Ethereum co-founder Vitalik Buterin has published a lengthy essay outlining a series of personal “digital minimalism” experiments, centered on three areas: cutting back on algorithm-driven social media, living with a phone that stayed off cellular data for 100 days, and reducing dependence on web front ends in favor of local tools and command-line interfaces.
Buterin describes digital minimalism as a deliberate way of controlling how one connects to digital technology, so that each use is a conscious choice rather than a default state. Physical minimalism still matters, he writes, but the current environment calls for a digital version of the same discipline.
Leaving X’s native interface behind
On social media, Buterin says Twitter, now X, has long had a toxic quality and that the feeling became stronger in recent years. For him, the low point came in late 2024 and early 2025. He writes that if someone spent a lot of time on X during that period, they would probably remember what he called an especially absurd phenomenon: it felt as if “everyone suddenly became a fan of Adolf Hitler.”
He adds that around the same time, Grok briefly went off the rails in what he describes as a “MechaHitler” episode, while “crypto Twitter” sank into a mood often described as “financial nihilism.” In his summary, that line of thinking treated gambling as blockchain’s natural endpoint and framed objections to it as elitist.
Buterin says he disagreed with both currents, but the larger issue was simpler: the platform’s style was highly engaging without being informative or emotionally healthy. That led him to run an experiment through most of 2026. He did not leave X entirely, but he stopped using X’s official client interface. Most of his posting moved to Firefly instead.
He describes Firefly as an all-in-one client that combines X, Bluesky, Farcaster and Lens, with the latter two partly based on Ethereum-powered decentralized social media. The app lets users read all four feeds in one place and publish a post across all four networks at once.
His original reason for using it was to support decentralized social media and spread network effects rather than reinforce lock-in around X. Over several months, though, he noticed something else: his mental state improved. The main reason, he suspects, is that Firefly lets him post and read posts without the core feature of X’s native interface — the recommendation algorithm.
To explain the effect, Buterin sketches out a loop. A user sees an extreme pro-Hitler post and clicks into it out of the same morbid curiosity that makes people stare at a car crash. The algorithm interprets that behavior as interest, then serves more of the same. One post turns into ten, then into hundreds.
He does note that over the past year and a half, that kind of extreme fandom on X has declined sharply. In his view, part of the reason is that such ideas burn out once more people see how immoral and foolish they are. Another part, he says, is that the recommendation system itself has improved.
Even so, he writes that when he occasionally opens X’s official interface, he quickly logs out and returns to Firefly. He also says he has become less interested in the endless disputes that dominate the platform. In one recent visit, he found the day’s hottest cultural argument was whether a historical figure was evil, then closed the app and moved on with his day.
His advice is straightforward: more people should try sharply reducing or tightly limiting their social media use, whether by using decentralized front ends such as Firefly or by restricting access to a set schedule, such as only using X on Tuesdays. The point, he says, is to capture useful information without getting pulled into recommendation loops that shape emotion and thought below the level of conscious awareness.
100 days with no cellular data
The second experiment focused on mobile connectivity. Buterin says he kept his phone in airplane mode for a full 100 days. During that period, he still used WiFi, but cut off all cellular network access.
He gives two reasons. The first is privacy. A phone that remains connected to cell towers at all times is, in his view, effectively a 24/7 personal location tracker. Telecom operators have a poor record on protecting user data, he writes, and in some cases have sold location information. That means a person’s movements can end up in the hands of intelligence agencies, military contractors, advertising firms and others.
Anonymous SIM cards are not enough, he argues, for two reasons. A device can still be linked through its IMEI over its lifetime, and detailed movement and browsing patterns can reveal identity on their own. Modern AI, he says, is already good at reconstructing the real person behind those patterns. His most complete answer for location privacy is simply not connecting to cellular networks at all. He also notes that he uses GrapheneOS and tightly restricts location permissions for apps.
