Note: this report assumes you have a basic understanding of Bittensor. If you need a refresher, check out my previous report.

The best investments are those that you can “sell” to all market participants across the bell curve.

Taglines matter. It makes no difference if a new piece of world-changing tech comes along. If you can’t distill its purpose into a catchy tagline, it will likely never achieve escape velocity. (Most people have the attention span of a coked-up Rhesus monkey.)

Put succinctly, Bittensor is an arena that’s purpose-built to accelerate decentralized AI, leveraging crypto incentives.

And what’s unique about it is that you could sell it to all curvoors - you just have to frame it for your audience.

Left curve: The Bitcoin for AI

In its purest form, the bull case is quite simple.

• AI has been, and (in my opinion) will continue to be, the most exciting narrative of the cycle
• $TAO is (in my opinion) the best Crypto x AI project and has been dominating the mindshare charts

Most people just want exposure to “the narrative” and will full-port the leader. This works to our advantage.

Also, each cycle, the market gravitates (and rewards) exciting ideas that are difficult to value. Whether it’s colored coins in 2012, ICOs in 2017, DeFi governance tokens in 2019, or NFTs in 2022 - people don’t know what a fair price would even look like, so they just bid into infinity.

I believe that “Crypto x AI” will occupy this position over the next couple of years. There have been brief previews of this hype, but I foresee this as being the center of a future mania phase. AI is the global zeitgeist, and TAO is crypto’s chosen fighter.

The best part, the most bullish catalysts for eyes to return to AI are still ahead of us.

1. The release of GPT-5 (and comparative frontier models from Anthropic and Meta) are expected to change the game when it comes to LLM capabilities. Whether they actually do is inconsequential, it’s the narrative that matters.
2. The full release (and subsequent backlash) of text-to-video models like SORA to the public.
3. The inevitable IPOs for OpenAI and Anthropic, which already command ~$157b and ~$20b valuations in private markets. You want TAO comps? Just wait for OpenAI to hit $1tn on the NYSE.

And when the market learns that Bittensor is “Bitcoin, but for AI,” the real party begins.

Mid Curve: Fuck Centralized AI

With more attention flowing to these closed-source AI superpowers, the question will undoubtedly shift to “who should control this?”.

Do you trust a company to “not be evil” when they can manipulate the system for profit? Or how about our incorruptible politicians? Surely, they would only use the tech for the betterment of their citizens.

The clear solution is an open-source alternative to these centralized models. An option to opt-out of the control while still benefiting from this technology, which is already shaping our world.

The problem with traditional open-source? It’s messy, it’s uncoordinated, and its goals are loosely defined. Thankfully, Bitcoin has shown us that crypto primitives are the great equalizer.

Because Bitcoin miners earn rewards for mining (securing the network), that’s what they do. And fuck me, they do it well.

The Bitcoin miner network is the largest supercomputer in the world. Larger than any company’s, larger than any government’s. Purely from well-defined incentives. That’s the power of crypto primitives.

If open-source AI is going to win, it has to learn from this.

Like Bitcoin, Bittensor miners (who are AI/ML engineers, in general) earn rewards for producing valuable work, except this time, the valuable output comes from building machine learning models. To recycle the parlance from my primer:

*Hits blunt*

“Dude. What if we got Bitcoin miners to build AIs instead of just solving useless math problems?”

Bittensor’s architecture allows entrepreneurial spirits to spin up their own competition (called a subnet), define their objective, set the rules, and invite participants to start competing for the prize of that sweet, sweet TAO.

Wanna make a super-fast LLM bot? Done (try it).

How about a machine that harnesses an army of God-like traders to play the markets? No problem (try it).

No, I wanna train foundation models. I mean, fine-tune them. Wait, no, let’s just make roleplaying LLMs so I don’t have to talk to real people anymore (try it).

‍

Bittensor takes the same crypto-economic principles that helped Bitcoin challenge central banks and applies them to an even more pressing threat: centralized AI. By harnessing the power of competition and incentives, Bittensor aims to create a decentralized AI ecosystem that's more resilient and adaptable.

