Scott Belsky Talk at South Park Commons

Often designs from frustration
Right now, greater skill is being brought by compute and developing a democratization of many things (code, design, etc.). Because of this, taste will probably be the most important skill
Taste is derived from culture and overlap of industries
Because of that, there is no where better to be than NYC
Has a principle known as the “Law of Displacement Speed”: in fast moving industries, things replace each other faster and faster. This happens in all platform sfhits in moments where something foundational is changing
Because of all these changes, how do you know what to do?
- You go back to your customers and use empathy
Many of the main companies for IA right now are major industries in NYC: healthcare, banking, etc.
New companies should do things that don’t scale as big companies cannot do those things
This happened at Behance, whose mission was to organize the creative world at work
- They conducted one focus group in the company’s history. When they told folks they were building some kind of portfolio site, designers told them they didn’t want another place they had to update. Yet, when they asked designers what problems they were really struggling with, the problems were pronounced
- To get traction, they started a blog and interviewed creators whose work they admired. They created a portfolio using the Behance tooling for each person profiled (around 100 portfolios of the people they admired most). There were 2-5 projects per designer, 12 projects per page. So when new users came to the site (and in the days of no infinite scroll), there was a roster of work to look at, pages of pages of it, so it was easier for others to want to join in.
For Scott, the true unicorn is successful M&A because it’s how companies change
- Adobe was made through acquisitions and the people they bring in through acquisition end up being in change
- It’s successful to keep Adobe thriving because they don’t make acquisitions function how they want them to (like a Google). They want to see what newness the acquisition brings into how they work
- People are at Adobe because they want to be there, to go deep on things that they can go deep on topics because these are tools built for pros. They can’t go as deep at a place like Apple where they have to make things for everyone. That can feel limiting.
“Founder Mode” is really about people being truthful on what they’re feeling about work (and sharing it with the top)
Every customer is in some way an exception with how they use your product. The happy path does not exist.
At large companies they have plans for the year, so they make them and have something innovative being worked on, yet after two years they wonder what people are doing in that part of the company. Founder Mode is about keeping that team where they are and not pulling their resources for something that’s important only in the immediate.
For startups, it’s an advantage that they don’t have customers, they don’t have to bring those people through changes
- At Adobe, they had to fork Lightroom, split it into two products, so they could make one that was more optimized for mobile and the cloud. They had to do this so they wouldn’t alienate existing customers
In a platform shift like the one were in, you’re able to help your customers understand what is new and how it could benefit them. You have a responsibility to really bring customers along.
It’s important to think about how to do things the right way: Adobe had to train their own AI media model on licensed content.
For the consumer, novelty does precede utility
- Big companies don’t prioritize surprise and delight enough
It’s hard for big companies to transition
For startups, the advantage is really about focusing on differentiation in both data and interface.
- In data, it’s about capturing data that can give users an experience that they can’t get elsewhere because of it
- In designer it’s about simplifying workflows in ways that competitors couldn’t
All products have first-mile experience that most people will see. In those first few moments of using a product, most people are lazy, vain, and selfish. They just want the product to work for them in that’s useful to them.
- You need to make it that a user quickly feels successful right in that first mile experience
  - This is product-led growth
- Yet often, these things are the last thing you work on: the onboarding and the defaults
Here’s what great companies do: they take something that’s really hard and make it into something that’s really simple.
- Users are refreshed by the easiness and thoughtfulness
After you launching something, you start listening to power users as that can help you monetize. Yet, they won’t give you feedback that improves your first mile experience as they’re not first adopters, those first customers anymore. These people are now pragmatists and have to use your software because someone told them too. What you focus on needs to be different for them.
The “messy middle” is about being in a place of constant anxiety, ambiguity and anonymity (as a company).
- Being a profitable startup with low growth is really hard because it’s hard to see the end in sight (i.e. acquisition)
- Here are some hacks for navigating the messy middle:
  - There’s an endurace component, about keeping the team together long enough to really figure out the work at hand. It’s about highlight the small wins, wins that aren’t important and won’t necessarily be monetarily focused at this stage.
  - As a CEO, you really need to narrate the journey for your team. You nee to continue the motivation for the team, contextualize what they’re trudging through and why it’s worth it.
