From Google to Decentralized AI: How Jacob Robert Steeves Built Bittensor's Incentive-Driven Network

Jacob Robert Steeves didn’t set out to revolutionize artificial intelligence through blockchain. His journey began in the most unlikely place—with Bitcoin, a brain-computer interface chip company, and mathematics. Today, as the founder of Bittensor (TAO), jacob robert steeves stands at the intersection of two transformative technologies, applying the mining economics that made Bitcoin revolutionary to the computational challenges of modern AI.

The Mathematician Who Left Google: Jacob’s Journey into AI and Bitcoin

Before launching Bittensor full-time in 2018, Jacob studied mathematics and computer science at Simon Fraser University in Vancouver, Canada. After graduation, he worked as a software engineer for a DARPA contractor developing brain-computer interface chips—an experience that shaped his fundamental understanding of computation and incentive systems. His mentor, also the company founder, was an early Bitcoin advocate who introduced Jacob to concepts like “energy-based computation” and thermodynamic principles embedded in Bitcoin’s design.

This early exposure proved transformative. Jacob realized that AI and Bitcoin shared a common DNA: both operate through feedback loops. AI learns through backpropagation, genetic algorithms, and reinforcement learning—all feedback-driven processes—while Bitcoin created the first programmable economic feedback loop at scale. Since 2015, jacob robert steeves has been immersed in both fields simultaneously, recognizing their natural compatibility rather than their apparent differences.

His Google tenure from 2016 onwards deepened this technical foundation. As a machine learning engineer, Jacob witnessed the publication of “Attention Is All You Need”—the Transformer paper that ignited the exponential expansion of large language models. More importantly, he absorbed practical knowledge from Google’s distributed AI infrastructure: parameter servers, model parallelism, and data parallelism techniques that would prove essential for Bittensor’s architecture. Yet while Google offered prestigious projects, it could not offer what Jacob truly wanted: the ability to apply decentralized incentives to AI at the network level. This realization led him to develop Bittensor initially as a side project, eventually becoming his full-time focus when he launched the mainnet in 2021.

Mining Economics Applied to AI: Bittensor’s Core Innovation

At its core, Bittensor represents a direct translation of Bitcoin’s mining philosophy to artificial intelligence. Jacob describes it as applying “Bitcoin-style mining incentive mechanisms” to AI computation—but this requires precise understanding. Bittensor is an open-source protocol with its native token TAO, currently operating approximately 128 specialized subnets. Each subnet organizes around distinct computational tasks: inference, training, reinforcement learning, coding agents, storage, and prediction/trading signals.

The fundamental innovation lies not in aggregation but in programmable incentives embedded directly into the learning process itself. Whoever provides more useful inference, training, or tools receives proportionally higher rewards. This creates a continuous optimization loop: market signals drive quality improvements, and poor-performing supply is naturally eliminated through economic pressure rather than administrative decree. The system transforms the traditional “miner-reward-consensus” paradigm into “useful AI supply-market reward-network consensus.”

From a practical perspective, developers can initiate or join subnets, contribute computing resources and models, and continuously earn incentives based on performance metrics. Demand-side participants can purchase inference services, computational power, AutoML capabilities, or prediction signals directly through the network. The entire structure remains permissionless and transparent, allowing any developer globally to participate fairly.

This represents a fundamental departure from traditional AI aggregation platforms that simply stack models together without economic optimization. Jacob emphasizes that the true significance extends beyond “Crypto + AI”—a phrase he considers intellectually superficial. The real innovation is using crypto-economic incentives to conduct artificial intelligence research itself, letting market forces continuously refine computational quality.

Chinese Developers and Fierce Competition: Building Asia’s AI Network

Jacob’s decision to visit China in late 2024 reflected a strategic recognition: Asia hosts the world’s fastest-growing and possibly most powerful artificial intelligence ecosystem. China alone produces 90% of the world’s semiconductor chips. When Bitcoin mining operated legally, China controlled over 50% of global mining power. These metrics underscore why Jacob sees China not as a peripheral participant but as essential infrastructure for Bittensor’s global network.

