Between 2005 and 2015, the center of podcasting was a two-person conversation that ran ninety minutes to three hours with no segment clock. This article treats that format — the long conversation — as a dated artifact and traces its decline through four documented changes: playback controls that edit time (2012–2014), listener analytics that priced pacing (2017), exclusivity deals that enclosed the exemplars (2019–2020), and a clip economy that selected against the form (2018–2022). Each change is checkable. The loss is itemized at the end.

The form, dated

The long conversation was native to open syndication. Apple shipped podcast support in iTunes 4.9 on June 28, 2005, and for the next decade the dominant listening device was a phone connected to an RSS feed, with no station clock and no program director. The form’s grammar held across its exemplars: one host, one guest, no segments, and a runtime set by exhaustion rather than by a format clock. WTF with Marc Maron began in September 2009; The Joe Rogan Experience in December 2009; Dan Carlin’s Hardcore History released Blueprint for Armageddon in six parts between 2013 and 2014, totaling more than twenty hours. The form’s peak as a civic object can be dated precisely: on June 22, 2015, WTF published a garage interview with President Barack Obama, conducted days earlier in the room where every other episode was made.

The mechanism: the player becomes an editor

Apple released its standalone Podcasts app on June 26, 2012, with a 2x speed control in the player; a 1.5x step followed by 2014. In July 2014, the independent developer Marco Arment released Overcast, a third-party player, and its second version, dated November 2014, added Smart Speed, which shortens detected silences during playback without changing pitch, and Voice Boost, which levels volume. Pocket Casts, another independent player, shipped silence trimming in the same period. None of these settings altered a single recording. All of them altered the time inside the recordings.

The arithmetic is plain. A three-hour episode at 1.5x takes two hours; at 2x, ninety minutes. Smart Speed returns a few minutes per hour of talk by deleting pauses. The interface presents this as recovery, and for a listener with a commute it is. The mechanism sits in what the word recovery assumes: that the pause was never part of the conversation. In the long conversation the pause was often the load-bearing part — the audible moment in which a guest decided whether to say the rehearsed thing or the true one. A player that deletes silence does not compress a conversation. It deletes the deliberation and keeps the sentence.

The dashboard: pacing becomes countable loss

In June 2017, Apple announced listener analytics for podcasts at its developer conference, and by that fall any podcaster in its directory could see, episode by episode, the second at which listeners skipped ahead or stopped. The data was private to each show, and it was decisive. Trade coverage and consultancy advice converged within a year: front-load the payoff, cut the cold open, keep the first sixty seconds clean. The advice was not hostile to talk. It was hostile to approach — the long conversation’s signature move, twenty minutes of circling before the disclosure.

A restraint is owed to the record. Dashboards did not forbid anything, and several long shows kept their pacing. But the analytics made one pacing style countable as loss, and countable losses get managed. By 2019, a runtime that had been a form had become a metric.

The institution: exclusives, rate cards, and the clip

On February 6, 2019, Spotify announced the acquisitions of Gimlet Media and Anchor, a combined deal reported near $340 million; it bought The Ringer in February 2020 for a reported $196 million. On May 19, 2020, Spotify announced a licensing agreement with The Joe Rogan Experience, reported at more than $100 million, and on September 1, 2020, the show’s catalog went exclusive to its app. The longest conversation in the medium did not get shorter. It got enclosed — removed from the open RSS ecosystem that had defined podcasting since June 28, 2005. Length survived as a premium library asset. The commons around it did not.

The advertising market compressed the middle of the field. The IAB’s annual series put United States podcast ad revenue at $708 million in 2019 and about $1.4 billion in 2021; the reports are published by the IAB and remain the industry’s reference figures. Inventory in that market is sold against listened time. Every retained minute is priced, every skipped minute is a discount, and every interface change that shortens listening — speed controls, silence trimming, chapter skipping — arrives in the rate card before it arrives in the studio. A show that holds attention across ninety minutes earns more mid-roll slots. A show that loses listeners at minute four earns advice.

The clip economy selected against the form from the other side. Musical.ly merged into TikTok in August 2018, and by 2021 the vertical sixty-second excerpt had become a main discovery channel for talk shows. A clip needs a quotable line, and a quotable line needs to arrive without its twenty-minute approach. Shows began recording with extraction in mind — a clean sentence every ninety seconds, a setup that survives cropping. The caption window has its own compressed history, documented elsewhere on this site. Here it is enough to date the effect: within three years of the merger, clip-first production was being described in trade coverage as standard practice.

