Some of it still true today.

But recently, things are changing. I was able to vibe code with Claude two versions of a simple NFT minting contract, which are, of course, dime a dozen in 2025, as compared to 2021, when it was harder to come by.

Deploying on Base is also significantly cheaper. Instead of $200 on ETH in the good old days, I spent about $1 each time. I was able to improve on the contract, as initially Claude made me a version that my contract pays for, and uses IPFS for storage. I realized that it means I'll have to maintain this forever, and it creates complicated debt that I wasn't willing to bear. Pinata, which I used for IPFS, had a severe file limitation so in less than a day with the current speed I'd be required to pay up $10/mo, potentially forever, for files in a fun side project that were just 400 bytes each.

So instead, Claude cooked a version 2 for me with improved requirements — this time, storage is not done with IPFS but directly onchain, and instead of the contract paying for minting (which was around $0.003 per mint but still require at least random checks once a month to see the balance and reload it as necessary), every user can now pay themselves.

However, I was dealing with SVG so closely for the first time, and so my PNG-to-SVG conversion meant that first NFTs were costing users $0.12, significantly more than before. It took all my prompting and available credits to redo the SVG design, and focus on minimizing everything that was extra. We done several rounds, reducing, reducing and reducing again to make SVG as compact as possible, even reducing extra spacing and such.

The new design, focused on minimalism, actually improved. Instead of decorative elements it's clear and purposeful. I think it turned out to be very nice indeed.

The result was an 84% reduction in code. The cost of each mint fell accordingly, from 12 cents to an average of 2 cents, a nice corresponding 84% cost reduction.

The system was ready, tested, and now producing live score collectibles.

And just like that, I went from a person who is scared of smart contracts and think smart contracts are too complex and too unpredictable to deploying and optimizing 2 smart contracts in under 24 hours to create a fully onchain scoring collectible that anyone can collect if they play Bitquiz on Farcaster.

PS: The app that I wrote about previously, has risen in ranks and now the #6 most popular app on Farcaster. I guess, you can just do things? ¯\ _(ツ)_/¯

About 10 days ago I started thinking of building a Farcaster mini app, after seeing early success of Bitquiz that was built a week prior (first commit happened on September 4). Using Claude Code, which had a "happy incident" of crapping out on me exactly around the start date, and Cursor, I committed myself to build a simple quiz app game that I could enjoy myself. Quick questions, fast answers, and the cycle of new questions every time to keep things fresh.

Vibecoding is a weird art form. It is very easy to start, but then to really plan around how it would look and how it would feel is hard. I use the term "cajoling" to describe the struggle of vibecoding and trying to convince Claude that things have to be certain way.

The unexpected benefit of vibecoding is that I get to try and test the app quickly. The ideas that I thought were good turn out to be bad once I get to play the app I've built. From the really-really old days of product planning back at earlier startups, writing documentation and spec was often done in a vacuum — you think something is good, but you don't know until you know. Vibecoding, surprisingly, fixes that. For example, I started the app with just 3 questions and thought that refreshing these questions every hour is good. It turned out to be a bad idea, because it's easy to get bored and then forget to try the app again. Instead, now it is 10 questions and you can play anytime, questions will be reshuffled. Many such decisions were made, then adjusted, then improved or discarded. Overall, it feels like I am on a major version 7 or 8 right now, since how the scoring works, how the prizes work, and so on differ dramatically.

After seeing early success between September 5 and 12, when a few of my friends played the app, I wanted to embark on a journey to build a mini app on Farcaster. The idea is that it could help distribution, as I've seen and used other apps that became popular there. However, getting with a simple initial version took longer than expected, about 3 days. I fought with mini app specifications, I fought especially long with integrating notifications (they are very simple still), and started to feel the limitation of my knowledge to cajole Claude into making workable integrations between various APIs and services.

While vibecoding and seeing Claude effortlessly generating tons of code, I have a sinking feeling that tech debt is the real friends we make along the way on our journey of vibecoding. There's so much code that doesn't work. There's so much legacy code from rewrites that just happen because Claude felt adventurous. This is probably the area where humans are different from AI for now. Best developers I know they are like chefs — they clean as they go. They write the initial code, and then they allocate time to clean and rewrite the code once more. It is meditative, but vibecoding doesn't have time for that.

The true peak of Bitquiz popularity came when with some help, I integrated native Farcaster wallet payments. Users can get rewards, ranging from several cents to several dollars and more for being the best at a game. I funded the prizes, and spent good amount on getting people into the game. Lots of players play daily! For a game, I have to frequently tweak the scoring algorithm or introduce new ways to score. For example, the latest feature is to invite a friend with a referral link to earn higher score.

Well, the journey is only 3 weeks long so far. I gave myself some time to take this app to a logical conclusion, so there's a few more tweaks and features I would love to introduce for the users. In the meantime, give Bitquiz a try and stack your prize Bitcoin. Maybe, finally, your obscure knowledge of culture, geography and history can come in handy!

From inside the crypto industry, it is clear that smart wallets are developing very quickly. When I first tried Argent, arguably a smart wallet due to multi-sig functionality, a few years back, I had to pay over US$200 to set it up. Gas fees were high and we were clearly in ETH-to-the-moon euphoria. Obviously, a normal human being is not going to pay $200 for a smart wallet, so a lot of work has to be done.

And so it was done. 

