Stop searching for the next Twitter. Build your island instead.

I know what it feels like to build something real on a platform — and then watch it disappear. For years, I ran RealScientists Nano, a rotation curation account where a new materials or nanoscientist took over every week to share their research, their life in the lab, their questions. Over 120 scientists from around the world took part. During the pandemic, it became a genuine lifeline — a space to connect, to laugh, to feel less isolated. In 2022 alone, the account generated close to a million impressions. Then Twitter changed, and slowly, that community scattered. I hear the same question constantly from scientists now: “Where do we go next? What’s the new Twitter?” I understand why they’re asking. But I think it’s the wrong question entirely.

Science Twitter was something rare

What scientists lost when Twitter changed was not just a platform. It was a specific, almost accidental kind of community — one that let a nanoscientist in the Netherlands stumble into a conversation with a materials researcher in India, or a PhD student ask a question publicly and get an answer from someone they’d never have met otherwise. That low-barrier, high-serendipity environment was genuinely unusual. It did not exist before Twitter in quite that form, and it has not been replicated since.

The grief scientists feel about this is proportionate to the actual loss. A Nature survey confirmed what most scientists already knew: thousands were cutting back on Twitter, feeling genuine uncertainty about where to build and maintain community. Climate scientists reported a 15 to 30-fold increase in hostile replies over two months. The platform that had been a relatively safe, intellectually rich space became something else entirely — and people mourned it.

That mourning is valid. What comes after it matters.

We’ve been here before — and we’ll be here again

After Twitter changed, scientists split across three platforms. Some went to Mastodon, drawn by its decentralised structure and science-friendly ethos. Some went to Bluesky. Some stayed on LinkedIn or retreated there. What happened next was entirely predictable, because we have seen it before.

Mastodon peaked at around 2.5 million active users in November 2022 — and then dropped back to roughly 880,000. Only 1.6% of scientists who publicly said they would move to Mastodon actually did, according to a New Scientist analysis. The platform was genuinely good in many ways, but it was too complex for most people to set up, and without the critical mass, the serendipity never returned.

Bluesky is a different story — and a more hopeful one, at least on the surface. It now has over 40 million registered users, and a Nature poll found that 70% of their readers are using it. The science community there is real and active. An analysis of 300,000 academic users showed that 18% of scholars migrated from Twitter to Bluesky between 2023 and early 2025. I made the move myself — Under the Microscope is now on Bluesky.

But here is the structural problem that does not go away: you are renting, not owning. Every platform is a landlord. The algorithm changes. The ownership changes. The business model changes. The momentum resets. You rebuild. You scatter. You rebuild again.

The platform era had an expiry date

This is not a Twitter-specific problem. It is a structural feature of how platforms work. They grow by aggregating attention, they monetise by manipulating it, and they decline when users lose trust or patience. The timeline varies. The outcome does not.

LinkedIn is already showing the signs. Approximately 25% of LinkedIn activity is now estimated to be AI-generated or bot-driven. AI-written comments are everywhere — generic, hollow, indistinguishable at a glance from real engagement. The platform’s own algorithm is trying to filter them and struggling. LinkedIn is owned by Microsoft, which also owned Skype — a platform that once connected hundreds of millions of people and was quietly left to decline after acquisition. The pattern is familiar.

What AI has done is accelerate the decay. When the cost of producing fake engagement drops to near zero, every platform that relies on engagement signals as a measure of quality degrades faster. LinkedIn is not collapsing — but it is becoming noisier, less trustworthy, and more exhausting to use. The scientists I speak to feel this already. They are spending more time on it and getting less from it.

The island era is already here

The shift I am watching — and living — is away from platform-driven community and towards what I think of as island-driven presence: owned, followable, multi-format spaces that do not depend on any platform’s survival.

Substack is the closest mainstream analogue. Nature published a piece in 2025 titled “Why scientists are flocking to Substack” — and the numbers are real. Epidemiologist Katelyn Jetelina built a newsletter during the pandemic that now reaches over 365,000 readers directly. No algorithm decides who sees it. No ownership change can scatter her audience. She owns the list.

This is what an island looks like in practice. It is not necessarily a website — though it can be. It is a space where your audience follows you, not a feed you happen to appear in. It can be a newsletter, a podcast, a membership hub, a combination of formats. The key distinction is ownership: if the platform closes tomorrow, your audience can still find you.

Under the Microscope became an island. The podcast and the rotation curation account are now unified under one brand, across platforms — but the podcast audience is ours regardless of where the episodes are hosted. That was a deliberate choice, made after watching Twitter change and understanding that community built entirely on borrowed ground is fragile.

The scientists who are building the most durable presence right now are not the ones with the most followers on any single platform. They are the ones who have diversified — a newsletter here, a podcast there, a consistent personal site — so that no single algorithm change can undo their work.

I do not know exactly when the shift will be complete, or what the dominant form of scientific community-building will look like in ten years. But I am confident of one thing: the scientists who wait for the next Twitter will still be waiting. The ones who start building their island now — however small, however imperfect — will have something no platform migration can take from them.

What island are you building?

Related reading on The Science Talk

This piece accompanies the full story of how RealScientists Nano made the move from Twitter to Bluesky — the practical and strategic reasons behind the transition, and what it means for the Under the Microscope community.

Browse all Perspectives or get in touch →