The loss, itemized

Four entries, in the register of a catalog.

1. The shared clock. Real-time listening gave every listener the same duration. Speed controls privatize duration: two people who heard the same episode no longer heard the same conversation at the same tempo. The form’s implicit claim — that a stranger’s thinking deserved your unedited hours — stopped being made.
2. The audible decision. The pause before an answer is where the speaker chooses. Smart Speed’s tally counts those seconds as savings. What is saved is the deliberation.
3. The unedited arc. The long conversation allowed a guest to be wrong for forty minutes and correct themselves on the record within a single take. The clip economy selects the confident second and discards the reversal, so the narrative shape that only length could hold stopped being recorded.
4. The commons. The 2020 exclusivities moved the exemplars into walled catalogs. The long conversation’s remaining audience listens inside proprietary apps, at negotiated speeds, on dashboards that report their minutes.

A note against nostalgia: the long conversation was not a pre-digital innocence. It was a native product of open syndication, born the week iTunes 4.9 shipped in June 2005 and matured in a decade of cheap audio hosting. What declined between 2014 and 2021 was not an outside but a pace — a specific agreement, held for about ten years, that unedited thinking was worth a listener’s real time.

What survives, dated

The form did not die; it became a boutique. Hardcore History installments still run four to six hours. EconTalk, which began in 2006, still publishes a weekly unedited hour. The Joe Rogan Experience still runs three hours, inside its exclusive enclosure. What changed is positional: between the changelog entry of November 2014 and the exclusives of 2020, the long conversation moved from the medium’s center to its edge, where it is scheduled and priced as an exception — the season ticket of spoken audio. The exception proves the catalog.

The Overcast page that opened this piece remains live as of this writing, and its tally is honest: it really does return minutes. The minutes it returns were the part of the conversation where nothing had been said yet. That is the whole of the loss, and it is enough.

Frequently asked questions

What was the long conversation in podcasting?

An unedited two-person talk format, typically ninety minutes to three hours, native to RSS-distributed podcasts between 2005 and 2015. WTF with Marc Maron (September 2009) and The Joe Rogan Experience (December 2009) are the standard exemplars. Its convention was the absence of one: the recording ran until the host ran out of questions.

When did playback speed change podcast listening?

Apple’s standalone Podcasts app shipped with 2x playback on June 26, 2012; a 1.5x step followed by 2014. Overcast 2.0 added Smart Speed in November 2014, and Pocket Casts offered silence trimming in the same period. By the late 2010s, speed adjustment had become a default expectation of the player interface.

Does Smart Speed alter the episode file?

No. It removes detected silences during playback; the downloaded file is unchanged. The change is cultural, not technical: listeners stop hearing the pause in which a speaker decides what to say.

Why did the long conversation decline?

Four dated pressures converged: playback controls that treated pauses as waste (2012–2014), analytics that priced slow pacing as loss (2017), exclusivity deals that enclosed the format’s exemplars (2019–2020), and a clip economy that rewarded extractable moments over extended thought (2018–2022). No single decision shortened the form; the interface, the dashboard, the rate card, and the crop agreed on it.

Are long podcasts still being made?

Yes. Hardcore History, EconTalk, and The Joe Rogan Experience all continue in the long format as of 2025. The decline is positional: the form moved from the center of the medium to a boutique edge between 2014 and 2021.

Method note. Dates in this piece are checked against company release notes, the IAB’s published revenue series, and contemporaneous trade coverage, part of the running index at lt90.org. Corrections are welcome. — Lena Torres

Poetry is a compression format. It is language organized for retention under constraint: meter, line break, stanza, refrain, rhyme, image. Prose is language organized for extension: paragraph, chapter, argument, scene, subplot, aside. Between 2005 and 2025, the interfaces that carry text changed from page-shaped to feed-shaped. That change did not treat both forms equally. Poetry, as a set of formal decisions, maps more cleanly onto the new interfaces. Prose, as a set of formal decisions, resists them. This article treats that mismatch as a design artifact, not a taste argument.

This is a cultural criticism of attention compression. The evidence is dated, verifiable, and institutional: character limits, feed algorithms, reading-position indicators, audio interfaces, and the layout grammar of mobile screens. The claim is narrow. Poetry did not become more popular than prose in absolute terms. It became more legible to the systems that now decide what gets seen.

