more focused, better crafted documents using this tool than I have in three years. I wish that were hyperbole. Now, am I writing more because I have sunk cost paying for a writing tool? Maybe. But I find it to be more engaging and enjoyable, much like the film camera I described above when compared to a digital one. The tool makes me want to write more, and aspire to be a better writer. I could’ve achieved the same end by sitting down to Google Docs, but... who wants to do that?

### A photo and link aggregator, like [mymind](https://mymind.com/)

### I want to believe that we’ll see a lot more of these in the near future, leveraging all the new AI tech coming out right now. The promise of these tools is a brain dump—a concept that has existed since long before the first computer, but has been getting closer and closer to its platonic ideal over the last few years. Place on object into an endless bag along with hundreds, thousands of other objects. Months later, when you need that exact object and objects like it, you reach your hand in and immediately retrieve it. That is the promise of these tools, and I’d like to believe it will enable better work, with more focus and clarity in time. It’s likely to kill any sense of information retention when used in isolation, but if you take the time to form opinions about each thing you place in, you’ll develop a strong semantic web along with any artificial web this tool layers over your artifacts.

### A niche social network, like [Posts](https://posts.cv/)

I’m learning a lot about how to put myself out into The Public more, both as a content creator and engager. It’s been an exploration in learning what I feel like sharing, finding content that both feels enjoyable and relatable, while expanding my mind in a fun way. Sounds crazy to say that finding a small community of internet strangers has inspired me to do better work, but it has. The community has taught me how to develop stories around what I share and talk about. This could take the form of friends, a community in your workplace, or a group of strangers you stumble upon online. But learning that your opinions are valuable and valid can do nothing but benefit you as you go about your life.

## Wrapping up

I can’t say what led me down this path to find new ways to work and think, but it’s been super enjoyable. I’ve been able to be authentic with myself and give myself the grace to explore things that interest me, even if they’re stereotypical or nerdy or off the beaten path. It’s been a good year.

New Tricks: A 2023 Retrospective

I have a hunch about what design will look like in 5 years. I believe we all know it's going to look significantly different. This is evident from the work being done by the teams building the future of design tools. But to understand the scale of the difference between where we are today and where we're going to be, we should examine design from various perspectives. Although I could discuss hardware engineering and physical design since that's where I began my career, I think that's a different topic for another article. For now, let's focus on digital design.

We can start with the surface you're looking at right now—a website that has been created by a team with specific goals in mind for the benefit of its users. In the near future, we will be able to go from designing in a tool, in low fidelity, directly to a polished, responsive, interactive website with just a few clicks. The transition from a polished design to a polished website is already possible with tools like Anima, which integrate with Figma and can generate websites based on a design. Anima preserves the design exactly as it is, and you can even create a responsive website if you configure your properties correctly. However, the key difference is that you still need a skilled designer to conceptualize and express your intentions in a tool like Figma or Adobe XD before this becomes an option.

Some members of the internal Figma team ([Sawyer Hood](https://x.com/tldraw/status/1726702203279384774?s=20) and [Jordan Singer](https://x.com/jsngr/status/1725614452123250792?s=20)) have made incredible progress this month by developing a way to generate responsive designs coded in Tailwind from basic wireframes and prompts using GPT-4Vision. I find this amazing because, at its limit, it resembles designers collaborating with an AI "partner" that rapidly creates and refines designs based on human feedback. To implement the new design, the designer simply needs to feed back, verify the changes are as intended, and click a button. Consider what it would mean to train the tool with your specific design system or an app with a style that you aspire to. The tool can also explore screens with different brand identities, helping designers (or anyone) discover preferences while in the exploratory phase. These tools will make designing and launching websites and apps incredibly accessible. The goal is to streamline design and shipping cycles. As more users adopt these tools, design cycles will shorten rapidly in the industry. Frontend work will be (mostly) automated, freeing designers from relying on frontend development. These tools will increase the presence of product thinking in designers' workflows and design work in product thinkers' workflows.

