(This article was adapted from a talk I gave at DesignOps Melbourne in August 2023 called Take the Pink Pill: why Best Practices are dead ends.)
I’ve worked in digital and advertising agencies for a decade, and I was a designer and developer; sort of this weirdo that was really interested in both things when most people are kind of only interested in one or the other.
I used to co-run the DesignOps Melbourne meetup with Ch’an Armstrong, and then I quit the meetup. I quit my job the week after Figma plugins were announced, and decided to start a company building Figma plugins full-time.
That company is called Hypermatic. It used to be called Figmatic, which was named after Illmatic, my favorite Nas album, until I got a very nicely worded letter from the Figma legal team saying I had seven days to rebrand. So I spent seven days reading about ten books on branding, wrote this crazy script to work through all the possible name combinations for the domain, paid far too much money for hypermatic.com, and here we are.
The goal of Hypermatic is really to help designers spend less time crying and more time designing. To help them get out of the corner or under the table wanting to go home, and actually enjoy their job rather than doing all these meaningless tasks.
One reason I’ve been able to work this way is that there are no meetings, no investors, no Agile, no Jira boards, and basically no distractions. I get to work on whatever I think is most important, because I just say no to everything I don’t like. It is amazing how much you can get done, even as a company of one, when you have hours of free time every day to work on whatever you want. You would never go back if you understood how good that reality is.
Human-centered design without the humans
One thing I wanted to get a quick check on during the talk was how often designers are actually talking to real users.
I think this is a weird paradox. We always claim that we’re human-centered designers, and yet I’ve found that consistently, designers barely talk to any humans. In a weird way, that’s a little bit depressing. You could almost think of the sales and marketing team as being more human-centered than everyone else, because they are actually talking to users every day. They understand the intricacies of their problems. The support team probably knows more than any designer actually would, and I think that’s a little bit strange.
Being the only person at my company, I’ve talked with thousands of Figma users from all over the world over the last few years, and I’ve learned a lot of stuff. The two key takeaways that are very consistent are that everyone is actually doing their best, and nobody has it all figured out yet. Even if you get a peek into all these huge companies that you know, the big tech companies, nobody actually has it quite figured out yet.
So this is the part where we dispense some red pills and follow the white rabbit into The Matrix.
Everyone is familiar with the Matrix: you take the red pill and see these uncomfortable truths, or you take the blue pill and go back to dreamland and wake up forgetting about it all. But if you upscale the 4K Blu-ray, you can see the contents of the blue pill, and it’s not something we want to be ingesting. It’s this sickly sweet file format that has been lingering around for the last 20 years, and even though we think we’ve purged it, it’s still around in a different form.
What I mean by this blue pill analogy is that we’ve basically come to accept that drawing rectangles in Figma, which we then hand off to be recreated in a different medium, is normal. We almost don’t even think twice about it anymore, and I think that’s sort of disturbing.
If you look at The Truman Show analogy, we accept the reality of the world we’re presented. A lot of people who are new to the industry just think, “that’s the way we’ve always done it, so we might as well keep doing it that way.” Unfortunately, I don’t think this is going to be a good approach for the future.
Best practices and dogma
Talking about best practices, I think you can think about it on two ends.
At one end, you’ve got extreme dogma, where nothing can be questioned because we know everything is certain. We’ve reached absolute truths, and there’s no point questioning them. Then, on the other extreme, you’ve got doubt, where nothing can be certain. I’m not wearing this ridiculous uniform, you’re not actually all here, and nothing can be certain at all.
Both ends of the extreme are dangerous. On one end, you can’t question anything, so you don’t do anything. On the other end, nothing can be certain, so why even bother?
I think best practices have gone way too far over to the dogma side, and we need to nudge them back the other way.
Ed Catmull from Pixar described this when they were working on Toy Story 2. “Trust the process” had become this mantra and this crutch that was distracting them from actually engaging with their problems. His position was that we should trust in people, not processes.