The Interface Shift, 2005–2025

In 2005, the dominant text interface was the browser window on a desktop monitor. The dominant reading posture was seated, forward-facing, with a scrollbar and a stable viewport. A long-form article could hold a reader for twenty minutes because the environment did not actively interrupt it. By 2015, the dominant text interface was the mobile feed. By 2025, the feed had become the default reading environment for most published language.

The feed is not a neutral container. It is a sequence of competing units, each with a preview, a timestamp, and a collapse threshold. The feed rewards units that can be completed before the next unit arrives. It punishes units that require a separate reading session. This is not a cultural preference. It is a structural condition.

Poetry, especially lyric poetry, is built for the unit. A sonnet is fourteen lines. A haiku is three. A ghazal is a series of autonomous couplets. Even a long poem is often modular: canto, section, numbered fragment. Prose is built for the session. A novel is not modular in the same way. A reported essay is not modular in the same way. A chapter can be excerpted, but the excerpt is a wound, not a unit.

The Character Limit as a Formal Constraint

Twitter launched in 2006 with a 140-character limit. The limit was an engineering decision derived from SMS protocol constraints, not a literary theory. But it created a new formal environment. A 140-character post is closer to a line of verse than to a paragraph of prose. It has a hard right edge. It forces enjambment. It rewards compression, image, and turn.

By 2017, the limit doubled to 280 characters. The change was announced as a response to cramming, especially in languages with longer average word lengths. But the formal environment did not change. A 280-character post is still a unit, not a session. It is still closer to a couplet than to a paragraph.

Prose writers adapted by serializing: threads, newsletters, multi-part posts. But serialization is a prose strategy that borrows poetry’s unit structure. The thread is a stanzaic form. The newsletter issue is a canto. The adaptation is real, but it is an adaptation toward poetry’s formal logic, not away from it.

The Reading-Position Indicator

Medium introduced an estimated reading time on articles in 2013. The feature was a design decision: a small line of text above the headline, reading “5 min read” or “12 min read.” It was meant to help readers budget attention. It also changed the reading contract.

A poem rarely carries a reading-time estimate. The omission is not an oversight. A poem’s duration is not predictable from its word count. A fourteen-line sonnet can take thirty seconds or thirty minutes. The reading-time indicator assumes a linear, extractive reading model. Poetry often refuses that model.

Prose, by contrast, is now labeled with its own duration before it is read. The label is a promise and a threat. A “20 min read” essay is a commitment. A “3 min read” essay is a snack. The label flattens prose into a time-management problem. Poetry escapes the label because it does not fit the model.

The Audio Turn

Between 2015 and 2025, audio interfaces became a primary reading environment. Podcasts, voice assistants, and text-to-speech tools moved language from the eye to the ear. The shift favored poetry in a specific way.

Poetry is already an oral form. It predates print. Its formal devices—meter, rhyme, assonance, repetition—are mnemonic devices designed for the ear. A poem read aloud is not a translation of the poem. It is the poem in its native medium.

Prose read aloud is a different object. A novel read by a narrator is an audiobook, a performance that adds a layer of interpretation. A news article read by a synthetic voice is a convenience, not a form. Prose does not become more itself when spoken. Poetry does.

The Playlist as Anthology

Spotify and Apple Music did not set out to change poetry. But the playlist format—a sequence of short, self-contained audio units—is structurally identical to a poetry anthology. A listener can move from a song to a spoken-word poem to a short essay without changing the interface. The poem fits. The essay, unless it is very short, does not.

This is not a claim that poetry dominates audio platforms. It does not. The claim is that poetry’s formal properties make it compatible with the playlist grammar. Prose’s formal properties make it incompatible unless it is cut into playlist-sized pieces.

The Mobile Screen as a Page Shape

The mobile screen is tall and narrow. It is closer to the shape of a poetry broadside than to the shape of a book page. A line of verse fits on a mobile screen without horizontal scrolling. A paragraph of prose wraps and reflows, but the reflow is invisible to the reader. The paragraph is a block. The line is a unit.

Line breaks are the most visible formal difference between poetry and prose. On a mobile screen, line breaks are also the most visible formal difference between a poem and a wall of text. A poem’s line breaks create white space. White space is a signal of brevity and approachability. A prose paragraph is a signal of density and duration.

This is a design observation, not a literary judgment. The mobile screen does not make poetry better. It makes poetry look easier. And in an attention economy, looking easier is a survival advantage.

The Screenshot as a Distribution Unit

Poetry spreads through screenshots. A poem that fits on one screen can be captured, cropped, and shared as an image. The image travels through feeds, group chats, and message threads. The poem becomes a visual artifact, not just a text artifact.

