Bittensor After dTAO: How Alpha Tokens Are Reshaping the Network
AI blockchain Bittensor’s long-awaited dTAO upgrade has now been live for a few weeks, shaking up subnet economics, reshaping incentives, and sparking plenty of volatility along the way.
With the dust finally settling, here’s a clear look at what’s working, what's not, and what to expect going forward.
Quick dTAO Refresher
Before dTAO, centralized validators controlled subnet rewards, leading to favoritism and slow growth for newer subnets. dTAO replaced this with a market-driven model:
- Each subnet issues its own Alpha token, paired with $TAO in liquidity pools. The token’s price determines the subnet’s $TAO emissions.
- Miners earn Alpha tokens, choosing between cashing out or betting on long-term gains.
- Stakers pick Alpha tokens instead of guaranteed $TAO rewards, adding risk.
- Validators must hold Alpha tokens, linking rewards directly to subnet success.
- Subnet owners are rewarded in their own tokens, requiring ongoing value creation.
dTAO So Far
dTAO’s first days were chaotic. Alpha tokens launched at 1:1 with $TAO but quickly diverged—some spiked to 5–10 $TAO on speculation, others fell to 0.1–0.2 $TAO.
Prices have since settled, largely because of automatic selling pressure (Alpha tokens being sold into $TAO to reward Root network stakers) and rapid profit-taking by early miners and subnet owners.
dTAO’s Timeline
To clarify where we’re at in the dTAO rollout, Chain of Thought’s Teng Yan created a timeline predicting how dynamics will evolve over coming months:
- Early Adoption Phase (Now): The Root subnet still dominates emissions but is losing share as more Alpha tokens circulate. Subnets are accumulating Alpha, validators are shifting from $TAO to Alpha, and subnet stakes now matter more.
- Day +100: Alpha and $TAO staking power should reach parity, marking a turning point where subnets start taking control of rewards.
- Long-Term: Network emissions will be fully market-driven, reducing Root rewards significantly in favor of subnets.
Top 5 Subnets by Alpha Token Price
So far, a few subnets have captured double-digit emissions.
Given the wide variety of subnets—each applying machine learning to specific AI tasks—the early rankings offer an interesting snapshot of which applications the network currently values most:
- Targon (SN-4): A decentralized testing ground for AI language models. Targon's decentralized verification also powers the user-facing platform, Targon.com, which provides fast and affordable leasing of advanced AI models.
- Chutes (SN-64): Built by Rayon Labs (behind Subnets 19, 56, 64), Chutes offers decentralized AI compute (which can be paid for in $TAO), allowing users to deploy, run, and scale AI models.
- Proprietary Trading Network (SN-8): A subnet using AI to provide and coordinate trading insights. On the network, miners provide sophisticated trading signals for crypto, forex, and indices, which can be purchased via Taoshi’s Request Network.
- Bitmind (SN-34): A subnet for AI-content validation and deepfake detection. It currently has a series of applications live like a web-based deepfake detection tool and a Chrome extension ("Agent Smith") that identifies AI-generated images online.
- Celium (SN-51): A decentralized GPU rental marketplace for machine learning and data analysis tasks. Within 60 days of launch, Celium claims to have generated over $30 million in revenue and maintains resources in multiple regions, including the US and Europe.
What Else Is Happening Since dTAO?
Beyond Alpha token excitement, the network has gained momentum with major players joining and projects gaining mindshare:
- OpenKaito Foundation’s Debut Sparks Excitement: Everyone’s favorite analytics platform Kaito now has the OpenKaito Foundation Twitter account for its SN-5. Kaito’s been involved in $TAO for a while, initially looking to build an AI-powered search engine for crypto. It’s unknown if that’s still the case but, with this new real estate secured, the rumor mill is alive with talks of a big application release soon.
- Templar Picks Up Momentum: Templar (SN-3), a decentralized AI-training subnet, also generated buzz by joining Twitter. It's hailed as among the only projects doing "true decentralized and tokenized training," so expect to hear a lot more from this subnet in the coming months.
- CZ’s Fund Invests in Tensorplex: Tensorplex, a decentralized AI company building tools for Bittensor, announced strategic funding from YZi Labs (CZ’s Binance investment arm). Their flagship product, Tensorplex Dojo, uses human feedback to improve AI models, aiming to become the "Human API" for AI-driven collaboration. This funding adds to their $3M seed round from DCG, Collab+Currency, and others.
- Bittensor-Powered Betting Agents: Recently, predictive sports-betting agents like Billy Bets and DKING have garnered popularity, with some seeing the social nature of sports betting as a ripe path for introducing the wider public to crypto, without them knowing its crypto. Both agents run on Bittensor, with BillyBets on SN-41 and DKING on SN-44. You can learn more about them in this week’s AI Rollup.
Zooming Out
If there’s one key takeaway here, it’s patience. There is incredible promise and potential in Bittensor and, while market-driven incentives create fairness, they aren’t foolproof.
Though emissions are more decentralized, bad actors could potentially exploit price dips to influence emissions by seizing a bunch of cheap tokens. Yet dTAO does represent a genuine, long-term opportunity for those willing to take it slow, monitor subnet traction, and accumulate tokens over a long timeframe.
Further, for those less comfortable navigating early subnet chaos, keeping TAO staked on Root remains a solid short-term strategy—APR will slowly decline but should stay respectable for a while longer.
In the meantime, here are some resources to get started with dTAO:
- Backprop Finance: dTAO trading & analytics terminal by Tensorplex
- SwordScan: dTAO mindshare & analytics terminal
- TAO Times: Top Bittensor ecosystem newsletter
- dTAO Token Guide: How to buy dTAO tokens