I think you should have a website. I also think you should have a blog if you have things you want to write about, but the website is the broader argument here. Put your projects on it. Put your writing on it. Put photographs, links, notes, music, stupid experiments, things you made five years ago and no longer particularly like, or just your name and a paragraph saying hello. It does not need to become a portfolio, a business or a personal brand. It can just be yours.
I started blogging in November 2022, when I was seventeen. The first version of my actual personal website came a few months later, on 31 May 2023, using Neocities. I did not have a domain, I did not have a deployment pipeline and I certainly did not have the increasingly elaborate website I have now. I had some files, somewhere to put them, and the growing realisation that the browser would let me do considerably more than the social platforms I had spent most of my life using.
That is important to how I think about all of this because I am not nostalgic for some version of the internet I remember from the 1990s or early 2000s. I wasn't there for it. I grew up in the boxed-in, containerised social media era, where being a person online generally meant creating an account inside somebody else's application and filling in the spaces they gave you.
You get an avatar. You get a display name. You get a bio of a specified length, perhaps a header image, and then your posts go in the same vertical column as everybody else's. The actual contents can be wildly different, but the shape of your existence is decided in advance.
For most of my childhood and adolescence, that was simply what the internet looked like, and I think that is fairly normal for people my age. Saying that I have a personal website still occasionally gets a more confused reaction than saying I have an account on basically any social network. Why would I need one? I already have profiles.
A friend once asked me almost exactly that: why do you even have a website?
The question makes complete sense if your model of being online begins with platforms. Why maintain a whole website when Instagram gives you a profile, TikTok gives you a profile, Bluesky gives you a profile, GitHub gives you a profile and everything you actually want to post can presumably be squeezed into one of those?
I think the squeezing is the problem.
I grew up inside the containers
There is an old shorthand for the handful of companies which came to dominate the modern technology industry: Alphabet, Amazon, Apple, Meta and Microsoft. Calling those the five companies which control the internet would obviously be too neat. They compete with one another, they do not control every part of the web, and the list already looks incomplete in 2026 because ByteDance owns one of the largest social platforms on Earth.
The European Union's Digital Markets Act reflects that rather nicely. Its original list of designated gatekeepers was not five companies but six: Alphabet, Amazon, Apple, ByteDance, Meta and Microsoft.
So no, there is not literally a committee of five people deciding what the internet is allowed to look like.
What is striking is how often the same names appear once you look at the individual layers through which we use it.
Ofcom's 2025 Online Nation report found that Meta and Alphabet alone accounted for 51% of the measured time UK adults spent online in May 2025. More than half of the average adult's measured internet time, across smartphones, tablets and computers, was going to services controlled by two companies.
That does not mean everybody was literally sitting on Facebook and Google Search for half the day. Alphabet includes YouTube; Meta includes Instagram, WhatsApp and Facebook. That is sort of the point. These companies have accumulated multiple enormous containers, and those containers now make up a substantial part of what ordinary internet use means.
Then there is finding anything outside them.
In January 2026, the UK's Competition and Markets Authority said Google accounted for more than 90% of general search queries in the UK. Search is one of the main routes from one part of the web to another, which gives the company operating it an extraordinary amount of influence over what gets discovered.
Before you search the web, though, you have to get to it.
Apple and Google effectively own that layer too. The CMA's 2025 investigation found that around 90–100% of UK mobile devices run on an Apple or Google mobile platform. Those platforms include not only iOS and Android but app distribution, browsers and browser engines.
An earlier CMA study put Chrome and Safari at around 90% of UK mobile-browser use between them and found that at least 97% of UK mobile browsing in 2020 ran on browser engines controlled by Apple or Google. On iOS in particular, Apple historically required competing browsers to use its WebKit engine, meaning that downloading another browser did not necessarily mean choosing an independently implemented browser engine underneath it.
That is not an abstract argument about corporate bigness. It changes what software can exist.
The CMA's final investigation into mobile browsers concluded in 2025 that the market was not working well and was holding back innovation. Developers had specifically raised progressive web apps as something whose capabilities were constrained by platform rules.
If you are building the application rather than using it, there is another concentrated layer waiting underneath.
Amazon Web Services and Microsoft are the two largest cloud-infrastructure providers in the UK. Ofcom estimated that they had a combined market share of 60–70%, with Google a smaller but still significant third provider. The concerns there will be familiar to anybody who has ever tried to move infrastructure between providers: data-transfer charges, committed-spend arrangements and technical interoperability restrictions can make leaving considerably harder than arriving.
