Sam has increased his agency and willingness to pursue diverse projects over the past 10 years, taking on things like Stargate, recruiting Fiji, and Johnny IVr—a shift suggesting either changed constraints, increased confidence, or reduced social inhibition.

AI Explainability Is Critical and Dangerous Gap

One of the most dangerous aspects of advancing AI is that many algorithms cannot explain why they made decisions or what criteria they used, so as both Twitter and society export more decisions to algorithms, the field of explainability—being able to state the 'why' behind every action—becomes critical.

Working in Public Modeled on Open Source

Dorsey prefers to work, make decisions, and make mistakes in public—a philosophy drawn from punk rock, hip hop, and open source where performers improve iteratively in front of audiences—and applies it to Twitter, accepting bruises in the short term to earn trust in the long term.

No Edit Button Due to Real-Time Fan-Out

Twitter lacks an edit button because the system was built on an SMS-style real-time fan-out where damage is largely done at send, and editing introduces problems: a delay harmful to real-time uses like NBA twitter, the risk of bad actors altering already-retweeted content (requiring change logs), and the legitimate need for clarifications which the company wants to solve deliberately.

Impartiality Over Neutrality

Twitter can no longer afford to optimize for neutrality—a passive, hands-off stance—but should optimize for impartiality, because people weaponize free expression to silence others (via doxing, troll armies, and threats of violence), especially targeting marginalized members of society.

Host Control Over Replies Tradeoff

Twitter is considering giving tweet authors host-like control to hide (not delete) replies, but this creates a tradeoff: it benefits the author yet risks creating filter bubbles and could let powerful figures heavily moderate their reply spaces, removing speech that holds power to account—so any such action would have to be transparently labeled as moderated by the author.

Like Button Is Empty and Destructive

The 'like' button, drafted reactively from Facebook and Instagram behavior, is empty and more destructive than understood, and a button expressing gratitude ('thanks') or learning ('changed my mind') would better incentivize valuable contribution to conversation.

Follow Mechanic Creates Echo Chambers

Twitter's single core action of following accounts creates isolated filter bubbles and echo chambers, because following accounts on one side of an issue (e.g. Brexit's vote leave) exposes you only to that perspective unless you do the deliberate work of following opposing accounts, which almost no one does.

Twitter Is Interest-Based Not Social Network

Twitter is not a true social network because it does not depend on the graph of people you know (the phone address book), but is instead an interest-based network where simply talking about a topic places you in a community without any deliberate join or leave action.

Journalist Following Asymmetry in 2016 Election

Research on Twitter behavior during the 2016 election showed that very few left-leaning journalists followed people on the right, while right-leaning journalists followed people on the left at extremely high rates, revealing an asymmetric filter bubble in the media sphere.

Square Enables Financial Inclusion for Underserved

Square addresses economic exclusion by using technology to give underserved people access to digital financial systems that previously required good credit scores, with the Cash App serving as the only bank account for a significant percentage of its users.

Character Constraint Drove Headline Outrage Culture

Twitter's character constraint attracted comedians and the hip-hop community for its rhythmic nature but also negatively fostered a culture of headline-driven outrage and fast takes, and the expansion to 280 characters added nuance mainly in replies rather than original broadcasts.

Twitter Unlocks Journalists From Publications

Twitter freed journalists from their publications by letting audiences follow them as individuals rather than institutions, so a journalist's direct connection with readers and sources persists as they move between outlets, which journalists found liberating.

Harassers Are Predictable Repeat Offenders

Harassers who sling slurs or dox in someone's replies have a high probability of doing it to other people too, so detecting that behavioral pattern allows Twitter to predict it and add friction to the spread of those tweets into shared areas like replies, search, and trends—without removing the content.

Conversational Health Measured Like Body Temperature

Twitter is developing a framework to measure 'conversational health' analogous to bodily health indicators like temperature, using four placeholder indicators—shared attention, shared reality, receptivity (toxicity), and variety of perspective—recognizing these exist in tension so optimizing one (e.g. variety of perspective) can degrade another (e.g. shared reality), with the goal of keeping them in balance rather than maximizing one.

Earned Audience Sees Everything

When someone explicitly follows an account or topic, that audience is earned and they will see every single tweet from that account with no algorithmic suppression; downranking and friction only apply in shared spaces like replies, search, and trends where anyone can inject themselves.

Toxicity Solved by Product Not Policy

Twitter will not solve its toxicity problems by changing or adding policy—a proliferation of case-specific rules is a weak, unnavigable position—but by looking at the product itself, the incentives it creates, and its role in what it recommends, amplifies, and downranks.

Pseudonymity Over Real Names

Requiring real names does not solve toxicity because platforms with real-name policies show the same problems, so Twitter aims to incentivize pseudonymity and built reputation rather than real identity, while using biometrics to label verified humans rather than fighting the losing battle of detecting ever-more-sophisticated bots.

Suspensions Based on Background Behavior Not Single Tweets

Twitter does not permanently suspend people for saying one particular thing (except specific violent threats with a location); permanent suspensions usually result from background behaviors such as controlling multiple accounts to harass the same people or ban evasion, but the company has failed to explain this because it lacks a transparent case-law-like system.

Voter Suppression Misinformation Scariest Case

The scariest misinformation is content intended to deceive people into off-platform actions, exemplified by a 2016 tweet with a fake number to 'register to vote' by text; while crowd wisdom corrected it (debunking tweets got 10x the impressions of the original), Twitter cannot rely on the crowd and must act far faster.