Prose does not screenshot well. A paragraph is not a complete unit. A page of a novel is not a complete unit. A screenshot of prose is a fragment that announces its own incompleteness. A screenshot of a poem is often the whole poem.

The screenshot economy is a distribution system. It favors forms that can be captured in a single frame. Poetry is that form. Prose is not.

Institutional Evidence

The survival claim can be checked against institutional data. The National Endowment for the Arts has tracked U.S. poetry readership through its Survey of Public Participation in the Arts. The 2017 survey reported that 11.7% of adults read poetry in the previous year, up from 6.7% in 2012. The increase was concentrated among younger adults, especially 18–24-year-olds. The same survey showed declines or flat trends for other literary forms over comparable periods.

The Poetry Foundation, a Chicago-based institution founded in 2003, expanded its digital presence during the same window. Its website, poetryfoundation.org, became a high-traffic poetry archive. The Academy of American Poets launched Poem-a-Day in 2006, a daily email and web feature. Both institutions made a specific design choice: distribute poetry in single-poem units, not in collections or issues.

Prose institutions did not make the same choice. Literary magazines continued to publish issues. Book publishers continued to publish books. The unit of prose distribution remained the collection, the issue, the volume. The unit of poetry distribution became the poem.

The Poem-a-Day Model

Poem-a-Day is a design artifact. It is a daily email containing one poem. The email is short. The poem is complete. The reader can finish it in the time it takes to read an email. The model is not new—it is a digital version of the tear-off calendar or the newspaper poem—but its digital form is optimized for the feed environment.

No equivalent institution distributes a daily short story or a daily essay with the same reach. The short story is too long for the email unit. The essay is too long. The poem is the right size.

What Prose Lost

Prose did not disappear. It changed shape. The long-form essay became the newsletter. The novel became the audiobook. The reported feature became the thread. Each adaptation is a form of compression. Each compression moves prose closer to poetry’s unit logic.

The loss is not in quality. It is in the reading contract. A prose work that requires a sustained, uninterrupted session is now a luxury object. It is the equivalent of a vinyl record in a streaming economy: valued, collected, but not the default.

Poetry’s survival is not a triumph of art over commerce. It is a case of formal compatibility. The attention age did not choose poetry. The attention age built interfaces that poetry already fit.

FAQ

Why does poetry fit the feed better than prose?

Poetry is built from units—lines, stanzas, short complete poems—that can be read in a single glance. Prose is built from sessions—paragraphs, chapters, arguments—that require sustained attention. The feed rewards units and punishes sessions.

Did poetry readership actually increase after 2005?

Yes, according to the National Endowment for the Arts. The 2017 Survey of Public Participation in the Arts reported that 11.7% of U.S. adults read poetry in the previous year, up from 6.7% in 2012. The increase was largest among younger adults.

Is this a claim that poetry is better than prose?

No. It is a claim about formal compatibility. Poetry’s design decisions—compression, line breaks, oral memory, unit structure—match the design decisions of feed-based interfaces. Prose’s design decisions—extension, paragraph flow, sustained argument—do not.

What is the most important design change behind this shift?

The shift from the page-shaped interface to the feed-shaped interface. The page invited sustained reading. The feed invites rapid unit consumption. Poetry was already a unit-based form. Prose was not.

Next Step for This Site

This article opens a recurring column: Interface Artifacts. Each entry will treat one dated, verifiable design decision—a character limit, a reading-time label, a screenshot format, an audio default—as a primary-source document for how attention compression changed literary form. The next entry will examine the 2017 Twitter character-limit change as a formal event in the history of the couplet.

Summary culture is the systematic replacement of primary engagement with compressed, pre-digested versions of texts, meetings, films, and conversations. Adjacent terms include attention compression, interface-mediated skimming, and secondary consumption. This article examines one specific design decision—the normalization of the summary as a default interface layer—and treats it as a primary-source artifact of the period 2005–2025. The audience for this publication is readers who notice that the container of information has changed shape, and who want a dated, verifiable account of how that happened.

The shift did not arrive as a single announcement. It accumulated through interface changes: the auto-generated video caption, the email thread collapsed into a “smart” preview, the book reduced to a fifteen-minute audio digest, the meeting reduced to an action-item list. Each of these is a small institutional change. Together they form a pattern. The pattern is now legible enough to describe without exaggeration.