This is the internet I mean when I call the one I grew up with containerised.
It is not just that five or six companies own popular websites. The same small set of companies keep appearing at different levels of the stack. They own social spaces where we spend enormous amounts of time, the search engine through which we find the wider web, the operating systems and app stores through which software reaches our phones, the browsers through which we render websites and a substantial chunk of the cloud infrastructure on which those websites run.
They are very good at making the containers convenient.
That is why we use them.
Convenience has a shape
I do not think any of this requires believing that large technology companies have produced nothing useful. That would be a slightly difficult position for me to maintain while typing this on a Mac, carrying an iPhone, watching YouTube and running much of my public software on free infrastructure supplied by large technology companies.
The problem is dependence.
There is a useful term for part of this in academic research: platformisation. David Nieborg and Thomas Poell wrote about the platformisation of cultural production in 2018, describing large online platforms extending themselves into the wider web until the production and distribution of other people's work becomes increasingly dependent on their infrastructure, markets and governance. Publishers and creators then have to adapt to changes in curation, interfaces, pricing and platform policy which they do not control.
You do not need to be a professional YouTuber for a mild version of that relationship to be familiar.
The platform decides how long your post can be. It decides whether links get previews. It decides whether you can change the page layout, whether you can export your archive, whether chronological feeds exist, whether third-party clients are allowed, what kinds of files can be uploaded, whether your followers actually see what you post and, ultimately, whether the particular form of expression you want to use exists as a feature.
If the answer is no, you work around it.
People write screenshots of text because the post format is too restrictive. They split essays into fifteen-post threads. They say "link in bio" because direct links are awkward to place or poorly integrated. Creators learn to phrase, crop, time and package things around whatever they believe the recommendation system currently wants.
Sometimes that belief is folklore and sometimes it is grounded in real platform behaviour, but either way the interface starts shaping the work.
Research on creators has been documenting this for years. Brooke Erin Duffy and Colten Meisner's study of algorithmic visibility among social-media creators describes people whose ability to reach an audience is tied to systems they experience as opaque and unpredictable.
I am wary of turning findings like that into the usual "THE ALGORITHM IS BRAINWASHING YOU" argument. Human beings are perfectly capable of becoming repetitive without machine assistance.
The broader point is simpler: interfaces and recommendation systems are not neutral pipes. They shape what becomes easy to see, easy to make and easy to choose.
A personal website is one small way of taking some of that decision-making back.
The web is allowed to be weird
What I like about personal websites is not only that the content can vary from person to person. The entire website can.
Mine is fairly normal. ewancroft.uk has navigation, pages, project listings, blog posts and a layout which is recognisably designed for reading text. I am not making you navigate a procedurally generated dungeon to find my GitHub profile.
I have still made the site mine in ways that would make absolutely no sense as product features on a general-purpose social network. Its colour system uses OKLCH and changes gradually through the year, interpolating between colours associated with the eight Sabbats of the Wheel of the Year. Even the neutral colours pick up a little of the current hue rather than sitting at pure grey. Pagan symbols and lunar details turn up throughout the design because they mean something to me. The whole thing is documented in far more detail than any reasonable visitor needs on my design page, because apparently documenting the design system of my own personal website is something I consider a recreational activity.
Somebody else's site can be nothing like mine. It could look like System 7. It could be a terminal. It could be a garden, a scrapbook, a collage or an entire 3D space where the pages exist as objects in a room. It could also be one HTML file with Times New Roman, a white background and twenty blue links.
On social media, visiting another person's profile mostly means looking at another person's information through the same application. I use Bluesky every day and have no particular interest in pretending that is bad, but somebody's Bluesky profile is still a Bluesky profile. Their posts and personality change; the surrounding structure mostly does not.
With personal sites, I can click a technical article somebody wrote and then discover that the rest of their website is dedicated equally to programming, local birds, an obscure album they bought in 1998 and a collection of photographs of badly designed drain covers. The next person's homepage might require me to drag windows around a fake desktop.
I like that variation because it reminds me that the web itself does not require everybody to live in identically shaped boxes.
The boxes are an application design choice.
A website doesn't make you independent of the internet
I should probably deal with the obvious contradiction before somebody else does.
My website is hosted on Vercel. Its source is on GitHub, which is owned by Microsoft. My domain needs DNS infrastructure. My AT Protocol data needs servers. The browser you are probably reading this in is overwhelmingly likely to involve either Google's Chromium project or Apple's WebKit somewhere in its ancestry.
