Five years is now long enough to make a conversation unintelligible. "6-7," "delulu is the solulu," "Labubu matcha Dubai chocolate" — the origins are tangled, the popularity viral, and the mechanism behind both is the social media algorithm. Linguist Adam Aleksic, who writes and makes videos as Etymology Nerd, takes on that mechanism in his book Algospeak: How Social Media Is Transforming the Future of Language. In conversation with Figma editor Jenny Xie, he traces how algorithms have reshaped not only speech but self-understanding, social interaction, and how people engage with the world.

Interface as semiotics

Product design carries communicative weight of its own. Swiping on Tinder is visual semiotics: the gesture implies people are commoditized, and that reframes how users treat one another. Even its direction matters — right-versus-left leans on the Western notion that "right is good," with the Latin root for "left" meaning "sinister." Marshall McLuhan's 1964 book Understanding Media: The Extensions of Man gave the idea its name: the medium is the message.

Every profile is a performance, a signal aimed at the kind of person the user wants to attract, shaped by what the interface affords. Hinge's prompts steer people toward storytelling; Grindr pushes users to sort themselves into tribes such as "bear" or "twink," which changes outward expression of identity and possibly the internal sense of who they are.

Context collapse and the speed of slang

Algorithms work in much the same manner. The internet opens spaces for communities to talk, then opens those spaces wide enough for their words to escape. What users experience is context collapse.

Consider "slay." When the term rose in the 1980s New York City ballroom scene, it percolated slowly: Black and Latino gay men, then white gay men, then the straight white girlfriends of those men. Today algorithms push trends captured through metadata, and metadata is no longer just hashtags — it is every word written or spoken. Influencers adopt trending language, their videos go viral, and the next viewer has no idea the term comes from African American Vernacular English. Because it surfaced on their "for you" page, they assume the slang is for them.

"Slay," "serve," "queen," "cooked," "ate," "that's bussin,'" "it's giving" — all ballroom slang that went mainstream. Aleksic's rule of thumb is that 95% of words come from either AAVE or 4chan, a figure he calls shocking in how often it holds. "Slay" no longer carries its old connotations, and language moves on regardless, but the machinery of change deserves attention.

Meme transmission and what gets lost

The words themselves are not the worry. If anything they expand the ways people can express themselves, and speakers are strikingly creative with them. The concern is culture: why, for instance, do so many incel terms — "looksmaxxing," the "-pilled" suffix — reach the mainstream? Memes carry them across bubbles, acting as a vehicle that Trojan-horses an idea which might otherwise repel people. For most, critical awareness blunts the harm.

Algorithms cut both ways. There is a democratizing side: voices once excluded by traditional gatekeepers now have representation, and Aleksic points to discoveries he has made through recommendation systems, as well as the platform they gave him personally. Calling any tool monolithically good or bad misses the point; what matters is what a given user draws from it.

Why words stick offline

Online there are incentives to use algospeak, but what drives adoption in ordinary speech? Some words start ironically and settle in permanently. A term truly spreads when it has genuine applicability to a new context and feels intuitive — when it does not stick out.

Staying current has its own pull; anyone who wants to seem tapped in has to know what middle schoolers are saying, which at the moment means "6-7," "chopped," "unc," and "aura." Meme coin markets track the same dynamic: as soon as a meme like "chopped" trends, the associated coin jumps, because the coin stands for a proportional slice of collective attention. The value of these transient assets is bound to how popular the meme is.

Predictive machines and the flattening of style

Merriam-Webster's 2025 word of the year is "slop": digital content of low quality, usually produced in quantity by artificial intelligence.

The deeper point is representational. Any attempt to represent language will be inaccurate — a dictionary is, and so are the inputs feeding ChatGPT. Those inputs overrepresented the word "delve," and as people began drafting research papers with AI, the share of abstracts using "delve" rose tenfold over three years. Represented words become available in the mental lexicon.

AI also arrives structured as conversation, mirroring the user's tone of voice, which encourages anthropomorphizing. The essential thing to remember about algorithms and AI alike is that both are predictive machines. Predicting the likeliest response means ignoring the outliers — and the outliers are sometimes the most interesting part.

Deliberate use over passive consumption

Finding those outliers starts with knowing exactly what a tool is good for and limiting it to that purpose, which will differ by person. Aleksic uses ChatGPT to spell-check essays or to suggest authors similar to ones he already likes. He treats his own discovery process as worth protecting, because how someone arrives at something carries meaning for them.

The same logic applies to the scroll state. Studies show people prefer the infinite scroll to having to work for content, yet they tend to feel more satisfaction when they have found material themselves.

His takeaway condenses to McLuhan: the medium is the message, and how we communicate ultimately shapes the communication. Literacy about how these media shape experience makes a difference — knowing which community a meme came from, or why something is being recommended. That is what it takes to act as something like conscious citizens of the internet, and of the world.