Vitalik's Digital Minimalism Experiment: 100 Days Without Mobile Data, Goodbye to Algorithmic Recommendations, and Reclaiming Attention
Author: Vitalik Buterin
Compiled by: TechFlow
TechFlow Overview: For years, Ethereum founder Vitalik Buterin has been well-known for his minimalist approach to physical life: carrying only a few daily necessities in a single 40-liter backpack, constantly traveling around the world. However, as smartphones, algorithm-driven feeds, and pervasive instant messaging have thoroughly permeated modern daily life, physical minimalism is no longer sufficient to shield against complete cognitive overload.
In this latest personal long-form blog post, Vitalik rarely but systematically shares his hardcore self-experiment with "Digital Minimalism" conducted over the past few months. Most notably, he proactively cut off his phone's cellular mobile data network, relying solely on fixed-location WiFi connections to sustain normal travel and high-intensity work life for a full 100 days.
This blog post not only documents Vitalik's specific technical configurations for escaping algorithmic echo chambers, using the Firefly aggregated client, and exploring offline maps and local small models, but also profoundly analyzes from philosophical and cognitive science perspectives how modern individuals can reclaim attention sovereignty from the addiction to infinitely scrolling screens. TechFlow has fully compiled every section and paragraph of Vitalik's sincere long article, providing a highly inspirational guide to mental relief for every modern person anxious in the torrent of information.
The full text is as follows:
For a long time, I have always been a dedicated follower of physical life minimalism: minimizing the number of physical items I own and striving to do more with fewer resources. Four years ago, I wrote a blog post detailing exactly what was packed in my 40-liter backpack that I carried while traveling the world.
However, the concept of physical minimalism was born and matured between the early 20th century and the 2010s, responding to the challenges and opportunities of that era.
I still firmly believe in the value of physical minimalism, though I also agree with some contrary views, such as owning certain "appropriate types" of physical items actually allowing you to more tangibly control your autonomy over life. But in today's era, we need a new kind of minimalism even more: Digital Minimalism.
The core of digital minimalism lies in consciously and purposefully controlling how you connect with digital technology, making each use a deliberate, conscious choice rather than involuntarily sinking into it under default conditions.
In this article, I will share a series of real self-experiments I have conducted in this direction.
Minimalism on Social Media: Saying Goodbye to Algorithmic Recommendations and Reclaiming Attention Sovereignty
Twitter (now the X platform) has always carried a certain toxic hue, and this trend has intensified in recent years. This platform undoubtedly possesses strong interactive engagement, but it is often far from being informative, and the way it captures you rarely leaves you feeling mentally and physically at ease.
For me, this negative experience peaked between late 2024 and early 2025. Many things happened, but if you spent a lot of time on X, you likely remember one particularly absurd phenomenon: suddenly, it seemed like everyone on the platform became a fan of Adolf Hitler.
Fig: Screenshot captured by Vitalik showing algorithmic radicalization and abnormal Grok outputs on the X platform interface
This doesn't even account for the ridiculous incident where Grok itself briefly went out of control and transformed into "MechaHitler."
Almost simultaneously, "Crypto Twitter" fell into a collective mood widely described as "Financial Nihilism."
This argument roughly sums up to: the natural endpoint of blockchain is pure gambling, and this gambling is good; if you dislike this phenomenon and expect higher pursuits from blockchain, then you are a pretentious, elitist bastard who has lost touch with the masses.
Naturally, I have firmly disagreed with both trends intellectually in the past and present. But more importantly, I began to seriously reflect on a fact: although the discursive style on this platform is highly engaging, it is neither inspiring nor does it make me feel at ease.
Therefore, for most of this year, I conducted a self-experiment: keeping a moderate distance from this chaos. I did not completely leave the X platform, but I stepped away from X's official client interface. Instead, most of my tweets were posted through Firefly:
Fig: Interface of Firefly, the multi-client aggregated tweeting tool used by Vitalik
Firefly is an all-in-one client that integrates X, BlueSky, as well as Farcaster and Lens (the latter two are decentralized social media platforms partially built on Ethereum). You can view updates from all four platforms in a single interface, and posting a tweet will automatically sync across all four platforms.
My primary motivation for doing this was to support decentralized social media and, while continuing to participate in X interactions, diversify network effects rather than reinforce their ecosystem lock-in.
But during these past few months of using Firefly, I discovered it brought an unexpected effect: I felt much better psychologically when using it.
I speculate the core reason for this is: although Firefly allows you to tweet and view others' tweets, it lacks the soul/core of the X official interface: Recommendation Algorithms (The Algorithm™).