As the Bittensor network expands, its intelligence grows exponentially. With each new block, the network becomes more robust and sophisticated. But what's truly significant is the way the competition among top miners drives the network's progress.

The fierce competition on Bittensor's subnets leads to a continuous cycle of improvement. Whenever top miners engage in a duel, the network as a whole benefits, leveling up with each new challenge.

Right Curve: Supercommodities

What really sets Bittensor apart from the countless others in its niche is that the more you dig in, the more bullish you become.

I’m not sure how to articulate just how frustrating it was to go through the linked list above. Projects would promise the world - throw around some impressive-sounding terms and wave their academic/institutional pedigree in your face, only to shrug their shoulders when you ask to see metrics, deliverables, or use cases.

Imagine Bittensor subnets as modern-day commodity-producing businesses, similar to oil refineries, corn plantations, and cattle farms. Bear with me for a moment.

While these businesses produce traditional commodities like oil and cattle, Bittensor subnets specialize in creating high-value, supercommodities that are both tamper-proof and constantly improving (e.g., models producing price predictions improve over time - it’s like if newer beef tasted better and better year after year).

Examples of these supercommodities could be:

• LLM Inference
• Computational Power
• Data Validation and Cleaning Algorithms
• Cryptographic Proof Generation
• Decentralized Data Storage
• Image and Video Processing Services
• Fraud Detection Models
• Decentralized Identity Verification

However, only the top-performing models in each subnet can deliver their commodities to end-users. By doing so, they enable the sharing of intelligence, solving of complex problems, and creation of innovative applications – all without relying on traditional gatekeepers.

We are in the early days here, to be sure, but look at what these subnets have been able to produce over the past few months. Truly impressive.

Just as nation states have built their economies around commodities like oil and rare metals, AI supercommodities have the potential to disrupt this paradigm.

By creating a new class of high-value commodities, Bittensor is opening the door to a decentralized, community-owned fountain of knowledge that can be tapped into whenever needed.

As AI supercommodities become increasingly valuable, they will enable a new era of open-source resistance against centralized powers.

The idea of a community-owned, decentralized intelligence network is no longer a distant concept, but a reality being shaped in the Bittensor discord today.

Competitive analysis

Now, the mere existence of these “TAO-killers” should tell you off the bat that Bittensor is onto something. The market understands that if there is anything that crypto is good for, it is coordination.

The idea of a “collective intelligence network” or “intelligence coordination network” is such a no-brainer once it clicks, and the late-blooming VCs would rather fund a dozen competitors than buy TAO at market price.

But let’s not get too set in our ways - let’s give some of the more interesting competitors a chance. Who knows, maybe they have some cool ideas.

BasedAI

Looking at the basic architecture of BasedAI, you may be forgiven in thinking that it’s a frog-themed fork of Bittensor. But there are some notable differences.

Instead of subnets, BasedAI is horizontally scaled with “Brains,” each hosting their own LLM in a zero-knowledge format. These “zk-LLMs” can process and respond to encrypted inputs from users without ever decrypting their original messages. Truly private LLMs.

BasedAI's innovation lies in their "Cerberus Squeezing" technique, which simplifies the process of transforming LLMs into zk-LLMs, drastically reducing computational complexity. This enables miners to run their LLM models and validators to verify their work more efficiently.

Having these private LLMs on a public blockchain is a huge feather in the cap of decentralization, and could be a game changer for industries such as healthcare and law, which require an assumption of privacy.

All this being said, I wouldn’t classify BasedAI as a true “competitor” to Bittensor. On the contrary, they even list Bittensor subnets as potential users of Brains in order to maintain confidentiality in a distributed AI network. Synergy, lads, not competition.

commune.ai

Commune was founded by an ex-Opentensor Foundation developer, with the goal of tweaking the mechanisms of Bittensor to iron out some inefficiencies. Sounds great off the rip.

