Ethereum co-founder Vitalik Buterin published a blog post titled Digital minimalism on Oct. 31, laying out a framework for using digital technology more deliberately. He described the idea as choosing how to use technology so each interaction is intentional, rather than drifting into products and interfaces built by tech companies.
The post centers on three experiments: posting through Firefly instead of using X directly, keeping his phone in airplane mode for 100 consecutive days while relying only on Wi‑Fi, and replacing some web-based actions with AI-written command-line tools.
Posting on X without using X itself
Buterin wrote that he had already embraced minimalism in physical life, noting that four years ago he described traveling with a single 40-liter backpack. In his view, that older form of minimalism addressed the challenges of the 2000s and 2010s. What is needed now, he said, is a new version shaped by digital systems.
On social media, he said X has long carried a toxic quality and that the problem has worsened in recent years. He wrote that from late 2024 into early 2025 he saw large volumes of pro-Hitler posts on X. At the same time, he said crypto Twitter was saturated with a kind of “financial nihilism,” the idea that blockchains are ultimately for gambling and that this should be celebrated.
Even so, he continued posting on X for most of this year. The difference was that he did it through Firefly, a social client developed by Mask Network. Firefly can display posts from X, Bluesky, Farcaster and Lens in one place, and a single post can be published across all four platforms. Buterin noted that Farcaster and Lens are decentralized social platforms built in part on Ethereum.
He said he originally switched to Firefly to support decentralized social networks and spread network effects across multiple platforms. After using it for several months, he found that his mood improved noticeably. He said the likely reason was simple: Firefly does not include X’s recommendation algorithm. As an example, he wrote that after opening one pro-Hitler post out of curiosity to read the replies, the algorithm treated that click as interest and began serving more of the same, first 10 posts, then 100.
Buterin added that the trend had cooled significantly over the past year and a half, and said part of that was due to improvements in X’s algorithm. Still, he wrote that when he logs into X now, he often signs out and returns to Firefly. He suggested that more people try reducing or constraining their social media use, whether by switching to Firefly or by limiting X to a fixed day such as Tuesday.
100 days in airplane mode, with Taipei at the top of his Wi‑Fi ranking
The second experiment kept his phone in airplane mode for 100 straight days. During that period, he used Wi‑Fi only and did not rely on mobile data. He said the reasons were privacy and mental health. A phone connected to a mobile network, he wrote, can be tracked continuously by location, and telecom operators have a poor record when it comes to storing that data. In some cases, he said, they even sell it themselves.
He argued that anonymous SIM cards are not enough to stop tracking. A phone’s IMEI can link usage records across the device, and movement patterns combined with browsing habits can be used to infer identity. Modern AI, he wrote, makes that process much easier. He also said he uses GrapheneOS and tightly restricts location permissions for apps.
There was also a psychological effect. Spending long stretches outside without internet access helped him rebuild the habit of living without constant connectivity, and he said it left him feeling calmer.
Buterin judged the experiment a success. He broke the rule only twice during the 100-day period. Once, he briefly enabled mobile connectivity to receive a verification code so his phone number would not be deleted. The second time came during a hike on a remote mountain trail, when friends insisted that he carry emergency communications equipment and check in every few hours.
The biggest inconvenience, he said, came in Singapore. He could not receive an SMS verification code, which meant he was unable to log into the ride-hailing app Grab and had to take the subway instead. After the experiment ended, he said he largely kept a “default airplane mode” lifestyle.
Without mobile data, public Wi‑Fi became essential, especially networks that could be joined without SMS verification. Based on places he has visited, Buterin ranked Taiwan, at least the Taipei area, first. He wrote that while walking in Taipei, or even while sitting in an Uber, there was more than a 25% chance that his phone could connect to a nearby hotspot, usually a free city government network or public Wi‑Fi from chain convenience stores.
His second tier included Ukraine, the United States, Canada and Singapore. Ukraine surprised him the most, he wrote, because multiple cities had large numbers of open hotspots. Austria and Hungary ranked lower because most public hotspots required SMS verification. Switzerland came last. He said that after walking 1.5 kilometers in a city center, he found only two hotspots that did not require SMS verification: one in a shoe store and one at a Starbucks. He noted the irony that Switzerland is known for privacy, yet finished at the bottom in his ranking based solely on whether a network could be accessed without SMS verification.
For navigation, he switched to CoMaps, an offline app using pre-downloaded OpenStreetMap data. His assessment was mixed: it was weaker than Google Maps for finding restaurants, stores and business hours, but better for hiking trails.
He also said he has been paying attention to offline knowledge apps. In his view, the amount of practical information needed to cover the world is not actually that large. He gave one example: the English Wikipedia, compressed and without images, is only 27 GB. He said he funded a bounty this month seeking mobile apps that can look up information offline.
From what he observed in the submissions, most entries relied on long reasoning runs by language models on the phone itself, taking anywhere from 1 to 22 minutes to answer a single question. That was too slow, he wrote. The best-performing app first used semantic search to pull results immediately from a local database, then passed them to a very small model to write the answer.
Replacing web interfaces with AI-written command-line tools
The third experiment focused on reducing dependence on websites. Buterin said that in his most recent ENS updates, he changed IPFS hashes through command-line tools written for him by AI. For payments through privacy protocols, he said he now mostly uses kohaku-cli, the command-line tool tied to Kohaku, the Ethereum Foundation’s privacy wallet project.
Earlier in October, he had publicly described a “paranoid mode” for AI agents built with zkAPI, Tor and local models. This new post follows the same direction: keep as much activity as possible on his own computer.
He listed three risks tied to web applications, with special attention to crypto-related ones. First, every time a user opens a website, the program is downloaded again from a server. That means trusting the server over and over, potentially thousands of times, and a single compromise can be enough to cause damage. By contrast, locally installed software requires trust at the moment of download.
Second, websites can observe browser version, operating system, screen size, mouse movement, click patterns and hundreds of other data points. He said that is often enough to identify a user quickly, and that the data is frequently shared with third parties. Third, commercial interfaces are not always designed in the user’s interest. Some details, he wrote, are dark patterns built to extract value from the user.
His alternative is a return to the Unix philosophy: many small tools, each doing one thing well, connected through text. With local AI added to that setup, users can tell the computer in English what they want done instead of writing Bash commands themselves. He argued that large language models can hand detailed control back to users without requiring everyone to become an engineer.
His closing argument for digital minimalism
Buterin concluded that computers and the internet made physical minimalism workable in the 2000s, while software advances over the past decade, combined with modern AI, now make digital minimalism workable as well. In his view, people can ask AI directly for information and depend less on social media than before. More tasks can be completed locally and offline, and AI makes command-line tools or custom interfaces easier to use. That, he wrote, makes the internet and the web less necessary than they once seemed.
He suggested that people try cutting “digital noise” from at least one important part of their lives, and continue if they like the result.
