In the second episode of “The TAO Pod,” recorded on July 15, 2025, hosts James Altucher and Bittensor expert JJ explore the transformative potential of Bittensor (TAO), a decentralized protocol revolutionizing AI through incentivized commodities. As TAO trades around $340 as of September 2025, with its first halving approaching in less than 90 days and the AI token market reaching $26.4 billion, the discussion highlights how Bittensor commoditizes AI resources—compute, data, and models—making them accessible and abundant. Drawing parallels to open source and Bitcoin, the episode positions Bittensor as an upgrade to traditional capitalism, blending intrinsic and extrinsic incentives for global innovation.
Decentralized AI: From Scarcity to Abundance
Bittensor operates as an infrastructure protocol, incentivizing participants to produce “digital commodities” like datasets, compute power, and trained models. JJ explains, “Bittensor enables every one of these scarce digital products to be abundant, decentralized commodities.” Unlike centralized AI labs requiring billions in capital, Bittensor aggregates resources permissionlessly, subsidizing costs for developers.
James likens Bittensor’s stage to the internet in 1991—powerful but lacking user-friendly front ends. He notes, “There’s not really a sense that, oh, we need to make a nice front end.” Yet, as subnets mature, Bittensor could democratize AI, turning proprietary tools into open commodities, much like how technology deflates over time.
Real-World Use Cases: Building AI for Emergency Rooms
The hosts brainstorm a practical application: an AI model for ER doctors analyzing incoming patient data (videos, audio, text) to provide instant diagnostics. James pitches, “I have a labeled dataset of videos… of patients coming into an ER… I wanna train an AI model so that when I have video of a new patient, it knows what to do.”
JJ outlines how Bittensor fits: Use Subnet 56 (Gradients) for training on open models like Llama 70B, uploading custom datasets while retaining control. Store data on decentralized subnets like Hippius for cheaper than AWS. For inference, leverage Subnet 64 (Chutes) or Subnet 4 (Targon), which handles video and text. Validators ensure quality, categorizing inputs (e.g., bullet wounds vs. heart attacks) via frame-by-frame analysis.
Without Bittensor, this requires hiring developers, buying GPUs, and building infrastructure—costing millions. With it, costs drop to 1/100th, as James notes: “I could do all this for probably one 100th of the price.” End-users like doctors interact via seamless interfaces, unaware of the underlying protocol.
Beyond Open Source: Bittensor as an Economic Upgrade
JJ boldly claims, “Bittensor is more disruptive in terms of the commoditization of technology than open source.” Open source relies on intrinsic motivations—love of creation, learning—lacking direct economics. Bittensor adds extrinsic rewards via TAO and subnet alpha tokens, creating liquid incentives.
James illustrates with Google’s history: Early search engines tweaked open-source web spiders. On Bittensor, subnets could incentivize faster spiders and better categorization algorithms, with miners competing and validators verifying. This fosters massive community-driven improvements, enabling anyone to “make a Google for almost nothing” using open source plus Bittensor’s incentives.
The protocol records incentives on-chain but computes off-chain, mirroring Bitcoin’s efficiency. JJ emphasizes, “Bittensor is literally just an incentive language,” scaling compute far beyond centralized data centers—thousands of times more than global supercomputers combined.
Spotlight on Subnets: Templar and the Path to Superintelligence
Diving into Subnet 3 (Templar), JJ describes it as a distributed training network aiming for multi-trillion-parameter models. Currently at ~70 billion parameters, its miners optimize loss functions across heterogeneous hardware worldwide. Though not yet revenue-generating, success could commoditize training, surpassing centralized labs.
JJ notes NVIDIA’s Jensen Huang acknowledging the need for distributed training to scale AI. Templar’s permissionless approach could achieve this at internet scale, potentially reaching 70 trillion parameters. As AI evolves to “continuous learning” (per the “Age of Experience” paper), Bittensor’s adaptive incentives position it ideally for AGI or superintelligence.
Future Vision: Incentivism and Global Impact
Bittensor redefines economics as “incentivism,” per James—pure incentives without regulatory friction. JJ suggests subnets for renewable energy, verifying solar farms or modular nuclear reactors via IoT, recreating carbon credits permissionlessly.
With predictions of TAO reaching $581-$996 by late 2025, and recent developments like Europe’s TAO ETP, Bittensor’s growth mirrors Bitcoin’s hockey-stick trajectory. JJ concludes, “Bittensor is really changing the ground rules… for how AI is incentivized… governed… and improves.”
As subnets integrate—compute, data, training—Bittensor could power an “age of experience,” where AI continuously evolves. For entrepreneurs, it’s a low-cost gateway to innovation; for humanity, a bulwark against centralized monopolies.
The episode emphasizes: Bittensor isn’t just building AI—it’s upgrading how intelligence emerges globally.
Watch the Full TAO Pod EP01 here:
The TAO Pod channel link.
Be the first to comment