Jon Li

We did about 26 web3 projects so far, average length 6+ months. A lot of design teams will be like "Oh we did 5000 projects last week." But us, we always end up working with people for a long period of time, some even over 2 years. - @KuruExchange raised $13.6M and had a cumulative DEX volume of $2.42B (now they hit $3B recently) - @Globalcoinrsrch raised $2.2M - @ZetaMarkets had cumulative perps volume of $12.09B I love taking on an idea, genuinely improving it with the team and watching it grow.

A lot of teams think shipping with AI is as good as hiring a designer, because the output looks similar and it lands faster. But similar-looking output doesn't mean the decisions behind it are sound. A signup page can look finished and still leave someone unsure why they should hand over their details. The work is explaining why that happens, deciding what to change, and checking whether it worked. When the fast path fails, who can defend the choice and take responsibility for fixing it? That's what I'd ask before comparing speed.

I stopped bringing finished answers into client reviews. A finished deck invites reaction. An unfinished decision invites the team into the call with us. For launch work (DEX brand, event checkout, AI product site) we open the black box: the question, the evidence, what changed our mind. Clients who help choose stay bought in when the site goes live.

Most designers share their work with the eye of beauty. Is it aesthetic? Is it trendy? Do the colors match? That's why Dribbble was such an issue amongst designers. It looks fantastic, but it doesn't solve any business problem. Or the design is so interaction heavy it kills the desire for someone to make a purchase. Design is not art. It's a visual business solution.

We rebranded our studio. We were quoting higher fees, and the brand wasn't living up to them. My co-founder put it pretty honestly: "I always look at people's brand and websites and think they always look better than ours." The new lime: Mat went for something simple, but not your regular studio brand, and nothing too funky. Understated, premium, personality without being chaotic. The main font we went for was sherif. You don't see serifs much in web3 or tech, and it makes us look a bit more sophisticated. Serif for titles, sans serif for body so it's still easy to read. The logo came from the O. We didn't want a literal lime, so he tweaked the O to look like a lime, which makes it unique to us. Black and white, with lime as the only accent. We wanted it understated and premium, something bigger clients take seriously. It's live on limestud.io.

This is freakin insane.

Am the only one having issue with the latest update from @bot ? My whole bots list won't load or show up. It's just empty.

When AI makes craft abundant, conviction gets scarce. Anyone can ship a polished homepage for a crypto product or an AI tool this week. Buyers still need a studio that can defend "this is right for the user" with evidence. Tools create the possibility. Design creates the conviction. That's the work we do at Lime: brand and product systems founders can stand behind when a buyer lands, signs up, or clears KYC.

I've been running most of the studio on Grokbot for about a month now. It's pretty good at business organization. But design? It is a different story. The first version of a repo I built with it was just horrendous. Like, garbage. I don't even know why it was so bad. I think the advantage these days is to know what AI tools to use for what, and Grok is definitely not for design.

Ex-Head of Design at X, Benji Taylor, released a skill that helps your agent make better faux UI stills, animations, bento grids, and more. It is a really useful skill, but I started to see less and less of them every week. To understand why is that, we have to understand what a skill really is. A skill doesn't give the model a new "ability", it's basically a markdown file of instructions that tells the agent what good looks like, so the output gets closer to what you had in your head or what is considered "good" instead of whatever it spits out by default. For example, the model could already make a bento grid, Benji's skill tells it how to make a better one. The gap that skills fill is between what the model CAN do and what it does by default, and it's exactly what the researchers improve with every update. When enough people (or talented people) keep writing the same instructions, the labs notice and train it into the next model, and the skill just becomes the model's default. That's why the models are progressing so fast, they're training on our data, and why it's wroth using them now, because they're at the fringe of the future. Benji's skill go further than most though, it comes with ready-made frames, cursor animations, and a script that films the result and checks it for glitches. A model can learn better taste from instructions, but it can't learn a tool kit, so this one is genuinely useful for the time being. Put this into any agent: npx skills add benjitaylor/skills --skill stageshow

We recently pitched to be the head of design for TOKEN2049. My co-founder @matty_xyz asked me: "Aren't you scared to say the price?" Always. I'm always scared to be shut down or denied. I have my reasons behind the price, the only thing left is to hear what they have to say. The best we can do is try, even if it is scary.