But a reason why we have best practices is because it’s actually really hard to do new things inside and outside of companies. You barely have any time to do your job. It’s unreal how much non-design work, or how much non-interesting work, we actually have to do at our jobs, let alone innovate or take any risk.
The risk factor is important. If you try something new and you fail, you’re viewed as an idiot and you’ll probably lose your job. Even if you succeed and try something new, you basically gain almost no credit. Whereas if you follow best practices and fail, you followed the best practice, so you’re basically just seen as being unlucky and you get to keep your job. If you follow best practices and succeed, you also get to keep your job.
So the risk-reward asymmetry is totally out of whack if you’re trying to incentivize people to try anything new.
I’m also interested in where these best practices even come from. We kind of inherit them and don’t really think too much about them. This is a very short list, and you could go on forever, but I only had about 30 minutes, so I just briefly covered a few examples.
What if the experts are wrong?
We can start with experts, and the question: what if they’re actually wrong? What if we’re just receiving all this wisdom from people we think have it all figured out, and actually they don’t?
Flash back to 2000 and Y2K, and there was this internet article about the internet which claimed that it was probably just going to be this passing fad. Everyone was kind of giving up on it. It was only teenagers playing on Neopets, there was no commercial application whatsoever, newspapers would never go online, music would never go online, and so on.
The people who came up with this literally called themselves experts from the Virtual Society project, which was groups of people from 25 universities across the US and Europe. Dozens of people all agreeing that the internet was nothing to worry about, just a passing fad, teenagers playing Neopets or whatever.
In that case, if you had just bet the opposite of exactly everything they predicted would happen, you would be a billionaire.
This also applies to industry standards. We look at “the industry” and assume they must have it figured out. They must know the industry practice. One example of this is the Spotify Model.
Back when I was in agencies, it seemed like every second person I talked to at different agencies was about to adopt the Spotify Model, this tribes model. It was so dumb at the time, but in retrospect, the people who were part of it at Spotify ended up writing an article to clarify it because they got so sick of people copying the model.
They wrote an article saying Spotify doesn’t use the Spotify Model, and neither should you. One of the key quotes was that even at the time they wrote it, they weren’t doing it. So they put out this big thing on the Spotify Model, and they weren’t even practicing it at the time.
That gets to the crux of what I’m trying to discredit about blindly following best practices. In this case, it didn’t even exist. There was nothing to copy in the first place.
Preference falsification and fitting in
Another piece of this is preference falsification.
Preference falsification is essentially misrepresenting your wants under perceived social pressures. What this results in is a public preference that is actually at odds with private preference, and there’s no easy way to gauge it. It’s slightly different from lying, but it’s in the same ballpark.
This extends even further into things like the Asch conformity experiments, where they would take five or six people who were all paid actors, then hire one person who had no idea what was going on and put them at the end of the line. They would ask these basic, obvious questions, and the first five or six would deliberately answer the wrong way.
They would say, yes, that’s definitely line C, and then you get to the person at the end who has no idea what’s happening. They did this time and time again, and 75% of people knowingly gave at least multiple wrong answers. The reason they gave was that they just wanted to fit in. They didn’t want to cause any trouble. They didn’t want to be seen as stupid. They actually went against what they could see with their own eyes just to fit in.
I think there’s this really weird dynamic with groups, where you go from the wisdom of crowds to the madness of crowds very quickly.
This also extends to innovation. We hear that it’s a best practice to get a bunch of people in a room to innovate, as if sitting down for 30 minutes with 12 people is going to lead to anything. But there are all these factors at play that totally derail and sabotage it.
You’ve got a natural authority hierarchy, with junior people, senior people, maybe the CEO in the room. The meeting is always dominated by a few people who just won’t shut up and won’t let anyone else talk. There’s the fear of social rejection. There’s bike shedding, which is the idea that meetings always gravitate towards spending a disproportionate amount of time on the most trivial thing.