What strikes Jacob most about Chinese developers is not merely their technical capability but their competitive intensity. Within Bittensor’s subnets, an observable phenomenon emerges: once Chinese miners enter a subnet, competition immediately intensifies dramatically. Many original participants choose to exit, not due to technical disadvantage but because the competitive dynamics become substantially fiercer. Jacob views this as entirely expected, given China’s university system’s competitive training structure producing “among the most competitive groups in the world.”

The concrete evidence validates this observation. Affine, one of Bittensor’s largest subnets, was built entirely by Chinese developers and has become one of the network’s most sophisticated competitive mechanisms. Lium, another major subnet providing GPU resources, demonstrates how Chinese computing infrastructure integrates into Bittensor’s permissionless marketplace. Many Chinese miners contribute GPU computing power (identifiable by IP addresses showing Asian origin), effectively bringing Asia’s computational resources to the global market through decentralized infrastructure.

These contributions represent something Jacob considers “very significant”—not simply technical participation but fundamental contributions to network resilience and competitive optimization. The engineering level among these teams is “extremely high, almost second to none,” according to Jacob’s direct assessment.

Beyond Aggregation: Why Bittensor is Fundamentally Different

Jacob directly addresses a persistent misconception: that Bittensor functions as an “AI model aggregator” that combines existing services. This misunderstanding fundamentally misses Bittensor’s architectural innovation. True aggregation platforms simply combine models without structural incentives for continuous improvement. Bittensor’s design embeds economic incentives directly into the feedback loops that drive AI learning itself.

The 15-year trajectory of AI advancement reveals a consistent pattern: breakthroughs emerge from adaptive learning based on feedback and rewards. Backpropagation, reinforcement learning, and other foundational techniques all operate through this principle. Bittensor’s innovation consists of embedding cryptocurrency and market incentives directly into these mechanisms, allowing real-time market signals to optimize both supply quality and network efficiency.

Decentralization serves essential functions within this framework. Permissionless entry means any individual or team can launch a subnet and compete directly with existing services. Good supply gets amplified by economic incentives; poor supply gets naturally eliminated. Resource distribution across nodes creates resilience against single points of failure—a property starkly illustrated when AWS experienced large-scale outages in recent months. While many projects claiming decentralization suffered serious disruptions, Bittensor’s distributed architecture remained operational precisely because it doesn’t depend on centralized infrastructure providers.

However, Jacob emphasizes that decentralization represents a means rather than an end. The fundamental driving force isn’t decentralization “for its own sake” but rather scaling useful computation through incentive-driven competition. This distinction proves crucial: Bittensor competes against traditional centralized AI platforms not through ideological preference for decentralization but through superior technical primitives and mechanism design.

Protocol Revenue, Prediction Markets, and Five-Year Goals

Bittensor’s economic sustainability flows from multiple revenue sources at the protocol level. The network generates income through selling inference services, computational power access, AutoML (automated machine learning) capabilities, and signals provided to prediction markets. This diversified revenue model prevents over-dependence on any single use case while creating multiple incentive streams for network participants.

Jacob displays particular enthusiasm for prediction markets as a breakthrough application. Platforms like Kalshi and Polymarket represent “real fintech applications” and “first consumer applications” that genuinely reshape human decision-making processes. Within Bittensor’s ecosystem, dedicated subnets are developing prediction market infrastructure, demonstrating the protocol’s capacity to support sophisticated financial applications.

Looking forward five years, Jacob articulates a singular aspiration: bringing Bittensor technology to “millions” of users while maintaining sustainable network operations. Currently, approximately 100,000 users actively utilize Bittensor’s technology. The pathway forward involves not only dominating inference services but expanding toward the application layer—aiming ultimately to serve billions of users globally.

