#440: Text-First Social Platforms Are Enjoying A Post-Election Surge, & More
1. Text-First Social Platforms Are Enjoying A Post-Election Surge
Real-time, text-first social news platforms like X, Threads, and Bluesky are surging after the US elections, each experiencing growth and engagement in its own way.
Elon Musk and Linda Yaccarino are touting1 record-high post-election engagement and usage on X, formerly Twitter, which—perhaps counterintuitively—might be catalyzing user growth on competing platforms like Threads and Bluesky.
During November alone, Instagram's text-based social network, Threads, added 15 million users and, during the past three months, more than one million per day, according to Adam Mosseri.2 In other words, since mid-August, the number of Threads users has increased nearly 50% to ~300 million.3 Having observed stronger engagement when users encounter fresh content, Threads plans to move away from Instagram's social graph to attract new sign-ups. Threads also is gearing up to introduce ads in 2025.
Post election, a smaller competitor with fewer than 20 million users,4 Bluesky has climbed to the top spot in the Free Apps section of the App Store, outranking both Threads and X. As part of the Fediverse5 along with Threads and others, Bluesky lets users view posts from multiple platforms, which could reshape the social media landscape.
As these platforms evolve and compete, their role in public discourse and news dissemination could increase significantly. A new era in digital communication and information-sharing appears to be taking shape.
2. Tether Has Launched Hadron, A Real-World Tokenization Platform
Last week, Tether, the largest stablecoin issuer, launched Hadron,6 a Real-World Asset (RWA) tokenization platform enabling users to tokenize virtually anything.
We believe that the launch of Hadron represents growing support for the tokenization of RWAs. As one of the largest companies in the digital asset industry, Tether hopes to capitalize on a nascent opportunity, as the total value of tokenized Real-World Assets is only $11 billion,7 a di minimis percentage of the ~$500 trillion total addressable market.8
RWAs are physical or off-chain assets—such as real estate, commodities, art, and financial instruments like bonds or equities—that are represented digitally, or tokenized, on public blockchains. Tokenization involves digital tokens that represent the ownership of these assets so that they can trade and serve as collateral on the Finternet.9 In other words, tokenized RWAs can transform illiquid assets into tradable instruments.
By lowering barriers to entry, tokenized RWAs are democratizing access to investment opportunities and disintermediating traditional middlemen. By purchasing digital tokens, for example, an individual in a developing country could own a fraction of a luxury property in Manhattan or invest in U.S. Treasury bonds.
Real-World Assets represent the deep convergence between traditional finance and digital assets. With regulatory clarity improving and institutional interest growing, we believe RWAs are likely to play a central role in the next wave of the digital asset revolution. In our view, Tether is extremely well positioned to be a leader in this space, as RWAs represent a natural extension of its core business—tying digital assets to fiat currencies.
3. Is AI Progress Hitting A Plateau?
Recent headlines10 suggest that artificial intelligence performance metrics are missing internal targets for next generation models, seemingly hitting a plateau. Our research suggests a more nuanced reality, suggesting steady improvement across multiple metrics.
On coding tasks, for example, foundation model improvements combined with new agent frameworks have produced consistent progress on the SWE-Bench benchmark.11 Agent frameworks combine a base large language model with prompting techniques and programming tools to solve the real-world software development tasks contained within the SWE-Bench benchmark. Anthropic’s latest models now solve 53% of cases within SWE-Bench, a staggering improvement over the ~4% solved by the best solution this time last year, as shown below.
The economics of AI models are evolving along two paths. On the first path, application programming interface (API) pricing, determined by the cost of using AI, has declined substantially, as shown below, enabling democratized access and enterprise adoption.
On the second path, the cost of building AI is growing dramatically, impacting the amount of capital required to train frontier AI models. As a result, AI labs like OpenAI and xAI have tapped private markets for multi-billion-dollar funding rounds at higher valuations. Hyperscalers like Google, Microsoft, Meta Platforms, and Amazon are corroborating the trend by increasing capex plans, currently expected to hit $300 billion in 2025.12
Importantly, among other measures to address the scaling dilemma, inference-time compute13 could curb the rise in training costs. In other words, allocating more compute to inference—letting models “think” longer—is accelerating gains in performance. According to our research, OpenAI and others are pursuing the possibility that inference-time compute will create a new scaling paradigm.
Despite the massive investment required to participate at the frontier of AI performance and the diminishing returns associated with traditional pre-training methods,14 our research on cost declines, informed by Wright’s Law,15 suggests that AI progress has not hit a ceiling.
-
1
Carr, D.F. 2024. “X Traffic Peaked After Election Day in US. So Did Deactivations.” Similarweb.
-
2
Mosseri. 2024. “Huge couple of weeks for Threads…” Threads.
-
3
ARK estimate based on data from Mosseri 2024 and Metha, I. 2024. “Threads now has 275M monthly active users.” TechCrunch.
-
4
Vanian, J. 2024. “X rival Bluesky gains 1.25 million users following U.S. election.” CNBC.
-
5
The fediverse (commonly shortened to fedi) is a collection of social networking services that can communicate with each other (formally known as federation) using a common protocol. See Wikipedia. 2024. “Fediverse.”
-
6
Tether. 2024. ”Hadron by Tether Platform Brings Simplified Asset Tokenization to the Mass Market.”
-
7
The Block Research. 2024. “The Future of Tokenization: How ZKsync is Changing the Game.” The Block.
-
8
This ARK estimate combines the estimated total addressable markets for real estate (~$250 trillion), bonds (~$115 trillion), and equity (~$140 trillion). Estimates are based on data as of November 15, 2024, from Statista. 2024. “Real Estate – Worldwide”; and Sifma. 2024. “Capital Markets Fact Book, 2024.”
-
9
According to the company, the Finternet aims to architect universal infrastructure that’s interoperable, unifies multiple assets, and supports multiple use cases to usher in low-cost, high-volume, and high-trust transactions.
-
10
Chowdhury, H. 2024. “OpenAI is Reportedly Struggling to Improve Its Next Big AI Model. It’s a Warning for the Entire Industry.” Business Insider.
-
11
SWE-bench is a dataset that tests systems' ability to solve GitHub issues automatically. The dataset collects 2,294 Issue-Pull Request pairs from 12 popular Python repositories. Evaluation is performed by unit test verification using post-PR behavior as the reference solution. See Jiminez, C. et al. 2024. “SWE-bench: Can Language Models Resolve Real-World GitHub Issues?” arXiv. doi arXiv:2310.06770v3.
-
12
Butler, G. “Morgan Stanley: Hyperscaler capex to reach $300bn in 2025.” DCD Channels.
-
13
Downing, F. 2024. “OpenAI's o1 Outperforms Other LLMs By ‘Stopping To Think.’” ARK Disrupt Newsletter. ARK Investment Management LLC.
-
14
Laird, J. 2024. “Open AI co-founder reckons AI training has hit a wall, forcing AI labs to train their models smarter not just bigger.” PC Gamer.
-
15
Wright’s Law states that for every cumulative doubling of units produced, costs will fall by a constant percentage. See Winton, B. 2019. “Moore’s Law Isn’t Dead: It’s Wrong—Long Live Wright’s Law.” ARK Investment Management LLC.