Tao Subnet Overview, 3 Minutes to Understand Key Subnets and Investment Logic 👇👇👇

Let me first share what I believe is the most critical investment logic for subnets:
1⃣ Invest in directions most likely to link with major AI giants in reality
2⃣ Invest in subnets backed by TAO core participants
Subnet Overview:
There are 128 TAO subnets in total. The overall situation varies, but some sectors have already started to stand out.
1⃣ The Covenant-led LLM large model sector
Covenant has three subnets:
Responsible for pre-training: @tplr_ai SN3
Responsible for post-training: @grail_ai SN81
Responsible for computing power: basilica SN39
This basically covers all aspects of current AI large model training. These are also the three most exposed and most interesting subnets moving forward.
2⃣ The Chutes-led computing power sector:
Currently, the larger ecosystems of computing power subnets include:
Targon SN4, Chutes SN64, basilica SN39
Decentralized computing power was one of the hottest tracks before.
Chutes is currently the largest computing power platform among subnets and has real revenue.
Targon recently released a paper jointly with Intel, causing its price to surge.
This is why I emphasize the importance of finding targets that can link with AI giants—because that’s how you capture attention.
3⃣ Stillcore Capital full package:
Stillcore Capital is the biggest supporter in this round of TAO, founded by @Jason, a leading TAO investor.
This fund has explicitly invested in two subnets:
Ridges AI: Building a Cursor-like AI coding platform
Score SN44: AI visual models, sports event dynamic prediction models
These are backed by institutions and have funding support. Since research thresholds and due diligence for subnets are quite high, institutional backing generally indicates “relatively” high quality and some ability to attract attention.
4⃣ Yuma full package
I won’t go into detail here. Yuma is one of TAO’s largest investors and a builder who has focused on TAO for years. Yuma has supported many subnets, but it feels somewhat like spreading too thin, so I won’t list each one individually. Many of these targets overlap with previous ones.
Overall, I believe the research threshold for TAO subnets is quite high. Non-professionals find it difficult to distinguish good from bad purely based on technology or fundamentals. In the end, it still comes down to the logic of capturing attention.
Whoever can capture the most attention, or even bring more attention to TAO, is the good subnet. That’s also why I invested in Temlar initially and later in Grail.