Having ewancroft.uk does not mean I have gone to live in a self-sufficient cabin in the digital woods.
Nor do I particularly want to.
The difference is that none of those individual services has to define the whole thing.
Vercel can host the current version without being my website in the identity sense. If I stop liking Vercel, I can deploy the source somewhere else and point the domain there. GitHub can host the repository without the public address of my site becoming github.com/ewanc26/website. Leaflet can be the editor I use for my blog without the canonical copy people read needing to be a leaflet.pub URL.
That separation is important to me.
The containerised model tends to bundle identity, distribution, presentation, storage and audience together. Your TikTok account is your identity on TikTok, your publishing tool, your archive, your distribution system and the place where your audience relationship exists. Leaving therefore means rebuilding several things simultaneously.
The web is much better at letting those pieces come apart.
I didn't start with what I have now
The chronology of my own site is fairly ordinary, which is one reason I keep mentioning it.
I started blogging in November 2022. I did not immediately decide that this required owning a domain and engineering a permanent personal publishing stack. For several months, the writing existed before the personal website did.
The first version of the website went online on 31 May 2023, hosted for free on Neocities.
That was enough. I could write HTML and CSS, upload it and slowly work out what I wanted the site to become.
Then, on 9 January 2024, a little over seven months after launching the site, I bought ewancroft.uk.
These days the site is a SvelteKit application hosted on Vercel, backed by Git, full of AT Protocol integrations and attached to enough build tooling that index.html has long since stopped being an accurate description of what happens when somebody loads the homepage.
I like it this way because developing the website has become part of the hobby. It absolutely did not need to begin this way.
Since May 2023 I have changed hosts, changed frameworks more times than I would advise anybody else to do, rewritten large parts of the frontend and gradually turned the site itself into one of my longer-running software projects. Sometimes those rewrites were justified by things I had learnt. Sometimes I suspect I simply enjoy rebuilding websites.
The domain is now the stable bit.
And the thing costs me £10.79 a year.
I think that is worth spelling out because custom domains are sometimes talked about as though they are some expensive final stage of becoming a Proper Website Person. They can cost substantially more depending on the TLD and registrar, obviously, and I also own croft.click, but the main address I use for myself costs me £10.79 per annum.
For that I get an address which is mine to point wherever I want.
I can move from Neocities to Vercel. I can move from Vercel to something else later. I can throw away the entire frontend and rebuild it. I can change where the blog is rendered and where the data lives. As long as I retain the domain and update the DNS, ewancroft.uk can remain the address people know.
Considering how much of my online identity now hangs from it, £10.79 a year is not much.
I still would not tell somebody to buy one before they have made anything. I didn't. My site spent its first seven months on its Neocities address, and nothing about that made it less of a website. The custom domain made sense once I already knew I liked maintaining the thing and wanted a more permanent address for it.
I hate WYSIWYG editors, personally
There is a slight problem when programmers start telling other people to make personal websites: we have a habit of making the activity sound far more technically demanding than it needs to be.
I am especially vulnerable to this because I personally hate building websites in WYSIWYG editors.
I do mean hate. Give me a visual editor where I have to drag rectangles around, click through panels and select from whatever properties the application has decided to expose and I will very quickly start looking for the source code.
That is a preference, not an argument that WYSIWYG editors are bad.
The abstraction which annoys me is the thing that makes them useful to somebody who does not want to spend an evening reading MDN documentation because a box has moved six pixels to the left. There is no virtue in hand-writing CSS if you hate writing CSS. If a visual editor gets somebody from "I'd quite like a website" to actually having one, it has succeeded.
I prefer writing websites manually because I want control at a level which would be irritating for most people. If I decide my theme colour should be mathematically derived from the current date, I want to write the function. If I dislike how something behaves at 768 pixels wide, I want to inspect the CSS. If I want to add some strange interaction that no page-builder component supports, I would rather implement it than work out how to persuade a visual editor to approximate it.
My site metadata page describes the website as a creative outlet as much as anything else, and that is probably the important distinction. I do not manually build it because that is the correct or pure way to make websites. I do it because making the website itself is one of the things I enjoy.
You also inherit every problem the abstraction would otherwise have dealt with. Once you take complete control of the markup, layout and behaviour, accessibility is your responsibility. Keyboard navigation, semantic HTML, focus states, reduced motion, readable line lengths, responsive layouts, colour contrast and browser compatibility all become things you need to think about.
If Safari decides that a perfectly ordinary CSS rule actually means something nobody else has ever imagined, congratulations, it is your problem now.
