Raine Luntta — Notes

Down the Color Vision Rabbit Hole

2026-05-27T00:00:00.000Z

Ever wondered how people with Color Vision Deficiency (CVD) experience your products and services, whether digital or physical?

Back in 2019 I wrote a small color library to help answer exactly that. It could simulate how people with different types of CVD see a given color, and it had the usual grab-bag of color transformation features you'll find in plenty of similar libraries. It grew out of some color manipulation expertise I'd picked up earlier in my career doing accessibility work.

The reason it existed was simple. I wanted to learn how different color spaces work, how to transform between them, and how the CVD math actually works under the hood. It was a rabbit hole, and I thoroughly enjoyed falling into it.

Over the years it quietly picked up a few thousand downloads on npm. Nothing grand, but more than I expected from a purely exploratory project. Most of those are probably bots, but I like to think at least a few came from actual humans.

Picking it back up

I recently went back to that old code, and a couple of things had changed since 2019.

First, AI coding tools have become genuinely useful (don't worry, the library still has plenty of good old artisan code in it too). Second, we got some lovely new color spaces in the meantime. OKLab and OKLCH both landed in 2020, and they're a real step up for this kind of work.

So I set out to finish the features I never got around to back in 2019, and to build a little playground so people could actually poke at the thing instead of squinting at API docs.

"Fixing a few things and adding a couple of features" quickly turned into a near-total rewrite of the library's core. At that point it didn't really feel like the same project anymore, so I retired the old library and gave it a new name: Swatch.

Have a look

You can try Swatch on its website, read the source on GitHub, or install it from npm.

I'd genuinely love to hear what you think. One caveat, though: I mostly build this to satisfy my own curiosity about color and color vision, not to dethrone the big established color libraries. So if you spot a bug in my math (or my AI agents' math) please let me know. And if you get some actual use out of it, I'd love to hear that too.

A Radar for the Shadow Fleet

2026-05-13T00:00:00.000Z

Ever wondered how openly Russia dodges sanctions, right under our noses?

There's been a lot of talk in the media about Russia's shadow fleet. Experts have been rightly worried about how Russia uses it to get around sanctions, and especially about the fact that these ships, often aging and poorly maintained tankers, haul enormous volumes of oil across the Baltic Sea.

Looking for myself

Last week I got curious enough to dig in. I combined AIS data, the position signals that ships broadcast over the Automatic Identification System, from Digitraffic, Fintraffic's open real-time traffic feed, with the available sanctions data.

The goal was a more realistic picture of two things: how many shadow-fleet vessels are actually moving around the Baltic, and how close they pass to the critical infrastructure on our seabed, meaning the submarine cables and gas pipelines.

What it shows

The result turned out interesting enough that I decided to publish it: a little map I'm calling Varjolaivasto-tutka, a "shadow fleet radar."

The short version is that a surprising number of sanctioned tankers (the ones highlighted in red) are moving through the Gulf of Finland at any given moment, and quite a few of them sailing very close to our critical cables and pipelines.

I know, and you should know too, that this is a demo built on open data, not an intelligence product. AIS can be switched off or spoofed, and sanctions lists are a moving target, so treat it as a rough picture rather than the truest truth.

Where it goes next

I already have a backlog of ideas for where to take this. But I'd genuinely love to hear what you think, what's useful, what's missing, and how you'd develop it further.

From open-source to partially open-source

2026-03-01T00:00:00.000Z

There have been many viable business models for open-source Software (OSS) for decades that has allowed the copyright holders and others to a build business around the OSS, while allowing the public to have unlimited access to the source code and everything that comes with it (like tests). It has a been a huge blessing for humanity. World runs on open-source and it's not often that you can give something away for totally free, while having a thriving business for that very same thing.

Well, Cloudflare has just announced that it has rebuilt Next.js with AI in one week and I think it's about to change open-source forever for the worse.

Most of the open-source projects society is most dependent on have 100% test coverage and extensive documentation. That means the test suite is basically a big ole' validation spec sheet for the software. What Clouflare did was they basically told an AI agent to recreate Next.js and validate it against the Next.js test suite. If a test passed the code was merged.

Now, let's say you are a creator, a maintainer or a copyright holder of an OSS and you also run a business around that OSS. Your risk management has basically leaned on the fact that your OSS is so popular and complex that it just isn't worth the hassle for anyone trying to recreate it. Its easier for everyone to just use the same OSS and contribute to it. Everyone wins, right? What Cloudflare just did was the total annihilation of that approach to your business risk management. Now anyone (with enough money to run Opus 4.6 for a week straight) can yoink your test suite and recreate the software you've spent thousands of hours and millions of dollars to create.

I'm betting there's quite a few OSS business owners scrambling for a viable risk management measures to mitigate this risk at this very moment. I think what we're going to see in the coming days, weeks and months is many of them removing tests from their repositories. If that happens, it's going to be a sad day for OSS. Transparency and openess suffers when contributors need to be part of some elite inner-circle to get access to the tests. Partially open-source just doesn't have the same ring to it as fully open-source.

