Gutenberg Lab — Morning Paper

The Rise of "Machinic Taste" and the Balkanization of Internet Culture

Featuring

Sofia Du × Sofia Puccini (MTS Podcast) × Ruby Justice Thelot (Designer, Cyber Ethnographer, NYU Professor)

Bottom line

The cultural crisis of the AI era has a dual structure: machines are beginning to develop their own aesthetic preferences, while human communities are fragmenting into mutually incomprehensible languages.

3-Line Summary

• Cyber ethnographer Ruby Thelot analyzes how AI and algorithmic platforms are fundamentally reshaping human culture, taste, and communication
• Americans claim to dislike "AI" while actively using ChatGPT—a language problem, not a utility problem, revealing a branding crisis for the industry
• Drawing parallels to 17th-century England, Thelot argues we need new frameworks for virtuous consumption amid AI-driven wealth creation

Three Key Points

1. The Emergence of Machinic Taste: With approximately 50% of internet traffic now bots (Cloudflare data), content creators face what Thelot calls a "three-legged waltz"—audience capture, algorithmic capture, and agentic capture. She predicts AI-generated content for AI audiences will reveal machines' inherent aesthetic preferences, independent of human input—a fundamental shift from human-centric culture

2. Balkanization and Babelification: The internet has fragmented from monoculture (5-10 TV channels everyone watched) into isolated cultural enclaves. Communities develop unique vernaculars like Galapagos species—the same word means different things across furry, incel, and high-fashion communities, making cross-community communication structurally impossible

3. Taste as Virtue, Not Preference: After the 1688 Glorious Revolution, British colonial wealth created a nouveau riche class. Philosophers feared excessive luxury would cause societal decay. The Earl of Shaftesbury defined "taste" as "harmony with nature"—a framework for virtuous consumption. Thelot argues we're in a parallel moment: an industrial revolution, immense wealth creation, and the need to answer how to deal with abundance without wrecking society

Editorial Perspective

What strikes me about Thelot's analysis is the reframing of taste from aesthetic preference to moral imperative. Just as 17th-century England needed "taste" to navigate nouveau riche excess, we need ethical frameworks for what to build and consume in the AI era. The stated-versus-revealed preference split—Americans dislike "AI" but love ChatGPT—isn't a technology problem but a narrative problem. The industry must reframe AI from "job-replacing threat" to "practical utility." Equally significant is the "babelification" of the internet into isolated cultural islands where the same words carry incompatible meanings. This fragmentation is a structural side effect of platform economics that cannot be ignored.

Source: a16z Podcast "Ruby Thelot on Internet Culture, AI, and the Future of Taste" (August 2, 2026)

https://a16z.simplecast.com/episodes/ruby-thelot-on-internet-culture-ai-and-the-future-of-taste-CvhHSpzt

Spoken source: a16z Podcast "Ruby Thelot on Internet Culture, AI, and the Future of Taste" (August 2, 2026). Guest: Ruby Justice Thelot

AI Disclosure: Produced with AI assistance; facts and analysis reviewed by our editorial team.

• --

Bitcoin's Mining Model Meets AI Inference—Bittensor's Permissionless Market Experiment

Featuring

Jason Calacanis × Jacob "Const" Steves (Bittensor creator)

Bottom line

The key to decentralizing AI inference scarcity is building the same permissionless competitive markets that made Bitcoin work.

3-Line Summary

• Bittensor applies Bitcoin's proof-of-work mining model to AI inference and training, aggregating global excess compute without gatekeepers
• 128 subnets (specialized networks for inference, training, storage) issue their own tokens and compete on price to stay in the top ranks
• The NG subnet already offers inference at ~50% the cost of competitors, and Jason uses it in production

3 Key Points

1. Measured cost advantage: The NG subnet provides inference for Qwen, Kimmy, and GLM-5.2 at roughly half the price of competitors. Jason uses it in production via OpenRouter, and Const explains that "permissionless entry lets anyone with excess compute join instantly, creating liquid supply that undercuts centralized providers"