The second reason is mental health. Spending extended periods, especially while outdoors, fully disconnected from the internet helps rebuild the habit of functioning offline, he writes. It also creates more calm and more room to think about healthier and more productive things.
By his account, the 100-day test was a success. He says he broke the rule literally once, when he briefly turned on connectivity to receive a confirmation text needed to maintain his phone plan. There was another case that did not violate the rule in a literal sense but did stretch it in spirit: during a hike in a remote and somewhat dangerous area, friends insisted that he carry an emergency communication device and check in every few hours.
The biggest inconvenience came during a stay in Singapore. Because he could not receive SMS verification codes, he was unable to log in to Grab, which he describes as similar to Uber in Southeast Asia. As a result, his dining options were more limited, and when public transit was inconvenient he still had to take the subway instead of a car.
After the formal 100-day experiment ended, Buterin says he occasionally turns cellular access back on, but still lives with airplane mode as the default most of the time.
Offline maps, local knowledge apps and small on-device models
For navigation outdoors, Buterin says he uses CoMaps, a local-first offline map app that downloads OpenStreetMap data in advance and then handles rendering and route planning on the phone itself.
His assessment is mixed. For commercial places such as restaurants, stores and business hours, he says it is weaker than Google Maps. For natural terrain and outdoor routes, he says it can be better, because it includes hiking trails that Google Maps does not know about.
It is not flawless. He recounts one case where the app marked a path that did not exist in reality, leaving him to push through several hundred meters of dense bushes before getting back to civilization.
That experience ties into a broader interest in local offline world-knowledge apps. Buterin argues that the amount of data needed to cover useful information is smaller than many people assume. A compressed plain-text version of the full English Wikipedia, without multimedia images, is only 27 GB, he writes. The global OpenStreetMap dataset, planet.osm, is about 2.2 TB, but if redundant information is removed and the data is compressed to 1% to 10% of its original size, it can still cover what a person is likely to need offline.
He also points to a developer bounty competition organized this month by Statuette, which he says he helped sponsor as a donor. The goal was to encourage developers to build mobile apps that run locally and offline while giving users practical knowledge on demand.
Among the entries, he highlights two apps in particular: FieldAtlas and Commonplace.
His broader takeaway is that offline tools are reaching a point where they are becoming genuinely useful. Most teams in the competition tried to run on-device large language models directly on phones and wait for the model to reason its way to an answer. He says that approach is too slow, with even simple questions taking 1 minute and in some cases as long as 22 minutes.
The stronger designs, in his view, combine semantic search with a lightweight language model. First, the app pulls precise results instantly from a local offline database. Then a small model generates a structured text answer. That combination, he says, can produce high-quality responses within a reasonable time.
Public WiFi varies sharply by location
Buterin also writes that when traveling, especially on days packed with meetings, some connectivity is still hard to avoid. On arrival, a person may need to contact someone to confirm the exact location or ask to be let in. At the end of one meeting, they may need to check the details of the next one. Without cellular data, public WiFi becomes the fallback, and the key requirement is that the hotspot must allow access without an SMS code.
Based on his travel experience while paying close attention to WiFi availability, he offers a rough ranking.
The top tier is Taiwan, at least Taipei. He says the city has an extremely dense public WiFi network, to the point that while walking on the street or even sitting in an Uber, the chance of finding and connecting to a hotspot at any given moment was above 25%. Those hotspots were usually provided free by municipal networks or convenience store chains.
The second tier includes Ukraine, the United States, Canada and Singapore. Ukraine surprised him the most, with a striking number of fully open free hotspots across multiple cities, some of them not even requiring a web portal step. In the US, Canada and Singapore, the total number of public hotspots is large, but many require SMS verification or mandatory login. Even so, he says, the number of fully open usable hotspots remains substantial.
He adds a side note from Toronto, where he says he came across a fraternity or sorority apparently named after SBF, or Sam Bankman-Fried, while searching for internet access.