After digging through the details, there are a lot of echos from Bittensor’s design:

• Built on Substrate
• Owners (they call them “Founders” though), Miners, and Validators
• Yuma Consensus
• Most discussion/building happening in their discord

All in all, the ecosystem kinda feels like just a cucked version of Bittensor’s own. There’s some development and competition occurring, but most subnets echo objectives that are already established on Bittensor.

A silver lining from Commune can be found in the screenshot above. See how SN0 is getting most of the stake?

Unlike in Bittensor, Commune’s Subnet 0 is known as the General Subnet - where things get a bit loosey goosey.

Commune's General Subnet has no founder, no Yuma Consensus, and no direct incentive mechanism. Instead, users can allocate incentives to whatever they value, whether it's code contributions, helpful widgets, or community members.

This allows the protocol to incentivize hard-to-represent things, but it's still early days, and I'm not aware of any major breakthroughs yet.

The dynamic between Bittensor and Commune is likely to be one of shark and minnow: talented ML developers will prefer the more established and lucrative network.

Sentient

After Peter Thiel walked up to the desk and slapped $85 million dollars(read: "his nutsack") on the table at the seed round, Sentient became the talk of the town at EthCC. It’s not every day that you see a crypto project raise that amount of money for a seed round (Story Protocol’s recent, and somewhat controversial, Series B was $80 million), but it does show a ton of interest in the “collective intelligence network” space.

Though not much is public about their intentions, the backbone of the project seems to be their OML (Open, Monetizable and Loyal) format, which they hope will directly align incentives of model developers and their end users.

Open - models are open-source and can be run by anyone.

Monetizable - model owners are compensated based on their contributions and how often their models are used.

Loyal - models behave consistently, so they can be relied upon for ethical use.

Beyond this initial blog post, not much is known about the scope of the project.

Personally, I’m all for new innovations that take decentralized AI forward. If the OML standard is useful, I hope it helps open the eyes of open-source contributors to the possibilities of developing with blockchain rails.

As far as a Bittensor competitor, I think it’s just too early to tell. Ironically, if the OML is shown to be a good way to evaluate models, I could see a version of it finding its way into a subnet incentive mechanism. That’s open source for you.

Rounding out the list

It’d be a practice of self-flagellation to detail out why every Bittensor competitor doesn’t match up to the king. But in order to be thorough and avoid the “you forgot about $MYSHITTYBAGS,” a couple notes:

Zero1 is not a serious competitor and their reasoning of “it is PoS so everyone has the chance to participate” holds less water than your uncle’s weak bladder after his fifth Bud Light. Mining should be a competition to produce the best models. It doesn’t make sense to democratize a competitive process.

Allora, which originally raised as UpShot in 2021 and has pivoted from NFT appraising,  is creating “self-improving decentralized intelligence,” and their “topics” has some overlap with a couple of the Bittensor subnets.

They have Workers and Reputers instead of Miners and Validators, and Topics instead of Subnets. One interesting aspect is that their emissions are governed by the “utility” of the topic, so models that are used more will receive more emissions. This does give some immediate feedback to the topic market, but sacrifices the longer-term objectives of solving difficult problems.

It would be foolish to suggest that Bittensor, in its current form, is a perfect machine and that these competitors contribute nothing to the conversation. These other coordination networks bring up some novel mechanisms, namely:

• BasedAI’s method of zk-ML circuitry
• commune.ai’s General Subnet sandbox
• Sentient’s OML standard for model creator incentivization
• And Allora’s reward mechanism for topic/subnet weighting

But do any of these individual tweaks move the needle? After all, the incentivization mechanism for most of these projects is the native token, and if that isn’t worth much, why not just port these improvements directly to Bittensor?

After all, “good artists borrow. Great artists steal.”:

There are over one hundred documented Bitcoin forks. People thought the idea of “hard money backed by code and compute” was a great idea, but the Corn could use some upgrades. How many of those survived?

Maybe Zero1 is just the MimbleWimbleCoin of its time.

Current problems

Now, after blowing the smoke up Bittensor’s ass for way too long, let’s take it down a peg. The current architecture works, but there is room for drastic improvement.