Smart wallets and major crypto activity migrated to L2s, becoming faster and more reliable. It is now possible to transact for 1 cent when the same transaction a few years back would cost $100, a very solid 10,000x improvement! With various improvements such as EIP-3074 for gas sponsorship, and many others, wallets became even more seamless and easier to use. That’s the technology we employ at Moai as well — making the wallet truly free to use, where the transaction fees are covered, and implementing various safety features that make onboarding about 10x faster than when you had to write down the seed phrase before you’d even know what you need it for.

But even more needs to be done.

Traditional (and I mean leather and textile here) wallets in themselves are quite dumb invention. A traditional wallet carries money, cards, or photographs (are these NFTs?!) and protects them from the elements, but does nothing beyond that. A smart wallet can and should do a lot more, because it is programmable in nature. I put these features in three broad categories: Opportunities, Social Recovery, and Protection.

Opportunities. A smart wallet should do more than just store money or be able to swap tokens. It should nudge you intelligently if the token you are holding is ripping or tanking, whether your USDC sitting idle can earn you some interest in liquid staking (or better yet just do it automatically if permitted… free money!), provide the option to participate in prediction markets, staking, and more, and overall help you grow your capital or invest it, wisely or unwisely, with your permission. Imagine a wallet that can employ an AI agent to take a portion of your portfolio to buy Base memecoins with specific parameters. If your risk is low, maybe you can allocate 5% of capital to this. If you are adventurous, increase allocation higher. Same with vaulting prized assets. If there are some assets that appreciated considerably, a smart wallet can automatically move them to your designated cold storage. 

Social Recovery. This has been talked about for many years by many, including by Vitalik in his recent post, but basically, accidents happen. It is important your wallet can be recovered and protected, with some social interaction. Maybe even between your crypto friends — i.e. if you swap or share money with certain people over and over again (real people, your other wallets), several of them can form a weak bond that can vouch for you to recover access if lost. Weak bond would mean that if you send money to one wallet, this wallet cannot singlehandedly sign a transaction to lock/unlock the wallet, but a group of several such parties can nudge a wallet in the right direction. Specific parameters can be set for people who are highly technical, and “easy mode” can be implemented for most.

Protection. Lastly, it’s not a secret that the way blockchain is designed is becoming untenable for many. The ability to spike tokens or NFTs to with malicious code makes transacting with them dangerous or being able to see every single transaction should be considered a bug. Privacy and provenance should go hand in hand. While various blockchain foundations are reworking the basics, the smart wallets can help temporarily by providing protection. For well-known malicious tokens, these should be isolated and removed from a wallet (many are already doing that part), for the spammy tokens, i.e. the ones that many active users are airdropped, while there are no protections on a blockchain level, these should be sent to “trash” addresses. Same for NFTs, anything that you didn’t ask for should be sent back, preferably with a note “don’t ever send this again”, and a smart wallet can have several levels of action that can be taken to protect the tidiness and integrity of a wallet. 

Some of these measures might take weeks to develop, some months. Some ideally require foundational blockchain changes, such as moving to zero-knowledge proofs or mixers being built right into every transaction, which might take years at the current stage. However, safe, secure, and smart wallets will make people richer and their lives easier and more fulfilling. 

Ethereum was riding high — at $3,615, and so was the gas. Today, the anniversary is marked by the largest and most anticipated event in crypto — Ethereum is switching to proof of stake on a 1 year mark of the first transaction on Sloika. Talk about coincidence! 

Everything gains clarity in retrospect, and it is incredible that in year we helped onboard 400 artists — including up-and-coming legends and world renowned photographers such as Iurie Belegurschi, Daniel Kordan, Stian Klo, Jordan Banks, Eric Pare, Justin Lim, Subodh Shetty, Hobopeeba, Oprisco, Tara Workman, half a dozen collectives, and so many more. 

As a photography first and (at a time) the only marketplace, we helped pioneer several new innovations that now became standard across the thriving industry, bear market be damned.

Photography rights are one of these. When we started, marketplaces insisted, deep in the terms of service, that all you get in exchange for your payment is a blockchain record itself and not the underlying art. We helped make it transparent, by putting rights front and centre on every photo asset, and embedding them on-chain. We even collaborated with a bunch of photographers and collector in real time to improve and expand it to match the evolution of the space.

We also helped pioneer more gas efficient contracts. From November of last year to September of this year, creators paid up to 1000 times less to mint their photos, with an average reduction of about 350 times. Till today, minting an edition or a photo series on Sloika will result in industry’s lowest gas fee. In the end, photographers are the ones who benefited the most from this. 

There are so many more things that we tirelessly worked on to bring into the world, such as Collectives, which allow a group of strangers to create cohesive works of art and share proceeds in a decentralized and trustless manner. Or take GACHA, our innovative on-chain randomizer that picks a random art piece from the series, turning collecting into a fun game. Or artist-owned smart contracts from day one, at a time when none of the marketplaces were doing it, because this was a controversial and overly technical solution. Now, it’s the standard as well. Or our work with enabling listing liquidity, allowing collectors to save money when listing on OpenSea, X2Y2, LooksRare, Rarible, Coinbase NFT, and soon, Sloika. 

To mark this anniversary, I also wanted to take a moment to say thank you to the first collectors who helped spark the revolution that is underway with Sloika. Thank you Daniel McAuley, Justin “hurtingbombz” Lim, Iurie Belegurschi, Chris Ye, octo9488, torman, 8F44F7, EDFDF4, 8c9c716018710c30ebb038da8d9eb4b3b8446943, and freeatnet.

Here’s to the future.