2. Subnet 51 revenue scaling: After tripling miner payments, compute resources tripled in two months—proof that "raise rewards, supply increases" market dynamics work in practice

3. Unsolved decentralized training problem: Training trillion-parameter models requires terabyte-scale weight synchronization per step. Const admits "how we stitch together all the compute in a way that gets around the bandwidth problem" is "the holy grail," but no solution exists yet

Editorial Perspective

Bittensor's bet is that market competition beats capital concentration. Bitcoin proved that permissionless entry can create the world's largest supercomputer. If that replicates for AI inference, OpenAI and Anthropic's capital advantage gets relativized. But decentralized training (model training itself) remains blocked by bandwidth limits, and success in inference markets doesn't automatically transfer to training markets. Const's design philosophy—"resisting adversaries is everything, the rest is fluff"—is faithful to blockchain's core insight: coordination through incentives, not trust. The real test is whether subnet 51's revenue scaling (3x compute in 2 months) can extend to the "holy grail" of training trillion-parameter models without centralized infrastructure.

AI Disclosure: Produced with AI assistance; facts and analysis reviewed by our editorial team.

Source: This Week in Startups, "Bittensor creator Const on Affine, dTAO, 'mining reasoning,' and more | E2326" (August 17, 2026)

https://4a885955-6823-4b22-a3e1-a526c25516a5.libsyn.com/bittensor-creator-const-on-affine-dtao-mining-reasoning-and-more-e2326

• --

The Repair Shop Owner Who Can't Keep Up With His Own Viral Success

Featuring

David Pierce (host) × Leo Mastrolia (owner, Digital Doctor Repairs; content creator)

Bottom line

Content drives customers, customers overwhelm operations, operations prevent content — and one person cannot sustain this loop.

3-Line Summary

Leo Mastrolia, 23, opened a repair shop during peak COVID and has since fixed 25,000 devices while building millions of followers through Meta glasses POV videos. Today he faces a structural paradox: 30 devices arrive daily, 300 sit in backlog, and 90% of customers come because they saw his content — but filming stops repairs, and repairs stop filming. Manufacturers restrict parts and refuse to share schematics, leaving independent shops supply-constrained despite overwhelming demand.

3 Key Points

1. Content-as-acquisition creates operational collapse: 90% of Leo's customers come from watching his videos, some driving 1.5 hours from New York City. But filming prevents him from clearing the 300-device backlog and 30 daily arrivals. Viewers in the Philippines and India will never visit but keep watching. He's one person trying to be both local service provider and global media company.

2. Repair demand migrates every 2-3 years, not because devices fail: A year ago, iPhone 13 was 10-15 repairs per day; now it's down to 1-2 per week. Current volume is models 15, 16, 17. Despite improved durability, consumer upgrade behavior hasn't changed. Leo observes "phones are top priority, laptops are lower."

3. Manufacturers weaponize serialization and withhold schematics: iPhone proximity sensors and batteries are serialized to specific devices — if Leo rips a sensor during battery replacement, Face ID breaks permanently even with genuine Apple parts. Motherboard schematics aren't provided to independent shops, forcing him to "guess power levels by comparing similar boards." Under a microscope, iPhone boards look like "an entire city" with components smaller than tweezer tips.

Editorial Perspective

Leo's story reveals a structural problem typical of platform transitions. When content becomes the customer acquisition engine, service businesses carry dual expectations — local customers and global audiences. But one person cannot serve both, forcing a choice: abandon the shop, disappoint the audience, build a team, or invent a new model. Meanwhile, manufacturers restrict repair access while independent shops face overwhelming demand. Leo says he's "not worried about competition — more people should enter this field," precisely because demand far exceeds supply. The repair sector is supply-constrained, not demand-constrained, in an era when every household has 10-15 devices.

AI Disclosure: Produced with AI assistance; facts and analysis reviewed by our editorial team.

Source: The Vergecast, "The Digital Doctor will see you now" (August 17, 2026)

https://www.theverge.com/the-vergecast

• --