Specifically, I suspect that when I used the X official interface, I fell into a meticulously designed vicious cycle trap as follows:
First, you see someone post an extreme tweet supporting Hitler. With a morbid curiosity akin to witnessing a car crash, you click into this tweet, reading its quoted tweets and comments.
At this moment, the "recommendation algorithm" thinks: "This user seems interested in Hitler, let's mark this as one of his personal interest tags."
Next, you see ten tweets supporting Hitler in your feed. You click into them again with morbid curiosity, much like someone stopping to stare at ten cars colliding and burning for the first time in real life.
The algorithm thinks: I don't need to say it, everyone can guess what it's thinking. Subsequently, you will see hundreds of tweets supporting Hitler in your feed.
To be fair, giving X, Elon, and their hardworking employees their due credit, the extreme fan groups on the platform have indeed significantly subsided over the past year and a half.
I believe part of the reason is that once the public sees through the unethical and foolish nature of such extreme ideologies, they inevitably burn themselves out. But another reason is undoubtedly that the operational mechanisms of the recommendation algorithms themselves have been genuinely improved.
Nevertheless, even with that, whenever I am occasionally forced to reopen the X official interface, I find myself quickly choosing to log out and retreat back into the peaceful world of Firefly.
At the same time, I also realized a profound shift in my mindset: I am becoming increasingly uninterested in the various disputes that so-called "thought leaders" on the platform are deeply entangled in and fighting fiercely over every day.
A short while ago, I logged back in once and found that the hottest cultural debate in the feed that day was actually people fiercely arguing whether a certain historical figure was evil or not. Witnessing this, my inner experience was like an adult accidentally encountering a snack, TV show, or game they were once extremely obsessed with as a child, suddenly realizing they have absolutely no interest in it now. I smiled faintly, promptly exited the app, and continued enjoying my fulfilling day.
My conclusion and advice are: I recommend more people try this approach. Explore ways to greatly reduce or strictly limit your social media usage, such as using decentralized frontends like Firefly, or restricting yourself to using X only on Tuesdays, finding the pace that suits you best.
Gather truly valuable informational dividends from platforms, but never get swept into the vortex of algorithmic recommendations, because these algorithms are quietly manipulating your emotions and mind in powerful subconscious ways you haven't noticed.
Mobile Cellular Data Minimalism: A 100-Day Pure WiFi Survival Experiment
Over a full 100 days, I launched a second self-experiment: keeping my phone permanently in airplane mode. During this period, I still connected to WiFi, but completely severed all mobile cellular network connections.
I conducted this experiment primarily for two objectives:
First is privacy consideration. A phone constantly connected to mobile cellular base stations is essentially a 24/7 personal location tracker. Telecom operators' terrible track record in protecting user data security is widely known, and sometimes they even sell this location data themselves.
Thus, your actual movement trajectories could very likely ultimately fall into the hands of intelligence agencies, military contractors, advertising/marketing firms, and other parties. Purchasing anonymous SIM cards is not a sufficient solution for two reasons: first, your phone can be associated and locked throughout its lifecycle via a unique IMEI serial number; second, fine-grained movement trajectories and browsing patterns easily leak personal identity, and modern AI can already very efficiently reconstruct and infer the real individuals behind these patterns.
Therefore, the most thorough location privacy solution is simply not connecting to any cellular mobile networks at all (in addition, I have been consistently using the privacy-focused GrapheneOS system and strictly limiting location permissions granted to apps).
Second is mental health consideration. Completely disconnecting from the internet over a longer timeframe (especially when moving outdoors) is an excellent way to rebuild the habit of "surviving offline," bringing you greater inner peace and mental whitespace to contemplate other healthier, more productive things.
Overall, I believe this experiment was completely successful. Over these 100 days, I broke the "no cellular network" rule literally once, but not spiritually: I briefly turned on the network to receive a confirmation SMS necessary to maintain my phone plan.
The other instance involved a spiritual breach but no literal violation: during a hiking trip to a remote and slightly dangerous area, friends strongly insisted I carry emergency communication equipment they prepared, requiring me to check in every few hours.
The greatest inconvenience encountered throughout the experiment occurred during my stay in Singapore. Because I couldn't receive SMS verification codes, I was completely unable to log into the local ride-hailing app Grab (a Southeast Asian version of Uber).
As a result, my dining options were severely restricted, and even when public transportation was highly inconvenient, I could only take the subway instead of hailing a cab.
After completing this formal 100-day experiment, although I occasionally turn on mobile cellular connections, up until today, I have maintained the lifestyle habit of "defaulting to airplane mode" for the vast majority of the time.
For outdoor navigation, I use CoMaps. This is an offline map application adopting a Local-First architecture; it downloads OpenStreetMap data in advance, then completes map rendering entirely on the phone locally and plans your route for you.