The third tier is Austria and Hungary. He says the two halves of the former Austro-Hungarian region felt similar: plenty of hotspots, but most public networks required SMS verification. Open networks did exist, though there were times when he had to wander around trying many of them one by one.
At the bottom of the list is Switzerland. Despite the country’s reputation for privacy, Buterin ranks it last for public WiFi friendliness. He says almost every hotspot required SMS verification. Over the first 1.5 kilometers of a walk through the city center, he found only two exceptions: a random shoe store and Starbucks. Starbucks allowed 10 minutes of free access, after which continued use required a code sent by email.
Moving away from web front ends
The third branch of the experiment is what Buterin calls minimalism for internet applications. Recently, he says, he has been trying to stop relying on web user interfaces for more and more tasks.
For example, when updating IPFS hashes on ENS domains, he says he has recently done the work through a command-line tool written with help from AI. For many on-chain payments made through privacy protocols, he has also switched to kohaku-cli.
He shows the process of sending a transaction through kohaku-cli and says the tool has become simpler to use. Users no longer need to specify the same value twice or understand what a tail call is.
In his view, traditional websites and web apps, especially those used for crypto asset interactions, carry several structural risks.
- First is repeated trust and takeover risk. Every time a user opens a web app, the browser is effectively downloading the full application code from the server again. That means the user must trust the server every single time. If the server is hacked or tampered with during any one of those visits, asset security can fail. A native downloadable application is safer in this respect because the trust assumption is concentrated at the moment of installation.
- Second is privacy by default in the wrong direction. Websites and web apps can detect browser engine and version, operating system and screen resolution. Once a user starts interacting, they can also collect fine-grained data such as mouse movement and click frequency, building a digital fingerprint. Those data points and metadata may then be distributed to third parties ranging from CDN scripts to analytics firms.
- Third is the use of dark patterns shaped by commercial incentives. Buterin argues that many web apps built to monetize attention are not designed around the user’s best interests. Small interface decisions can be tuned to capture more attention. He contrasts that with what he calls “bright patterns” aimed at education, understanding and the protection of user interests.
His proposed alternative is to revive the Unix philosophy and combine it with local AI so that ordinary users can issue instructions in natural language instead of having to write Bash scripts themselves.
He describes the Unix philosophy in simple terms: build small, modular tools, let each one do one thing well, and have them communicate through the most universal interface available — plain text. That gives users fine-grained control over how tools are combined.
Modern web interfaces, by contrast, often bundle many functions into a closed black box controlled by a single development team. Taken to the extreme, he says, that trend produces super-apps, concentrating power in platform developers and increasing security risk through complexity.
Large language models, in his view, make it possible to return to the elegance of the Unix approach. Users can regain decision-making power without every user needing to become a professional programmer.
From physical minimalism to digital minimalism
Buterin closes by placing the experiment in a longer arc. The rise of physical minimalism in the 2000s and 2010s was made possible by computers and the internet, he writes. A lightweight laptop could replace dozens of physical tools, making it possible to live with little more than a computer, a few changes of clothes and basic charging and hygiene items.
Before the internet, physical minimalism was possible but came with much higher costs and trade-offs. Now, after a decade of software improvements and the recent surge in AI capabilities, he argues that the technical stack exists for a new wave: digital minimalism.
In that model, people no longer need to depend on social media in the same way because high-value information can come from local intelligent tools. They no longer need to depend on constant internet access because far more can be done in local and offline environments. And they no longer need to remain tied to web interfaces because command-line tools and custom interfaces can handle tasks more efficiently, with AI helping users operate or build those tools.
He argues that these technical conditions are maturing just as the weaknesses of the previous digital paradigm — along with secondary harms from misused AI — are reaching a peak. That makes digital minimalism feel more urgent.
Still, he does not present a single formula. The path will differ from person to person, depending on needs and pace of life. His final recommendation is to try clearing digital excess and mental clutter in at least one important area. If the result feels calmer and cleaner, keep going.