The example is that they’ll be talking about building a nuclear power plant, and it’s an hour meeting, and they spend the first 55 minutes talking about where to put the bike shed out the front. That’s where the term comes from.
What ends up happening in these groups is that you actually end up with the average opinion of the average person.
The average opinion of the average person
This is why every website at some point started looking the same. If you removed all the logos, you wouldn’t know which website was which. They have absolutely no brand and no sense of self-identity, and I think this is a reflection of what happens in these brainstorming groups.
One of the strongest things I’m convinced of is that a lot of this can be described through this innate mimetic desire that we all share.
Rene Girard discovered that most of what we desire is actually mimetic and not intrinsic, and that imitation plays a far more pervasive role in our society than anyone else had openly acknowledged. People are intrinsically on autopilot to desire what other people are doing.
It leads to things like the exact same Instagram photos, Airbnbs and interior design that all look identical, car manufacturers that used to be defined by which country they were made in now all looking exactly the same, cafes all looking identical, and brands all falling into the same trap.
You can look at a bunch of toothbrush brands, for example, and every single one is a different brand but they all look the same. Are you telling me every single one of these brands got together, came up with these unique strategies independently of each other at different periods in time, and deployed them all at the same time? There’s just no way. They’re all copying each other, and this just happens every day.
We’ve kind of distracted ourselves with all these things to convince ourselves that everything is normal and we don’t really have to do much. We’ve settled on all these absolute truths, so there’s no point doing much more. We’ve got all these best practices that we’ve learned in university and in the workplace. We’ve got group consensus on everything, so there’s not too much to worry about.
But I think we really do need to get back to the future, and I believe the way out is to accelerate.
We need to get away from the extreme of dogma and go back to the side of doubt, where we can start rethinking some of these things, because everything is not certain. We are not at the end of history.
React, AI and the worst it will ever be
React kind of proves this.
If you remember when React came out, there was a very strong, visceral negative reaction in the developer community. It was seen as crazy. People were making jokes about Facebook rethinking established best practices, and now you’d be struggling to find a developer who doesn’t agree that it was obviously the right direction to move in.
I was trying to talk about this in 2018 when I gave a talk called Insanely Inevitable. I was saying that machine learning, AI and automation were actually super underestimated in design. I remember talking to people afterwards, and half of them were laughing at me, and half of them didn’t quite understand what I was saying.
Now I think it’s happening.
I made the joke at the time that we really haven’t done ourselves any favors to avoid the machines taking over when we basically design the same stuff over and over. All that’s changed since then is that we just copy different things.
For example, every tech platform now looks like Linear. Obviously they all independently came up with this in their own group settings when they were innovating, and they all happened to ship it at the same time. Linear just got unlucky with the timing.
Even the reactionary movement to this sameness, the brutalist kind of ideas, in theory kind of works because you’re doing the opposite. But you can’t just add a negative sign in front of what we’re not happy with, because what happens is they also copy each other, and they also all look the same. Again, this mimetic desire is inescapable.
I think AI will essentially be a forcing function to upend some of our best practices. This stuff is not going anywhere, and we can already see from the very early concepts that it’s going to be a real game changer.
One example I showed was Galileo AI, founded by a few people from Facebook, where you describe an app UI you want, and it creates real Figma designs and code. You describe an onboarding flow with certain fields, and it gives you the UI. You describe a reading app featuring a certain author and a list of books, and the UI gets designed automatically.
I don’t think we need to panic about this, but I do think it’s worth acknowledging that we are not at the end. This stuff is coming, and we can’t be complacent and expect that nothing is going to change.
This is also the worst that AI is ever going to be. Don’t make the mistake of trying to cope with what the future might hold by looking only at what the tools can do right now. There was an article where an illustrator was showing their drawing next to an AI illustration and saying it was obviously a piece of junk, that it would never get better, and therefore it was impossible to create a professional and useful image from a text description.