The freedom is worth that to me. It might not be worth it to you.
You can make a website for free
This is also why I do not want "make a website" to quietly mean "learn frontend development first".
If you want something conventional with a visual editor, WordPress.com still has a free plan. You get hosting, themes, its block editor, unlimited pages and posts and limited media storage. There are the normal restrictions of a free hosted service, but it is enough to make a perfectly real personal website without learning HTML.
Carrd is lighter. Its free version is particularly suited to a one-page personal homepage with a short introduction, links and projects. I would personally be trying to escape the editor within twenty minutes, but that is a statement about me rather than Carrd.
Blento is particularly interesting if you already use Bluesky or anything else on AT Protocol. It is a visual bento-style personal-site builder where you can arrange cards containing links, media and data from elsewhere in the Atmosphere. The records behind the page live in your AT Protocol PDS rather than only in an isolated Blento account, which makes it a nice example of a visual builder that does not need to reproduce the most restrictive parts of the platform model.
You can use a friendly interface without requiring the friendly interface to become your entire identity.
Then there is Neocities, which I can recommend from direct experience because it is where this all started for me.
A free Neocities account gives you static hosting, a subdomain and an in-browser editor without advertising. More importantly, you get somewhere to put actual web files. You can write index.html, make a stylesheet, refresh the page and immediately discover that you have misunderstood CSS.
This is educational.
The community around Neocities also demonstrates how wildly personal websites can vary when people are not all designing inside the same product system.
For developers, GitHub Pages is another easy route: put a static site in a public repository and GitHub will host it, with support for a custom domain later. Cloudflare Pages gives you a more modern Git-based deployment setup if that is what you want.
Vercel is where mine lives now. I use it because the free infrastructure is enough for my personal site and most of the little public tools I deploy, while still giving me the SvelteKit, Git deployment and server-side features I want. In practice, my domains are basically the recurring cash cost of most of this.
Again, the version I launched on 31 May 2023 did not need any of that. Neocities was fine.
Put your writing somewhere too
The blog is the other half of this for me.
Social media is very good at conversation, but it is not especially good at giving thoughts a permanent home. I post something on Bluesky and, unless it has some reason to keep circulating, within a fairly short period it has vanished into thousands of newer posts. It still exists, technically, but finding something I wrote two years ago usually involves remembering enough of the wording to search for it.
Blog posts behave differently. They have titles. They have URLs. I can write something, link to it six months later and reasonably expect the link to still mean the same thing.
My blog is currently published with Leaflet. Leaflet is the editor and publication application I use, while the underlying documents live in my AT Protocol repository and blog.ewancroft.uk is my own frontend for reading them.
Leaflet, pckt.blog and Offprint all work with the broader Standard.site ecosystem for long-form publishing on AT Protocol. They have their own editors and interfaces, but they are working with shared document and publication structures rather than each inventing a completely isolated model of what a long-form post is.
I have more than five hundred site.standard.document records in my repository now. Some are blog posts; others are project documentation, development logs and creative writing. I did not deliberately decide I wanted five hundred long-form documents attached to my identity. I have just been publishing things for several years and apparently this is what happens.
The portability side matters because I have migrated my writing before. I used Hugo before my current Standard.site setup, and that migration was painless largely because the source was Markdown.
That experience is why I built Bismuth, which converts Standard.site documents from Leaflet, pckt and Offprint back into Markdown. The live format can be richer and more structured, but I can still get the writing back into something extremely boring that a huge number of other tools understand.
Boring formats are underrated.
I am happy using Leaflet. I am not planning an escape from it. I just prefer being happy with an application while knowing that leaving it would be a technical task rather than an archaeological expedition.
The site doesn't have to be one subject
Another thing I like about personal websites is that they do not really require a coherent content strategy.
Mine certainly does not have one.
I write about software because I spend a significant amount of my life writing software. I write about AT Protocol because I have apparently chosen to make that everybody else's problem too. I also write about language, disability, religion, games, technology, music and whatever else has occupied enough of my brain that posting about it three times on Bluesky has stopped being sufficient.
The website has projects, writing, status information and whatever other things I decide belong there. The common subject is basically me.
That would be terrible advice for a commercial publication. If you are building a business, having a clear audience and subject probably matters quite a lot. I am not building a business.
My blog is called Ewan's Blog and its description is "I ramble, enjoy."
This has so far proved sufficient editorial guidance.