My overall evaluation of CoMaps is: when exploring commercially built facilities constructed by human society (such as restaurants, shops, business hours, etc.), its performance is inferior to Google Maps; but when navigating natural landscapes and outdoor trails, it is even superior to Google Maps because it includes many hiking paths that Google Maps doesn't know about at all.
However, it occasionally malfunctions, marking non-existent wild trails on the map. Once, this led to me having to crawl through hundreds of meters of dense shrubbery with both hands and feet before clumsily returning to civilization.
During this period, I also developed a strong interest in the concept of "local offline world knowledge base applications." From a theoretical perspective, the volume of data packages covering globally valuable information is actually not as massive as imagined.
Compressed full-text English Wikipedia data (excluding multimedia images) is only 27 GB. The OpenStreetMap global dataset (planet.osm) is about 2.2 TB, but if redundant information is removed and it is cropped/compressed to 1% to 10% of its original size, it is sufficient to encompass all the information you might need while offline.
This month, a developer bounty competition was organized by Statuette (I participated as one of the sponsors), aimed at incentivizing developers to create mobile applications that can run offline locally on phones, allowing users to conveniently access practical knowledge anytime.
Below are two particularly outstanding offline applications among the submissions:
Fig: Interface of the offline map application FieldAtlas
Fig: Interface of the local offline personal knowledge base application Commonplace
My overall feeling is: we currently happen to be right at the critical inflection point where these offline tools begin to hold genuine practical value.
Most participants in the competition tried calling large language models directly on the mobile device, letting the model perform lengthy reasoning computations before giving an answer. But I think this approach is too slow: answering a simple question often takes 1 minute or even up to 22 minutes.
In contrast, the highest-performing applications cleverly combined semantic search technology: first instantly presenting precise retrieval results from the local offline database, then immediately invoking an extremely lightweight small language model to automatically generate structured text answers. This combination can produce exceptionally high-quality answers within extremely reasonable response times.
Finally, when you are out and about, especially with meeting schedules, it is indeed inevitably necessary to connect to the network occasionally while on the move.
When you arrive at a location, you often need to instantly contact the meeting person to confirm the exact spot or ask them to open the door; at the end of a meeting, you frequently need to verify the details of the next itinerary. If you lack cellular mobile data, public WiFi is your only lifeline, and crucially: this public WiFi hotspot must allow you to connect directly without needing an SMS verification code.
The friendliness of public network environments worldwide varies extremely drastically. Based on my travel records during the period I specifically focused on public WiFi availability, I have compiled a quick tiered ranking:
First Tier: Taiwan, China (at least in the Taipei area). Taipei boasts an extremely dense public WiFi hotspot network. When I strolled on the streets or even took out my phone while sitting in an Uber, the probability of successfully searching for and connecting to hotspots around me at any given moment surprisingly exceeded 25%. These hotspots are usually provided for free by municipal public networks or chain convenience stores scattered throughout the streets.
Second Tier: Ukraine, United States, Canada, Singapore. Among them, Ukraine's performance surprised me the most: in a staggering number of areas across multiple cities, there exist completely open free hotspots, a considerable portion of which don't even require webpage redirect verification.
The situation is similar in the US, Canada, and Singapore: the total number of theoretical public hotspots is huge, but a large portion requires SMS verification codes or mandatory login thresholds. Nevertheless, even so, the number of truly completely open and usable hotspots remains quite substantial.
Additionally, while looking for networks in Toronto, I also accidentally discovered that a local fraternity (or sorority) appeared to be named after SBF (Sam Bankman-Fried):
Fig: Vitalik encounters a network identifier named after SBF while searching for public WiFi in Toronto
Third Tier: Austria, Hungary. These two halves of the former Austro-Hungarian Empire territories are similar: there are plenty of hotspots, but the vast majority of public networks require sending an SMS verification code to log in. Although completely open hotspots can be found occasionally, several times I had to try countless networks one by one and linger on the streets for a long time.
Bottom Tier: Switzerland. Despite Switzerland's extremely high reputation for privacy protection, I rank it dead last in public WiFi friendliness. Almost all public hotspots here strictly require SMS verification codes. During my 1.5-kilometer walk through the city center, I only found two exceptions: a randomly passed shoe store, and Starbucks (Starbucks allows free connection for 10 minutes, and if you want to continue using it, you must enter the verification code received via email).
Web Application Minimalism: Saying Goodbye to Frontend Black Boxes, Embracing Unix Philosophy and Local Tools
Lately, I have started trying to completely eliminate reliance on web-based user interfaces for an increasing number of task processes.