In my mind, the three immediate problems are:

• Weight copying
• Inter-subnet competitiveness
• Token holders have little sway

Weight copying

In the 3024 Galactic Olympics, a Martian diver receives scores of 6, 6.5, 5.5, 6, and 10. The judges' scores are suspicious, particularly the one from Mars.(That spiky, orange bastard)

How do you score the scorers?

Bittensor's Yuma Consensus would flag this judge for giving preferential treatment. And if the pattern continued, this judge would be kicked out.

A problem with this design is that, in the world of Bittensor, judging doesn’t come free. Bittensor validators usually have to run hardware to benchmark and test miner models, and will offset this cost with the TAO they earn over time.

A strategy that some employ is to copy the answers of other validators and submit them as their own. This sidesteps the validation computation and, since validators are compared to the average response of other validators, actually comes out to be more profitable than acting honestly.

“root” validators = the largest 64 validators, which vote on weights for subnet TAO emissions

In the above chart, Owl Ventures is fairly well-known for weight copying, and Datura claims to weight copy on subnets where, from their perspective, it doesn’t make business sense to validate honestly - whether that is due to startup or maintenance costs. (Note that, since the implementation of child keys, Datura no longer weight-copies - sick)

While some community members lambast these root validators for their “heresy,” it is not as if copying weights breaks the “letter of the law” of Yuma Consensus in its current form. You could argue that it’s perfectly within their rights to maximize returns for their delegators and weight copy until their heart’s content, but it is clear that this is not optimal for the project long term.

Inter-subnet competition

Today, once a project has: launched, showed a minimum viable product, and received weight from the Opentensor Foundation (OTF), they are pretty much in the clear. Many other validators look to OTF for approval and will follow suit with sending emissions to subnets. And once a project has made it to this point, it can become easy to rest on their laurels.

Validators can check in on projects to ensure progress, but there's no direct penalty for not doing so. This can lead to a "ghost town" of subnet owners collecting TAO emissions with minimal effort. To fix this, there needs to be a personal stake (skin in the game) for those deciding subnet emissions.

Token holders are second-class citizens

As alluded to in my above gripes, in today’s architecture, the largest validators hold the bulk of the power in the ecosystem. While it is true that tokenholders are the ones that delegate their stake to these validators and therefore can “vote with their TAO,” this is an imperfect system.

In reality, most tokenholders only glance at validators' strategies before delegating, and don't follow up when they change.

TAO holders have no incentive to continually reassess their nomination. In a perfect world, TAO holders would be nudged towards supporting validators that are benefiting the ecosystem, but there’s no mechanism in place just yet.

In case you were asking my opinion, I’d say to delegate to a validator that allows TAO holders to assign weights themselves, such as Miner’s Union.

The proposed solutions

You probably noticed how I kept saying “just yet” or “in the current design” above - and it’s worth hammering home.

Bittensor is an evolving project. Its current form (Finney) has only been around since March of 2023 and has since grown from a single subnet to 52 individual competitive ecosystems.

Is it perfect? Of course not. Lessons are being learned while the chain continues to empower new accomplishments month after month.

A couple upcoming changes are set to (in my opinion) drastically tune up this well-oiled machine we call Bittensor - Dynamic TAO and Liquid Alpha.

Dynamic TAO

If you’ve been hanging around with TAO moonboys (takes one to know one), you’ve probably heard about Dynamic TAO (or “dTAO” for short). In essence, it takes the power away from root validators and puts it back into the hands of TAO holders like you and me.

Basically, this upgrade will create “subnet tokens” that TAO holders can swap with their TAO. If you think a new subnet is really cool and will change the world, you can swap some of your TAO into their dTAO token. When it does indeed change the world, you can swap a bit of your dTAO into a lot more TAO.

Subnets earn emissions based on the relative value of their subnet token against all of the others.

Dynamic TAO turns emission allocation from a pseudo-DAO governance into a completely market-driven machine.

Prove your worth, and get rewarded with glorious TAO emissions to fund your incentive mechanism.