We have all heard or considered the potential drawbacks of relying on such tools—innovative, new products might become scarce, similar to the standardization of architecture (do you know the name of the architect who created the plan of your office, apartment, or home?) unless deliberate efforts are made, since training relies on existing content. On the other hand, as more people develop purpose-built, opinionated software, niche products will flood into our lives. Brand will continue to be king in a lot of software circles, I doubt that will change.

Is it good that design is heading this direction? I don't think that's really for me to say. The momentum is here, so I'm already thinking about how I'll adapt. There are obviously some caveats here—things like backend work, animation & interaction design, and software-hardware integration aren't covered by these recent changes. But could they be? I think we'll find out.  As for the present, I'm staying up on the news and development of these tools, and as soon as the barrier to entry is low enough, I'll be integrating them into my workflow to see how to develop this new skillset. The times are changing.

Tightening the Loop

Building something new takes time. I've been working in an org where we've been doing a ton of building from first principles for about the last two and a half years. We're starting to get the product we've thought and theorized about so much out into the hands of the people who we've been building for this whole time and it's an incredibly gratifying, cathartic, and scary feeling.

Building always takes time, especially at the intersection of hardware and software. Sweating the details is table stakes, and you're going to make mistakes. That's ok, you'll build your way out of them. The difference now is that it isn't internal stakeholders giving feedback on my designs, it's the people I want to help with my work. That makes things incredibly real, and has really lit a fire to move quickly and build even more on the experiences that we've been sweating for so long.

Our press launch is today, and the nervous energy has been palpable in the office, in every meeting for the past two months. A lot of people have been holding their breath for a long time. It's worth reminding ourselves of what we believe to be true about the product we've been working on, and how much that matters.

Something New

I watched a fantastic talk by Amelia Wattenberger the other day. Since then, I’ve been thinking about how the concept of varying the level of zoom on a problem or topic to accomplish different aims, often referred to as a ladder of abstraction, can be applied to just about any problem in today’s world. I’ve been thinking about health and wellness, specifically. This concept allows us to focus on different aspects of a problem or task at different levels of detail, depending on our immediate needs and goals. The same is true of the users of tools, services, and products. The ladder is a designer’s tool to help someone tackle a larger problem that they might not be an expert in but are still interested in addressing.

Consider the example of writing a book. At the highest level, we have the entire narrative. This is where we can influence the tone, plot, and overarching actions of the story. It's akin to viewing the curve of a plot, where we can shape the trajectory of the narrative.

As we zoom in, each chapter comes into focus, allowing us to delve into specific events and character developments. Zooming in further, we see the individual paragraphs where we can manipulate the pace of the story, introduce new ideas, or build tension.

At the sentence level, we can polish the language, refine the voice of the characters, and ensure the flow of the text. Finally, at the word level, we can select precise vocabulary to evoke specific emotions or imagery.

The same concept can be applied to someone’s health. A user's goals at different levels might look like this:

1. I want to be healthy (highest level)

2. I want to improve or maintain specific aspects of my health

3. I want to control a specific element of my health

4. I want to improve a specific health metric (e.g., vitamin B12 levels)

5. I want to manage a specific physiological response (e.g., cytokinal response)

These *goals* can be mapped to *system actions*:

1. Automate overall health maintenance (e.g., prescription refills)

2. Address specific health goals or chronic issues (e.g., weight loss)

3. Address a specific health issue (e.g., back pain)

4. Track specific health metrics (e.g., LDL cholesterol)

5. Influence specific physiological pathways (e.g., cytokine activation)

In essence, the ladder of abstraction allows us to tailor our approach to a problem, depending on our immediate needs and goals. This concept enables us to navigate complex systems effectively. It’s our job as designers to help our users navigate the systems we build for them in a similar fashion—tailoring your interfaces and content to correctly help someone know what actions to take next, every time.\
\
For more on the topic, check out Amelia's talk at AI Engineer Summit 2023 [here](https://youtu.be/PAy_GHUAICw).

The Ladder in Practice

### **Basis**

Notifications serve as means of engaging the user around timely or important information from different facets of an app—its core content, marketing, etc.