The Interface Decision: Summary as Default

In 2006, Twitter’s 140-character limit was a constraint on expression. By 2017, the thread feature had turned that constraint into a format for argument. By 2023, the platform’s own “read more” truncation and algorithmic ranking meant that most users encountered only the first two lines of any post. The summary was no longer a user choice. It was the default view.

This is the core design decision under examination: the summary moved from a tool to a filter. A tool is something a person picks up for a purpose. A filter is something a person passes through without noticing. The distinction matters because filters shape what counts as information before the reader can ask a question.

Consider the email client. In 2010, Gmail introduced Priority Inbox, which sorted messages by predicted importance. The user still saw subject lines and sender names. By 2020, Gmail’s “summary cards” began extracting dates, tracking numbers, and event details from message bodies and displaying them above the message. The user could act on the summary without opening the email. The email itself became a backup artifact, not the primary object of attention.

This is not a complaint about convenience. It is a description of a change in the unit of reading. The unit shrank from the document to the digest. The digest is not a shorter document. It is a different object with different properties: it has no tone, no sequence of reasoning, no hesitation, no asides. It is optimized for scanning, not for understanding.

Institutional Changes That Made Summary Culture Legible

Summary culture became institutional through a series of dated, verifiable changes. Each one is a primary-source artifact in its own right.

2005–2010: The Rise of the Auto-Generated Preview

Google’s search results had shown snippets since the late 1990s, but the snippet was a quotation from the page, not a summary of it. The shift came when platforms began generating previews that were not quotations. Facebook’s 2007 introduction of the News Feed with collapsed stories meant that a user saw a headline, a thumbnail, and a one-line description. The description was often written by the publisher, but the interface treated it as sufficient. The full article was one click away, and the click was optional.

By 2010, the “read more” link had become a standard element of blog templates. The decision to hide the full text behind a click was a design choice, not a technical necessity. It trained readers to expect that the first paragraph was a summary, and that the summary was enough.

2011–2015: The Digest as a Product

In 2012, the app Summly launched with a promise to condense news articles into 400-character summaries. It was acquired by Yahoo in 2013 for a reported $30 million. The acquisition was not about the technology alone. It was about the premise: that a generation of readers would prefer a summary to the source. The premise was correct.

In the same period, the newsletter format shifted. The Daily Skimm, founded in 2012, delivered a conversational summary of the day’s news. It did not link to full articles as a primary feature. The summary was the product. The source was an afterthought, listed at the bottom for readers who wanted to “go deeper.” The phrase “go deeper” itself became a marker of the new hierarchy: depth was now a destination, not the default.

2016–2020: The Meeting and the Book, Compressed

In 2016, Slack introduced threaded messages, which allowed conversations to be collapsed into a single line. The thread was a summary of the conversation, but it was also a way to avoid reading the conversation. The same year, the app Blinkist, founded in 2012, reached a million users. Blinkist offered fifteen-minute summaries of nonfiction books. The marketing language was explicit: “Read or listen to the key ideas from 5,000+ nonfiction books.” The word “key” did the work. It implied that the rest of the book was not key, and therefore not necessary.

In 2020, the COVID-19 pandemic accelerated the shift. Meetings moved to Zoom, and Zoom introduced automatic transcription. The transcript was a summary of the meeting, but it was also a way to avoid attending the meeting. The phrase “I’ll read the transcript” became common. The transcript was not a summary in the traditional sense—it was a full record—but it functioned as a summary because it allowed the reader to skip the experience of being present.

What Is Lost When the Summary Replaces the Source

The loss is not information. A good summary preserves information. The loss is sequence. A summary removes the order in which ideas were encountered. It removes the false starts, the digressions, the moments when the author changed their mind. These are not defects. They are the texture of thought.

Consider a book like The Shallows by Nicholas Carr, published in 2010. A summary of the book would say: “The internet is changing our brains, making us less capable of deep reading.” That is accurate. But it is not the book. The book is an argument built over 200 pages, with evidence from neuroscience, history, and personal experience. The summary removes the experience of being persuaded. It delivers the conclusion without the reasoning. The reader who only reads the summary has not read the book. They have read a sentence about the book.

The same applies to meetings. A meeting summary might say: “We decided to delay the launch by two weeks.” That is the outcome. But the meeting itself contained the discussion that led to the outcome: the disagreement, the data presented, the moment when someone changed their mind. The summary removes the process. The process is where understanding lives.

The Tradeoff: Speed vs. Understanding

Summary culture is not a mistake. It is a tradeoff. The tradeoff is speed for understanding. A person who reads summaries can cover more ground. A person who reads sources can understand more deeply. Both are valid choices. The problem arises when the summary is presented as equivalent to the source, and when the interface makes the source harder to reach.