  - You need to also decide if you want to stick it out. If you’ve gained conviction through the continual grind of the work, keep going. If you haven’t, there’s no shame in quitting.
The idea of “AI knobs”: All products these days are about interest graphs and not follow or friend graphs. Social platforms are always bringing you to the edge of something of what you knew. In something things (like art maybe) that’s great. When it’s politics and conspiracy theory, it can be very dangerous as the knobs rachet up something more wild that gets engagement.
- The leaders at these companies are the ones playing wit these knobs and they’re not making the changes to them that they should be
- We need some kind of designer code of ethics
Will AI take over design?
- The “prompt-based” AI era is almost over. Next generation will be control based. It will be about given people tools that are similar to the tools in software like photoshop, where with precision and a sense of taste, user can control an AI’s output.
- AI is lowering the floor and raising the ceiling: More people can design but also designers can focus on higher-level things. With Adobe’s Regenerative Recolor, designers don’t need to spend hours testing different color combinations for their vectors. They can focus on making the design in those vectors even better. It helps them do the higher order work.
- Another example of this is Adobe’s “generative extend” that extends a scene just by a bit rather than a whole crew and cast needing to to costly and time consuming reshoots.
- Some people are jumping at the possibilities of AI, others are nervous about what it will do for their career
- It always goes back to the same things: focus on hiring for skill, taste, and experience. Those three are really something that aren’t changing.
“Creativity is the world’s best recycling program”, it’s always about taking what exists and incorporating it into something new (references etc, that create a new idea).
- But right now, the AI model that uses creativity is incorrect. If your model is impersonating someone else’s style, they should be compensated for it.
- Maybe creatives could make money while sleeping if there was a market they could lend their style to an AI to use and get paid for it
With Behance, they pushed their algorithm to show things to users that were outside their field as creative people are stimulated by creative endeavors in other fields
With AI, we have a chance to get a really human experience wherever we go because of aI agents. Every shoe business should just know your size automatically. You shouldn’t have to always place it in. There’s an opportunity to “feel special as a service”.
You can cultivate a good group of people in the now very large startup space by finding people you admire and being each others board of directors, engaging with and challenging one another
- People challenging each other, with respect, is key.
Right now, there’s somewhat of an AI agent pyramid:
- At the bottom, AI is glorified help
- The next level is automation. It’s about letting me doing something for you.
- The final level is proactivity. It’s about helping someone see that there might be a better way to what they’re doing and nudging them to understand why that is.
  - For this version, the tone needs to be right. You can’t just have an AI telling people what to do. They need to see it as a tip.
Right now, there’s a debate of the multimodal model: having an AI model that you can throw any kind of content into it and get any kind of output you want.
- Some think that’s the be all and end all yet this kind of version will be very expensive
- The future, is most likely, about the orchestration layer, figuring out how models can work together in new ways. A lot of the models will probably be based locally. It won’t just be about working with one model, it’s about how you orchestrate these models.
  - Let’s say an AI agent needs to pay different vendors for things, what’s the enabler that helps get that payment done?
- This orchestration layer will be a winner and is huge
  - Models will eventually have a lot more APIs that will give greater access to third parties in the future (AWS did this back in the day).
Adobe isn’t building LLMs. They’re partnering with them. But they are building media models and working on the orchestration layer of leveraging these larger LLMs.
Founder Mode is really about going right to the source of the talent, not to their manager or their manager’s manager. It will keep those people at the company longer and the people between you (as a high level manager) and them need to realize you’re not side-stepping but that the way is really to talk right to the people. You need to explain to them that it’s about a direct connection and that should be okay and really betters talent retention.
We need to remember, as humans, we’re find with some things being public. We want some things private though, and we want the choice. Few companies leverage their first party data well.
Some company should make it easier to have portability of preferences, that the right things will be shared with trusting AI. Most likely this will be an AI we pay for, one where it’s the product and were not, that it’s working for us.
Scott has a litmus test for ideas that will resonate: we’re always kinda wanting things that take us back to how things were, where people knew you and knew what you liked, but at scale and efficient. It’s taking the idea of living in a small place but making it something that can be do on a large scale and well. We want to be known, we don’t just want to be anonymous. But we want to be known on our own terms.

Why Centralized AI Is Not Our Inevitable Future

Apple Intelligence is Right On Time