### **Permissions**

Before an app can send a notification, iOS offloads perms acquisition onto each app, apps will make the most convincing argument to why they should receive notification perms to the user. Often this takes the form of some enticing messaging paired with a design that primes the user to accept the native-served Perms modal. The apps often also give users the chance to opt out of notifications without opting out through Apple’s perms modal, as each app only gets the chance to serve that notification once, and the choice is binary, no escape valve. If the user accepts, the app gains the ability to send notifications out into iOS’ aggregation point, the Notification Center. Settings has more fine-grained control about which permissions and styles of notifications are available on an app-by-app basis. Notifications from apps are also allowed or disallowed on a per-app basis by different focus modes on Apple’s products.

### **Notification Center**

Notification Center is iOS’ aggregation point for all notifications from apps with permissions. It allows users to view and manage notifications, by either dismissing or tapping to engage with a notification.

Notification Center is present on a user’s lock screen by default, meaning it is often one of the first things a user sees when interacting with an Apple product. It is also omnipresent when using an apple product, hidden behind a gesture but with the ability to override the screen from anything below it (other than Apple Pay/Wallet). Its prominence allows users to receive timely and important pushes from the apps they subscribe to, serving as a parallel to organic engagement with apps, as I mentioned previously.

### **How Notification Center handles notifications**

Notifications from apps are served in chronological order. If multiple notifications are present from the same app, Notification Center groups them to prevent unnecessary context switching. A “Stack” of notifications from each app is also ordered chronologically, meaning that the believed importance of the notification sent by a creator doesn’t override recency.

### **Interruption Levels and Focus Modes**

Interestingly, Apple requires builders to set an interruption level for every noncommunication notification their application sends. This ranges from Passive →Active→Time Sensitive→Critical, with the final two breaking through focus and overriding scheduled delivery. Critical level notifications override Silent mode on iPhone, however you have to request perms from Apple to use this level.

I don’t get the sense Apple restricts the use of the other three Interruption levels, but lays out best practices for using Time Sensitive notifications as they break through conditions the user has enabled to prevent notifications. The implication is that Apple wants creators to believe that users will disable all notifications for their app if trust is broken. That being said, Apple also provides guidelines for promotional or marketing notifications that it seems most apps don’t follow.

Focus modes are user-defined modes that allow additional controls on which apps can push notifications to a user when engaged. These modes, if blocking a notification from an app, will queue into a scheduled delivery when that mode is disabled, or rather when the user enters a mode where that app has notification perms again. Scheduled delivery is also available outside of Focus Modes as a concept and users can choose which apps are included in the scheduled summaries delivered at user-selected points.

### **Implications**

1. Today, health apps by default have permission to notify a member at any point or engage the member in any way. We may earn the right for that not to be true at some point in the near future.

2. Our notification tray is a similar centralized point to iOS’ Notification Center, but isn’t accessible other than from the Homepage. There are practical benefits to exploring other ways of allowing engagement from any point within MBA, especially if we’re worried about thorough notification engagement or completion in the future.

3. Notification management is a concept that we have opted to not take on up until this point—we have a naïve system that equates tapping a notification to completing whatever action is associated with that notification. This is true of iOS’ system as well, with the insight that as long as we enable the proper routing structure (deep-linking to content or interaction), the onus for having a user complete the desired action should fall on the Health App. Interestingly a notification that prompts the user into a high-friction action like a CarePod visit likely won’t fit well into this system.

### **Meta Commentary**

1. Interesting that our structure today feels very similar to iOS’, however fails the user in multiple ways. We only recently allowed correct deep-link routing, for example. We also do not present the notifications up front, I’d be much more likely to engage with a health app that has a badge than the notification bell in the top right, as other apps have trained me to believe these bells with red dots are often useless.

2. iOS offloads the optimal content onto the apps using the platform. If we centralize our content, it’s likely that we’ll scale the system of possible templates rapidly, but making sure these templates are relevant to the Health App in question becomes a larger task.

3. iOS also offloads the engine for pushes to each app as well, outside of OS-related blocks like Focus Modes or Scheduled Deliveries. We have no notion of an engine outside of a naïve if statement around the presence of new content (intentionally, we didn’t need anything else).