Creator Economy & Web3
Web3’s relationship to the creator economy is bifurcating: as AI algorithms tighten control over content distribution and IP, decentralized tools offer a structural alternative. The case isn’t primarily technical — it’s about attention sovereignty, energy politics of streaming, and who owns the training data that AI was built on. (→ [[devconnect-argentina]])
The Algorithmic Attention Crisis
Adrián Garelik (Flixo) argues AI already won the attention war — not through robots but through recommendation algorithms deployed 15 years ago. (Devconnect Argentina, The Creator Economy)
The progression:
- Meta, TikTok, YouTube collect faces, contacts, and video to train AI systems
- Content is evaluated on “effectiveness” (engagement maximization), not quality (promotes thinking)
- Prediction: within 5–10 years, AI generates fully personalized VR content to maximize engagement indefinitely
The “generation swipe-up” phenomenon:
- Young people condition to 3-second content confirmation via algorithmic feeds
- Mental health peaks with iPhone front camera (2008) and Instagram launch (2012)
- Current generation shows highest-ever anxiety, depression, and suicide rates
The carbon dimension:
- Streaming consumes ~3% of global electricity — a number conspicuously absent from public debate while Bitcoin’s energy use receives constant scrutiny
- AI data centers require massive electricity; Bill Gates shifted carbon-zero deadline from 2030 to 2045 under this pressure
Decentralized Distribution: Flixo
Flixo is a decentralized film financing and distribution platform built on B2P2P (Blockchain Peer-to-Peer) streaming. (Adrián Garelik)
Energy case:
- B2P2P streaming reduces energy consumption by 95% compared to centralized CDN streaming
- Centralized streaming: data center → CDN → viewer (three energy-intensive hops)
- B2P2P: content cached and served by viewers themselves (one hop, no data center)
Economic model:
- “Death of a Comedian” film had 10,190 co-producers via tokenized financing
- NFT-based “Ticket 3.0”: holding the NFT grants exclusive streaming access (3 months on Flixo before Amazon Prime availability)
- Decentralized financing removes studio gatekeeping while creating community ownership
Sovereign digital identity:
- The goal isn’t to replace streaming — it’s to add privacy-preserving identity layers
- Show only what’s required at the door (age verification) without revealing full identity
- Blockchain enables contextual disclosure without credential aggregation
IP in the Age of AI
Panel discussion (Isha Kim, Ezequiel Canle, Gabriela Cortes) on how intellectual property law applies when AI is both a tool and a potential infringer. (Devconnect Argentina, The Creator Economy)
The Gema case (Germany): German collecting society Gema filed against OpenAI, with a court ruling that OpenAI owes compensation for training on artist works without licensing. This is the first major European precedent.
Key legal questions unresolved:
- Who owns the training data? (creators, platforms, or ML companies?)
- Did artists consent to AI training when they published work?
- What are the payment obligations for AI training?
- Is style protectable? (Generally no — but specific stylistic expressions may be)
Style vs. work protection:
- Style is generally not protectable under copyright law (any jurisdiction)
- The specific expression of style in a work is protectable
- Exception emerging: where a model’s output is substantially similar to a specific training work, courts may find infringement
The authorship shift: The industry is transitioning from “author/authorship” to “content creators” (TikTok reels to fine art, same label). This terminological flattening has legal consequences: creator platforms may not provide the same IP protections as traditional publishing.
Co-Authorship with AI
The emerging norm (Isha Kim): Artists increasingly co-author with AI, crediting it in the credits. The anxiety among educators is real, but the consensus is forming: “creators must learn to use AI or others will use it on them.”
Blockchain for provenance:
- LoRA (Low Rank Adaptation) weights stored on-chain: training data provenance becomes auditable
- Creators can prove which works contributed to which model weights
- Enables licensing retroactively even for models already trained
Why this matters: The off-chain world has no chain of custody for AI training data. Blockchain provides the immutable record that copyright law requires to prove infringement — or to license proactively.
Web3 as Structural Alternative
The Creator Economy track argues Web3 isn’t competing with Web2 distribution — it’s building a parallel layer with different properties:
| Property | Web2 (Algorithm) | Web3 (Blockchain) |
|---|---|---|
| Content ranking | Engagement maximization | Creator-set rules |
| Identity | Platform-owned | Self-sovereign |
| Monetization | Ad revenue share | Direct-to-fan (NFTs, tokens) |
| Financing | Studio/platform gatekeeping | Community co-production |
| IP record | Platform ToS | On-chain provenance |
| Energy model | Centralized CDN (high energy) | B2P2P (95% reduction) |
Critical caveat (from panels): Web3 tools are currently for early adopters. The “normies” still discover content via TikTok and YouTube. The bridge is hybrid models: Web2 distribution → Web3 monetization and ownership (the “Ticket 3.0” NFT model).
Connections
- Cypherpunk Values & Philosophy — Attention sovereignty and algorithmic control are the mass-market version of the surveillance capitalism critique; Garelik’s framing is the accessible version of Eva Galperin’s cyberfascism analysis
- On-Chain Agents — AI agents amplify both the attention economy problem (algorithmic curation at machine speed) and the Web3 solution (provenance-verified, agent-accessible content markets)
- Decentralisation Accelerationism (d/acc) — d/acc shares the structural critique of algorithmic concentration; Garelik’s “AI won 15 years ago” resonates with dacc’s defensive tech framing
- Metadata Privacy — Sovereign identity (show only what’s needed at the door) is the creator-economy expression of metadata minimization
- zkTLS Infrastructure — ZK proofs of LoRA weights or training data provenance are a natural application of zkTLS to IP verification
Open Questions
- Does the Gema precedent propagate to US courts, or remains a European-only constraint on AI training?
- Can B2P2P streaming achieve sufficient reliability and CDN-like performance for mainstream video?
- At what point does the “generation swipe-up” crisis force platform-level intervention — and what form does that take?
- Is the “Ticket 3.0” model (NFT-gated exclusive streaming windows) a durable economic model, or a novelty?
- Does on-chain LoRA provenance create enforceable licensing claims, or remain a provenance curiosity without legal teeth?