For the last few updates to the IPFS hash values on ENS domains, I completed them directly using a command-line tool written for me by an AI. As for the many on-chain payment transfers I conduct through the Privacy Protocol, I have now fully switched to using the kohaku-cli command-line tool:
Fig: Terminal demonstration of sending privacy transactions via the kohaku-cli command-line tool in low-bandwidth and offline environments
The image above shows the process of sending a transaction via kohaku-cli. The command-line tool has now become much more concise and user-friendly; users no longer need to specify numerical values twice, nor do they need to understand what a tail call is.
In my opinion, traditional websites and web applications (especially those handling crypto asset interactions) have several fundamental hidden dangers:
First is repeated trust and hacker hijacking risks. Every time you visit a web application, your browser is essentially re-downloading the entire application code from the server thousands of times, meaning you must fully trust that server every single time you open the webpage. If that server suffers a hack or malicious tampering even once during your thousands of visits, your asset security could be completely compromised. In contrast, traditional native installation programs are safer: you only need to trust that the server hasn't been breached at that one precise moment of downloading and installing.
Second is default infringement of user privacy. Websites and web applications are extremely unfriendly to user privacy under default mechanisms. When you visit a webpage, it can easily detect your browser engine and version number, operating system, screen resolution; once you start interacting, hundreds of fine-grained data points such as mouse trajectories and button click frequencies are rapidly collected, effortlessly sketching your digital fingerprint. Moreover, webpages often default to distributing this data and metadata to countless third-party service providers ranging from CDN scripts to various analytics companies.
Third is commercial-interest-driven dark pattern design. As we see in social media, web applications carefully crafted by third parties with commercial monetization intentions often do not originate from the user's best interests. Many subtle interface interaction logics are deliberately designed as "Dark Patterns" to induce harvesting user attention, while what users truly desire are "Bright Patterns" aimed at enlightening cognition, providing education, and protecting personal rights.
Facing these drawbacks, what alternatives do we have? The core idea is: revitalize the classic Unix philosophy and leverage local AI to transform it into a more user-friendly and accessible tool for ordinary people: allowing you to directly give instructions to the computer in natural language, rather than being forced to write cumbersome Bash scripts.
The essence of the Unix philosophy is actually quite straightforward: build simple, modular tools, where each tool focuses exclusively on doing one thing well, and communicate/collaborate with each other through the most universal standard interface (i.e., plain text).
As a user, you possess extremely fine-grained autonomous choice rights, freely splicing and combining different tools. In contrast, modern web interfaces often force-countlessly complex functions into a closed black-box tool controlled by a single development team. This trend evolved to its extreme gave rise to so-called Super-Apps, which concentrate supreme power highly in the hands of platform developers, with internal operational mechanisms that are extremely complex and obscure, thereby exponentially amplifying system security risks.
Empowered by large language models, we can finally embrace the elegant philosophy of Unix once again: returning refined decision-making power back to ordinary users, while completely removing the high barrier that requires everyone to become a professional programmer.
Conclusion: From Physical Minimalism to Digital Minimalism, Regaining Your Own Focus and Freedom
The physical world minimalism that flourished from the 2000s to the 2010s became possible thanks to the popularization of computing and internet technology waves. A thin, light computer is enough to substitute dozens of physical tools. The result is that with just a laptop, a few sets of clothing, and basic charging and toiletries, you can live a highly free, extremely pure and simple life.
Before the advent of the internet, physical minimalism was feasible but often meant huge lifestyle compromises and high costs; the appearance of the internet completely and dramatically lowered the threshold for living a physically minimalist life.
With the significant progress in software development quality over the past decade and the explosion of modern AI technology, I believe we have now ushered in the underlying tech stack capable of similarly spawning the next wave of minimalism: Digital Minimalism.
We are no longer highly dependent on social media as in the past, because you can directly obtain high-value information from local intelligent bots (and for interpersonal emotional connection, the most classic and pure solution still exists: stepping out of the house and having face-to-face conversations with real people in real life);
We no longer rely on internet connectivity as continuously as we used to, because there are vastly more things that can be accomplished in purely local and offline environments;
We are no longer bound to be manipulated by web-based interfaces as before, because with command-line tools or customized interfaces, you can complete operations in a more convenient and efficient manner, with AI able to assist you in utilizing or customizing these tools at any time.
The maturation of all these technological conditions coincides precisely with the peak of the drawbacks of the previous generation's digital paradigm (as well as the secondary harms brought by misusing AI), making the practice of digital minimalism more urgent and necessary than ever before.
The path to digital minimalism varies from person to person, as everyone's actual needs and life rhythms differ.
But I wholeheartedly advise everyone to at least try clearing your digital redundancy and mental impurities in at least one important dimension. If you enjoy the refreshing calmness brought by this change, then steadily keep walking forward.