The silver lining is that we're probably going to rid ourselves of the tyranny of some of the legacy OSS maintainers by building new and modern rivalling solutions to the same problem, but at the same time I dread the fragmentation of the OSS community.

Product Design After the AI Panic

2026-02-26T00:00:00.000Z

I'm a design leader at a company that invests in and values good design. I spend a lot of time making decisions about strategy, hiring, and the kind of design work we invest in. I've also spent my fair share of time getting to know the latest AI tools and technologies.

I'm writing this because I spent the past few days sick in bed doom-scrolling social media. The public conversation about AI in product design (or UX and UI design) seems to often be split into two camps:

panic and doom
denial and copium

On one side, there are the doomers: design is dead and we're all redundant, learn to code asap (or don't, since that's dying too). On the other side, there's the copers: AI is just a tool and everything will be fine if we just focus on "human-centered design." I think both positions are a bit lazy. The reality, like always, is not black and white.

These are my current thoughts. My opinions are subject to change, since the world around me changes constantly. I'm aware of the environmental impact of AI and the net losses these AI companies are currently running. My feelings are conflicted, but I wouldn't be doing my job if I didn't analyze the potential.

AI as design material

Designers have always needed to understand the material they work with. A product designer who doesn't understand engineering constraints designs fantasies. An industrial designer who doesn't understand manufacturing produces beautiful things that can't be built. Service designers need to understand customers, stakeholders, and the service landscape.

AI is now part of the material we work with and design for. It comes with unique possibilities and unique constraints.

What I mean by that is that designers should understand the technology well enough to make informed design decisions and use it effectively. Can an LLM reliably fetch structured data from messy documents? What happens when you put an AI agent into a customer-facing service (I bet we all have some experience already)? Where does this shiny new RAG make some features suddenly viable?

If we don't understand these things, we're designing blind. We're either: under-designing and missing opportunities the technology makes possible, or over-designing and proposing stuff that fails in the most spectacular ways.

My driving force has always been curiosity (lame and cringe, I know, but it's true and I don't care). I think people who are genuinely curious about new tools, who experiment a lot, and who understand the world around them are usually making better design decisions. Not because they're "AI designers" or prompt engineers (which is dead too, btw), but because they understand the material they're shaping and working with. I think that's what has always been expected from great designers.

Exploring this type of new material has always been one of the more fun parts of design for me. Solving real problems while learning new things. And the people I see actively engaging with these new tools and technologies seem to be having a much better time than the ones anxiously refreshing LinkedIn and X to find out if their career is over. And you know what? It's just more fun to be competent.

No one is going to inject this competence in you with a needle. Your employer might (and should) support you in developing your competence. Some professional communities might host some AI workshops. But in the end, the responsibility of your own competence is yours. When the internet happened, no one waited for permission to learn CSS. When mobile happened, the designers who thrived were the ones who started before anyone told them to. This is no different.

Product Design: a stormy sea

I'm trying to be honest and realistic about what's happening in digital product design, because this is where I can see the most visible and severe changes. I don't want to underestimate or undersell the disruption it has in service design, strategic design, business design etc, but I want to focus on product design (including UX and UI design) specifically.

Current AI tools can generate functional user interfaces from text prompts. They can produce variations, translate designs into working code, and handle a growing share of the work that used to fill designers' and engineers' weeks. Tools like Claude Code (with or without Figma MCP), and OpenAI Codex are not theoretical. They work today at a level that would have been unthinkable a few years ago. Take a look at this Batch Image Editor or this Marey's Train Graph I recently built with Opus 4.6. They're not just some generic bullshit form-based web UIs. They're custom (and niche) WebGL and WebGPU UIs built with just a few prompts. The image editor could be deployed as an actual product somewhere and someone would probably pay for it. The train graph is just an exploration of an idea and not even meant to be prod ready, but it could be worked into a working product easily.

The quality of the outputs is rising fast. A year or two ago, AI-generated UI was total garbage. I know, because I tried. Today, with the right models, prompting, and iteration, you can get output that's close to production-ready and sometimes fully production-ready. Even in environments where shipping AI-generated design and code is, questionable at best or straight up forbidden (like safety critical software with a ton of risks from Human Factors), AI can still help you build inspiration, concepts and PoCs in minutes instead of hours. That leaves you with more time to think, understand, explore, research and validate.

To put it plainly: if the current trend continues, a significant portion of UI work will become automated. It's already happening.

Not all hope is lost. What AI can't yet do is the first half of design work. AI can't understand the why behind your design decisions. It can't navigate the tradeoff between onboarding simplicity and power-user efficiency when both matter and you can't have both. It doesn't know your exact user context, trust issues, regulatory constraints, or why a similar feature failed last time. It doesn't understand the Human Factor issues that rise with specific circumstances. It produces plausible interfaces based on the information it gets.

