OpenAI's Sora app and the threat to social media's survival

Executive overview

Social media giants built durable advantages from two things: the right users and painstakingly constructed social graphs. When they abandoned those advantages to chase TikTok's pure-engagement model, they made themselves vulnerable to anything more engaging that came along.

OpenAI's Sora app is that thing. It offers algorithmically curated AI-generated video — the same brain-stem stimulation as TikTok, but without the constraints of reality or human creators.

Each platform's only viable defence is to retreat behind the social graph and curated user base it already has — the one asset no new entrant can replicate.

Why the incumbents were once unassailable

• Facebook's moat: the specific people users knew were already there, plus a dense social graph built from billions of manually confirmed friend relationships
• Instagram's moat: a large installed base of expert and influencer creators producing visually compelling content, plus user-curated follow lists
• Twitter's moat: culturally interesting people (reporters, comedians, commentators) in an exponential follower graph that acted as a distributed human curation machine
• All three required new entrants to replicate both the right user base and the social graph — effectively impossible at scale

How TikTok bypassed every competitive moat

• TikTok ignored who its users were and who knew whom — it needed neither famous people nor friend graphs
• Its recommendation algorithm surfaced compelling content without any user effort
• The short video format worked at a primal, pre-social level — visual novelty and direct emotional expression
• When incumbents chased TikTok, they walked away from their own moats and entered the open plain of pure engagement
• On that plain, they now compete with every other attention source: Netflix, podcasts, slot machines — and Sora

What Sora changes

• Sora combines a TikTok-style feed with AI-generated video: any content imaginable, on demand
• Human creators are constrained by location, budget, equipment, and risk; Sora has none of those limits
• The result can be more visually arresting than anything a typical TikTok creator can film
• OpenAI launched with minimal IP or content protections to generate viral attention — a deliberate, cynical move
• A key constraint: generating video is computationally expensive; $20/month ChatGPT Plus gives only 50 low-resolution videos, and serious use requires $200/month

Platform-by-platform survival strategy

• Twitter/X: stop the algorithmic video-autoplay pivot; recommit to interesting people saying interesting things; build a transparent, balanced content-moderation council; a $3–4B revenue business is achievable and AI-proof because distributed human wit cannot be replicated by algorithm
• Instagram: retreat from algorithmically served Reels; use the algorithm only to suggest new creators to follow; curated expert content chosen by the user is differentiated from slop
• Facebook: return to the friend graph — show what people you know and organisations you chose are doing; lean back into the original social compact of mutual attention
• TikTok: may survive this wave because Sora's economics are prohibitive for most creators, limiting the inventory available to its recommendation engine; creators will strip watermarks and repost Sora clips to TikTok anyway; but the pure-engagement model means more waves will follow

Sora's economic Achilles heel

• Recommendation algorithms need inventory — the more potential content, the better the match for each user
• High cost per video means far fewer creators, which means thin inventory relative to TikTok
• TikTok's computation is distributed to users' own phones; Sora burns expensive GPU time on every generation
• Fewer creators also means jailbroken Sora clips flow onto TikTok, boosting TikTok's own inventory

Will Sora kill film and television?

• Lower-quality, high-engagement content has always coexisted with higher-quality work — TikTok didn't kill Oppenheimer
• Sora occupies the same stratum as TikTok; it does not threaten the stratum of auteur filmmaking
• The complexity of producing a great film is not reducible to a text prompt
• Historical pattern: the introduction of a more engaging lower stratum has never collapsed the strata above it

What Sam Altman's announcements signal

• OpenAI is pivoting toward an attention/ads economy: TikTok-style video feed, relaxed chat personality constraints, planned adult content for verified users
• This shift away from "we will automate white-collar work" rhetoric suggests GPT-5 did not produce the expected capability leap
• Agents built on LLM queries remain unreliable — no one has successfully automated complex workflows
• Circular investment deals between AI infrastructure companies echo patterns seen before the dot-com collapse
• The pivot to brain-stem engagement looks less like a strategic choice and more like a search for any revenue model that works

Defensive strategies for social media alternatives

• Newsletters and indie media: built on webs of trust rather than algorithmic curation; each link in the chain is a real, verified relationship
• Substack lowers the barrier to newsletter publishing but its algorithmic discovery features undermine the trust-chain model
• Podcast hosting is structurally cheap (files on a server, RSS feed); email delivery at scale is expensive
• The ideal outcome: if all incumbents chase Sora-style engagement and fail, users may migrate to higher-quality indie media — podcasts, newsletters, long-form writing