Next year, the team behind Token2049 and SuperAI goes from 3 events to 5. Dubai in April. London in May. New York in June. Singapore in September, and again in October. Three of them back to back, that's never happened before. The way the team works today would never support that. So we're building the systems that will. 1. One place for brand asset. People keep asking for marketing assets, and people waste time looking for them. So we're building one source of truth. 2. A better way to brief a designer. Currently, it's basically a bunch of form fields, so whatever gets typed in goes straight into a ticket and then someone still has to manually claim it and assign it. The briefs most of the time come in without making any sense. 3. Event pages the teams can edit themselves. The hackathon and startup competition pages change non-stop. Right now every new headline is a ticket that goes to the front-end developers. There are two of them, for both brands. With their own seat in Webflow, the teams change the text themselves. We have November to March, quite a few months before the push starts. Building these will help a lot.

wow I am trippin'

We worked on @ZetaMarkets (now @Bulletxyz) from 0 to 1 and they became Solana's top perps DEXs. But the start was rocky. 2021. Zeta launched as an options platform on Solana. $8.5M raised, brand new, unproven. Options on-chain was still finding its footing. But crypto users weren't structured traders. They were speculators. Speculation is easier than any other form of trading, it's essentially gambling for most people. Options needs patience and sophistication. That's a hard sell to a market wired for fast moves, so they pivoted into perps in 2022. By 2023, the perps product was maturing and volume was building. By early 2024: 24x growth in MAUs. $3.4B in cumulative volume. Nearly half of Solana's perps market. Zeta is a good example that to become a top web3 product, it is hard to predict the market. You have to let the market tell you what it needs.

I will say this until my face turns blue. Less clicks does not equate to a better product. It does not make for a better experience. I will die on this hill.

Things design school never taught me: · How to say no to a client · When to stop iterating · How to present, not just design · Taste - and how to develop it · Reading a room before a pitch · The business logic behind every brief · When "good enough" is the right call

Last Friday HSBC launched RedCoin, a Hong Kong dollar stablecoin, inside PayMe. PayMe is basically Hong Kong's Venmo. That's 3.3 million people and over 100,000 merchants who can now pay in stablecoin, instantly. So what happens to the smaller stablecoin apps? My thoughts, and I might be wrong. The bank wins on users. PayMe already has a giant user base that's local to Hong Kong, where as a startup spends years trying to get that. But RedCoin doesn't leave HSBC. You can send it to another PayMe user, pay a merchant, or turn it back into Hong Kong dollars, that's about it. No outside wallets, no swapping it for USDC. So it's a stablecoin, but it behaves like a bank balance. The smaller apps are much more international. They're built to move across wallets, across borders, onto a card you can use anywhere. So I'd use the bank at home, and a stablecoin card when I travel. That's where the smaller apps still win. The day HSBC's own credit cards support stablecoins, then it will be a different conversation. Also, I'm not sold on the name RedCoin, maybe they're obligated to by regulation, but that word is still too closely attached to crypto (which has negative word association).

Is the disdain for "scroll jacking" real? I've long pondered on this question, and apparently there aren't many tests on this topic. So I thought why not make use of agents to sprawl the web and kind of do a pseudo-research? Admittedly, it's a bit long, but I found the results to be interesting. I'll definitely be doing this more in my work process moving forward. I definitely think agent research for design isn't talked about enough. scroll-jacking-research.verc…

"b-o-o" 👻 do you know? limestud.io

My co-founder @matty_xyz loves logos that are born from the wordmark. He's done it for @ZetaMarkets, @KuruExchange, HQ (accquiried by @gnosis_) Ask what the company actually is Find the letter that can carry it, usually the first one Make it work as a symbol on its own For example, @ZetaMarkets there's a Z in the negative space, and the arrows point up and down like candles on a chart. Or for HQ, a finance app that connects all your money stuff, one curved line flows from the Q into the H. Or even for our rebrand, we took the O from the brand and made it into a lime. A lot of logos now are just text. It's easy to add a symbol next to the name, but building it into the letters gives you something no other brand can use.