Last mile of customer and user experience is another thing AI can't truly understand. Some say AI doesn't have taste, but I disagree. AI does have taste, and often better taste than a designer, but AI doesn't truly care. You have to do that. Looks and features get users, but that extra bit of care you put in that shows the users they're cared for is not something AI can do. AI can guess, but it doesn't have that human connection. I'm not saying you can't prompt an AI tool to put in that last bit of care, the last bits of delightful interaction that makes or breakes the product, but you are the one who needs to care first.

The gap between an "interface that looks good and right" and a "product that works great" is where product designers either prove their value or they don't. In the end, pixel-perfect interface design is just an artifact. If your value begins and ends with the pixels, then yeah, you might be in trouble.

And yes. The market will be filled to the brim with products that suck to use. Use that to your advantage, don't just dread it.

The uncomfortable truth

If you're reading this rambling as a product designer, I want to speak directly to you, as a fellow designer and as someone who leads designers.

Some of what we're protecting isn't worth protecting.

Not every design task is meaningful. Some of what we call "craft" is really just slow production. Some of what we call "process" is just a ritual that doesn't generate new insight or value. Some gatekeeping around who gets to make design decisions is less about quality and more about control. AI, similar to any other big tech disruptions, is forcing an honest discussion about which parts of design work actually create value, and which parts we defend because they are familiar and because they are ours.

I get it, it's uncomfortable for me as well. And I know I don't have all the right answers. But I think it's healthy to give up something once in a while.

From my perspective, the design tasks that AI can automate are often the ones that bored our best people anyway. Automation almost always starts where it's easiest to apply. The same is true with AI: it's exposing how much of the work isn't that meaningful anymore.

The real question isn't "how do we protect design jobs." It's "where design actually matters", and are we capable of delivering the value in this new world. If AI accelerates the production layer, that's an opportunity to align talent toward the work that makes a genuine difference. The problem-framing, the systems thinking, the organizational navigation, the strategic sense-making. The work most design teams never had enough time for because everyone was too busy pushing pixels.

I can't force you to be curious. I can't make you spend a Saturday experimenting with a language model to understand what it's good at. I can't make you read a technical paper about RAG because you want to know how it might change the product or service you're designing. That drive has to come from you.

If you treat this moment as someone else's problem to solve, if you're waiting for your manager, your company, or your professional network to tell you what to do, you will probably fall behind. Not because AI replaced you, but because you are opting out of your own relevance as a designer. I know it's harder to hear than "AI is taking our jobs", but I think it's unfortunately closer to the truth. And yeah, maybe this is just a bubble that's about to burst, a fad that kind of slows down in a few years, but the technology and capability is here to stay.

And I get it, I truly get it. The world is on fire right now. You're overworked or unemployed and just trying to make ends meet. Your family, little kids, and aging relatives leave you with little time for yourself and self-development. You have better things to do in life than to learn some new tricks to boil the ocean.

But still, endless hunger for learning, and curiosity towards the unknown makes you a better designer. It opens up possibilities you couldn't see before. Positive attitude for change makes the work more fun. The best people I work with aren't anxious about new things or change. They're intimidated, sure, but also excited in the way that competent people get excited when their field suddenly gets more complex and more interesting.

Don't outsource critical thinking

One last thing I wanted to touch is that you shouldn't outsource critical thinking. I've already seen some terminal cases, not just in design, but in general, who outsource all their thinking to their dear LLM models. It's insufferable. Don't do this. While you can outsource many of your menial tasks to it, as can be seen in product design already, you shouldn't give up critical thinking.

"Redesigning the design department"

2022-05-06T00:00:00.000Z

Latest McKinsey report "Redesigning the design department" was a joy to read and validated a lot of my thoughts about design departments. Couple of key findings:

Designers should fully integrate into their project, product and service teams as peers rather than act as service providers. The design department is there to provide support, training and mentorship for the designer. The project team is there to provide purpose and possibilities. This cross-functional organization maximizes the surface area of design across the business and enables greatest impact. Centralized design department is a breeding ground for scaling, accountability and trust issues within the organization.
Old schoolish (e.g. non-agile) organizational structure doesn't mean that cross-functioning project, product and service teams are out of your reach. You can unlock design's potential without reorganizing the whole company. Don't look for excuses, just do it.
Designers should also design the business, not just the products and services - and leadership should be calling that out to the business. As a designer, building the trust of business people from the bottom up takes too long and is risky. People move around and that trust gets reset too easily. That's why design needs a seat at the executive team - to get CEO, CFO, CHRO etc. on board.
Designers should do more than design - although I think this applies to every member of the organization. Understandment, integration and cross-functional expertise of the business and its functions is at the core of making better decisions and attracting and retaining talent. "You have to have an equal position with the other functions before you've earned the right to challenge anyone."

Feels nice to get some outside validation for our efforts. Let's just stay on track now.

"We used to tell the designers what to do, now they're showing us what is possible — and it's so much better."