I know roughly sixty people subscribe to it. I like that people read what I write, obviously. I like getting replies and finding out that somebody has linked a post somewhere. I just do not particularly care whether the number becomes seventy by the end of the month, and I do not want knowing the number to change what I choose to write.
Once every subject needs to justify itself against engagement metrics, I have recreated one of the things I wanted the personal site to get away from.
Social media is still useful
I am not arguing that everybody should delete their social accounts and return to an imagined golden age of manually checking bookmarks.
I like social media. I use Bluesky far more frequently than I publish blog posts, and most thoughts do not need two thousand words and a permanent URL. Sometimes I want to complain about a compiler, post a song or make a joke and then continue with my day.
The two forms work together perfectly well. I publish an article, post the link on Bluesky and people can talk about it there. Somebody else might find it through RSS, search or another website. The article does not need to live inside the place where the conversation happens.
The same applies to AT Protocol more generally. My website can render information from the same repository that multiple completely different applications understand. Blento can turn an AT Protocol identity into a visual homepage. Leaflet can turn it into a publication. Bluesky turns part of it into a social feed. My own site presents bits of the same underlying data in yet another way.
That is one reason I am so interested in open protocols. I grew up with applications where changing the interface usually meant changing the service, account, social graph and data store at the same time.
One identity being able to have several interfaces feels much healthier.
Having a website feels weird now, which is weird in itself
This is the bit I keep coming back to when people my age find the website unusual.
Having your own webpage should not really be an especially strange thing to do on the World Wide Web.
Yet it often feels like one.
I have had conversations where saying I have a personal website gets a more confused reaction than saying I have a Bluesky account, Discord account, GitHub account or almost any other kind of platform presence. Those make immediate sense because an account is now the normal mental model for existing online. A website prompts the additional question of what you actually do with it.
I think that says quite a lot about how successful the container model became.
If you are around my age, the mainstream internet you grew up with probably did not teach you that making a homepage was a normal part of participating online. It taught you to sign up.
The page had already been built. You supplied the content.
The colours, dimensions, allowed media, navigation, discovery system and monetisation had already been decided. Increasingly, so had the mechanism deciding who actually saw what you supplied.
That arrangement is exceptionally convenient, which is why I have no expectation that it will disappear. Most people do not want to build a webpage before sharing photographs with friends and should not have to.
I just think the convenience has become so normal that we have forgotten the container is optional.
You can have the accounts and a website.
I do.
I discovered the personal web late
People occasionally announce that blogging is back, or that personal websites are having a revival. Maybe they are. I am not sure it matters.
They never completely disappeared.
What disappeared from the mainstream was the expectation that an ordinary person's internet presence might primarily be something they shaped themselves. Platforms became easier, everybody was already there, and the account gradually became the default unit of online identity.
I grew up after that had mostly happened, which is why I find the nostalgia around personal websites slightly funny. I cannot be nostalgic for maintaining a homepage in 2003. I was born in 2005.
My nostalgia is much more likely to involve an old version of YouTube.
Personal websites are interesting to me because I encountered them from the other direction. I learnt the containerised version of the internet first, where you joined an application and adapted yourself to its available fields. Then I discovered that the browser itself does not actually care about any of those conventions.
You can make a document. You can link it to another document. You can make it ugly, beautiful, useful, stupid or completely unlike anybody else's.
You can host it for free. If it sticks, you can buy a domain later, and that does not necessarily mean taking on some enormous recurring expense. My main domain costs £10.79 a year, and for me that is easily worth having a stable address I can keep while everything behind it changes.
You can stay on Neocities forever or somehow end up with a SvelteKit application on Vercel whose colour palette tracks the Wheel of the Year.
Apparently I chose the latter.
Use WordPress.com if you want WordPress. Use Carrd if one page is enough. Use Blento if the visual builder and AT Protocol integration appeal to you. Make a Neocities account if you want to start poking at HTML. Use GitHub Pages, Cloudflare Pages or Vercel if code and Git are already familiar.
If you want to write, try Leaflet, pckt.blog or Offprint, or put a blog on whatever site you already have.
You do not need to settle the architecture of your permanent online identity before making the first page. I started blogging in November 2022, put the first version of my personal website online on 31 May 2023, bought ewancroft.uk on 9 January 2024, and have changed almost everything underneath it since.
The first version still counted.
I grew up on an internet where an extraordinary amount of ordinary online life had been compressed into containers owned by a handful of enormous companies. People my age sometimes think it is slightly weird that I made a website instead.
I still have plenty of accounts inside those containers. I will probably continue having them.
I am glad none of them is the only place I exist.