This is the institutional change that matters most: the default changed. In 2005, the default was the full text. The summary was an option. In 2025, the default is the summary. The full text is an option, and often a hidden one. The reader must now make an effort to reach the source. The effort is small—a click, a scroll—but it is an effort. And effort is exactly what summary culture promises to eliminate.

The result is a two-tier system of reading. Those who make the effort read the source. Those who do not read the summary. The gap between them is not a gap of intelligence. It is a gap of attention. And attention is the resource that summary culture is designed to conserve.

Dated, Verifiable Examples of the Shift

The following examples are dated and verifiable. They are not opinions. They are artifacts.

• 2006: Twitter launches with a 140-character limit. The limit is a constraint on expression, but it also trains users to compress.
• 2007: Facebook introduces the News Feed with collapsed stories. The full post is one click away.
• 2010: Gmail introduces Priority Inbox. The inbox is sorted by predicted importance, not by chronology.
• 2012: Summly launches. The app condenses news articles into 400-character summaries.
• 2012: The Daily Skimm launches. The newsletter is a summary of the day’s news, with the source as an afterthought.
• 2013: Yahoo acquires Summly for a reported $30 million. The acquisition validates the premise of summary as product.
• 2016: Slack introduces threaded messages. Conversations can be collapsed into a single line.
• 2016: Blinkist reaches a million users. The app offers fifteen-minute summaries of nonfiction books.
• 2020: Zoom introduces automatic transcription. The transcript becomes a substitute for attendance.
• 2023: Gmail’s summary cards extract dates, tracking numbers, and event details from message bodies. The email becomes a backup artifact.

Each of these changes is small. Together they form a pattern. The pattern is the replacement of the source with the summary as the default unit of information.

The Museum Label: How to Read This Artifact

If summary culture were an object in a museum, the label would read something like this:

Summary Culture, 2005–2025. A set of interface and institutional changes that made the compressed version of a text, meeting, or conversation the default unit of attention. The summary is not a shorter version of the source. It is a different object, optimized for scanning rather than understanding. The shift was gradual, accumulating through design decisions that were individually small and collectively transformative. The result is a two-tier system of reading, in which the source is available but no longer the default.

This is the tone this publication aims for: declarative, evidence-first, elegiac but restrained. The elegy is not for a lost golden age. It is for a specific way of reading that is now harder to reach.

What This Means for Readers

The practical takeaway is not to abandon summaries. Summaries are useful. The takeaway is to notice when the summary has replaced the source, and to make a conscious choice. The choice is not between speed and depth. It is between two different kinds of attention. Both are legitimate. The problem is when the choice is made for you by an interface.

Here are three concrete practices:

1. Read one full source per week. Not a summary. A full article, a full chapter, a full report. Notice what the summary would have left out.
2. Check the default. When an interface shows you a summary, ask: Is the full text one click away? Is the click worth it? The answer will vary. The question is the point.
3. Keep a reading log. Note what you read in full and what you read in summary. Over time, the log will show the ratio. The ratio is a measure of attention, not of intelligence.

Frequently Asked Questions

What is summary culture?

Summary culture is the systematic replacement of primary engagement with compressed, pre-digested versions of texts, meetings, films, and conversations. It is not a single technology but a pattern of interface and institutional changes between 2005 and 2025 that made the summary the default unit of attention.

When did the summary become the default?

The shift was gradual. Key dates include 2007, when Facebook introduced the News Feed with collapsed stories; 2012, when Summly and The Daily Skimm launched; and 2020, when Zoom introduced automatic transcription. By 2023, Gmail’s summary cards had made the email itself a backup artifact.

What is lost when a summary replaces the source?

The loss is not information but sequence. A summary removes the order in which ideas were encountered, the false starts, the digressions, and the moments when the author changed their mind. These are the texture of thought, and they are where understanding lives.

Is summary culture a mistake?

No. It is a tradeoff: speed for understanding. The problem arises when the summary is presented as equivalent to the source, and when the interface makes the source harder to reach. The practical response is to notice the default and make a conscious choice.

Next Step for This Publication

This article is the first in a series on attention compression. The next article will examine a single interface element—the “read more” link—as a primary-source artifact. Future articles will cover the auto-generated video caption, the email summary card, and the book digest as institutional changes. If you have a specific interface decision you would like examined, send a note through the contact page. The question is the beginning of the archive.