That failed to see that progress is not linear. In that case, it was exponential. Six months later, you could do illustration, interior photorealistic shots, photorealistic portraits, grayscale sketches, and all this crazy stuff.
The point is that you have to think past the immediate and think about what the exponential version of what we’re looking at might be.
AI could be applied to almost everything: UI design, documentation, vectors, design to code, accessibility, video and audio generation, self-improving A/B tests that you can let loose, user testing, insights and summarizing things, design linting, translations, localizations, and all this sort of stuff.
The distance between design and production should be zero
That brings me to the part where I try to convince you to take the pink pill.
All this AI stuff is here. It’s getting tons of investment, and there are lots of very smart people working on it. It’s not all here right now, but it’s trending in that direction. In the meantime, this is some of the stuff I’ve been doing.
I believe the distance between design and production should actually be zero.
None of this “bridge the gap” stuff. We don’t need more red lines. We don’t need diagrams showing where the padding needs to go. Developers don’t like that stuff. Just give us the code. We don’t want to do that.
We should never send a human to do a machine’s job.
What I mean by that is not this idea of replacing jobs or taking over designers’ jobs. I don’t agree with that at all. I think AI and computers can be very complementary to what we need to do as creative people and as developers. We just want to take the parts that make sense for a computer to handle, and let humans do the creative side, working in a complementary fashion with each other.
This is the idea behind a lot of the Hypermatic plugins. Pitchdeck lets you turn Figma designs into presentations, or export them to PowerPoint, Keynote, Google Slides, PDF or a web presentation. The point is taking the same design from Figma and automatically doing the work that we want a machine to do instead. You can design in Figma, get the benefits of that, and send the result to people on your team who need PowerPoint or Google Slides so they can make content updates there.
Bannerify lets you animate and export production-ready banners from Figma to HTML, GIF or video formats. You can take your Figma frames, animate the banners directly in Figma, export them for platforms like Google Ads, and get production-ready code with the click tags, CSS, JavaScript, HTML and image assets. If you’ve ever worked on a banner campaign, you know they have about 7,000 rounds of changes. If you’re a developer on the other end of that, changing layers around, re-exporting assets and figuring out new CSS positions is a nightmare. The whole point is to make that iteration fast enough that you can animate, export, change and re-export a banner campaign in about 90 seconds.
Commentful is about supercharging Figma comments and gathering feedback from stakeholders without them needing to use Figma. You can take native Figma comments and put them into a custom board, create review links, send them to stakeholders with no accounts or logins, and collect comments directly on the design. The useful part is that feedback is context-specific to the design, so you don’t have to manually action all of it. You can apply text and image changes back into Figma instead of trawling through 50,000 Word documents with screenshots pasted into them.
Weblify lets you inspect Figma layers as HTML, Tailwind, React or Vue code with one click. If you’re a developer, you don’t want more handoff documents. You want to click the element and get the code, preview it, copy it, or download the assets.
Convertify lets you import and export designs into and out of Figma with one click, whether that’s Adobe XD, Illustrator, After Effects, Google Docs or other formats. If you need to keep working with clients or teams using other tools, the point is to avoid manually rebuilding everything just because the file format changed.
Emailify is the plugin I’m probably most proud of, because it solves the pain I hated most when I was working in agencies. It lets you design and export responsive, production-ready HTML emails from Figma. Every email marketing platform is a nightmare in its own special way, and this integrates with all of them, which is why a lot of my support comes through that plugin. But the whole idea is the same: build the email in Figma, preview the real HTML, export it, or upload it automatically to email platforms through their APIs, without rebuilding the design somewhere else.
Think different
To finish on the last shot of The Truman Show: best practices can be useful. We’ve touched on why they exist, why they can be useful, and why they’re so hard to change.
But they certainly aren’t always true.
I think we need to get back to being more on the side of questioning them, rather than blindly adopting them.
To take the famous Apple marketing line, we really need to think different.
Read as Markdown Related resource index Site index for AI agents
Pro Bundle