The economic advantage underpinning this vision centers on cost-efficiency. Ridges, a major Bittensor subnet focused on coding agents, demonstrates this principle. By leveraging distributed optimization across worldwide miners, the network achieves dramatic cost reductions: scenarios where centralized providers charge $1,000 monthly subscriptions for $200 of actual value can be undercut with $10 network prices reflecting only $6 in actual costs. This economic scaling law—impossible within centralized architectures—enables global reach and adoption.

Jacob draws a historical parallel to Bitcoin’s success: Bitcoin outperformed centralized systems not through ideology alone but by adopting superior technical primitives and mechanism design. While Bittensor hasn’t achieved this advantage across all domains, it has demonstrated this principle in specific computational areas. Moreover, millions already use Bittensor services daily without conscious awareness—the network operates transparently across application layers.

The competitive dynamic ultimately reduces to a simple proposition: if Bittensor maintains technological superiority in key dimensions like performance, speed, and cost-effectiveness, centralized AI providers operating under traditional infrastructure economics cannot ultimately compete. Conversely, failure to maintain these technical advantages would render the entire premise moot. This clarity of purpose—achieve genuine technical superiority or fail meaningfully—defines Jacob’s strategic vision for Bittensor’s five-year trajectory.

Market Evolution and the First Halving Cycle

As of March 2026, Bittensor (TAO) reflects the market dynamics inherent to any successful protocol. The token, initially listed in March 2023, has experienced market cycles reflecting broader cryptocurrency and AI sector dynamics. Current market conditions show TAO trading at $197.10 with a flowing market capitalization of approximately $1.89 billion. This represents the natural price discovery process in permissionless markets.

Jacob’s perspective on 2025’s first TAO halving—now completed—emphasizes supply dynamics rather than speculative narratives. The halving tightens token supply, but Jacob explicitly states this alone does not alter Bittensor’s fundamental incentive mechanisms. Economic incentives for developers remain robust regardless of token supply schedules. The network’s core value proposition depends on computational utility and market-driven reward distribution, not on supply scarcity alone.

This measured perspective reflects Jacob’s engineering mindset rather than trader psychology. The focus remains on network utility, developer incentives, and competitive optimization—the metrics that determine genuine technology adoption rather than speculative token price movements.

Asia’s Strategic Position: Open-Source AI and Decentralized Infrastructure

Jacob perceives a decisive strategic shift in artificial intelligence development. China, Singapore, and East Asia collectively lead what he characterizes as the “open-source AI revolution.” Top open-source models including DeepSeek originate from Chinese teams. Hong Kong and Singapore, with superior regulatory flexibility and capital infrastructure, facilitate industrialization and cross-border technical collaboration. This regional dynamic creates natural compatibility with Bittensor’s decentralized model.

Beyond commercial development, top Asian universities including Peking University and Tsinghua University have made fundamental contributions to AI academic advancement. This combination—open-source models, engineering implementation focus, and academic rigor—aligns precisely with what decentralized AI requires: transparent development, competitive optimization, and deep technical competence.

Vision Beyond Partnership: Competition and Technical Primitives

When asked about cooperation possibilities with traditional AI laboratories and major technology companies, Jacob maintains philosophical clarity. Partnerships with teams like DeepSeek, Kimi, and Moonshot appear naturally compatible—these organizations can launch subnets on Bittensor, monetize their models through the network, and simultaneously consume network-provided services. Centralized American laboratories, by contrast, “prefer to consolidate and control” rather than embrace openness and permissionless participation.

Jacob frames this not as hostility toward traditional AI companies but as technical inevitability. Either centralized providers adopt Bittensor’s decentralized training approaches, or they face long-term competitive disadvantage as cost structures and performance metrics increasingly favor incentive-driven, distributed computation. The outcome ultimately depends on technological execution rather than market positioning or capital advantages.