Not delivering results? Token holders notice and the smart ones run for greener pastures. No more waiting around for validators to recognize when projects lose steam. No one is more aware than players with skin in the game.

There are, of course, a lot of technical details to buff out, but here is the original proposal (and TLDR) in case you want to check it out. At the time of writing, a series of testnets are taking place in order to prove out each mechanism involved in dTAO.

Liquid Alpha

This one may have flown under the radar for many casual observers, but I honestly think it’s just as big as Dynamic TAO. Liquid Alpha changes Validators from passive judges into talent scouts.

In the current system, as long as you agree that the best miners are the best and the worst miners are the worst, it doesn’t matter how long it took you to come to your conclusion.

With the upgrade, validators that “discover” outperforming miners before the pack are rewarded handsomely. This turns the act of validating into a competitive sport - with the laid-back validators earning fewer rewards than their over-achieving peers.

After Liquid Alpha and Dynamic TAO go live, the Bittensor machine will be whirring from all fronts:

• TAO holders will be incentivized to sniff out the best subnets
• Validators will be incentivized to uncover the best miners
• And, as always, miners will still be incentivized to crank out the best models

The bear case

As always, if you don’t study the bear case of your investments, you’re destined to be the one left holding the bag when your would-be invalidation criteria pass you by. With that in mind, what am I worried about?

1. 🏃 Compounding and complex changes being added too quickly. Dynamic TAO, on its own, is a radical change to the network and the full ripple effect of it will take time to play out. If Bittensor decides to compound this change with more updates in a short period of time, there would likely be consequences that are difficult to foresee.
2. ↘️ Decentralization as a lower priority. Because the Opentensor Foundation needs to move fast, full decentralization has taken a back seat. A potential issue that arises from this is that of the insider threat. If there are billions to be gained, what’s to stop a motivated individual from messing with the state of the chain.
3. 🧛 Vampire attacks from competitors. And with all of these competitors launching, I have to imagine they will be pelting money at the miners (ML engineers) in order to steal away some of their time and resources. While I don’t think this would kill Bittensor, I think it would slow down progress. Miners are what drive the subnets forward, after all.
4. 🐞 Misallocation of emissions to parasitic subnets. There is a worry that “the market” will over-allocate to flashy subnets that are too short-sighted in their mindset. These subnets may be looking to jumpstart their own product without contributing to the network, or perhaps they just are peddling vaporware. Either way, this would take away resources that would otherwise be driving progress for the network
5. 📉 Lack of demand for subnet intelligence. Of course, you must consider the case that there won’t be demand for subnet output. Whether it is due to better APIs from centralized competitors, difficulty involved with integrating with subnet endpoints, or something else entirely, a lack of demand would seriously hinder Bittensor adoption.
6. ⏰ It’s too late for open-source alternatives. My base case (illustrated below) is that the aligning nature of incentive mechanisms is able to coordinate a huge mass of talent and resources to open-source AI development. Combine that with the fact that centralized AI companies have to contend with regulatory pressures, and it isn’t hard to see the path where open-source wins out. But what if this is wrong? What if centralized AI achieves “AGI”/”the God model” (a load of horseshit, but I digress) before open-source catches up? In that case, we’re probably in a dystopian hellscape anyway, so it probably doesn’t matter that my net worth is tied up in magic internet coins.

All this being said, I think market commentators fall into the trap of “there is an issue in the current design, therefore the project is doomed.” As we’ve mentioned, Bittensor has overcome so many hurdles that would’ve destroyed lesser projects, and has made it out the other side stronger.

There will always be challenges when you are throwing a wrench into an ecosystem that is so heavily dominated by powerful players. But if it were easy, it wouldn’t be so damn exciting.

How hype returns

If you have been around for any stretch of bullish market movements, you’re aware of how fast narratives can shift. And if you aren’t dialed in enough, you can be the one buying the tops and providing sweet sweet exit liquidity to the trend setters. We’ve all been there.

My plan is to wait for the inevitable AI wave to return, and not be the top-buyer you see seething in the comments.

But how does that wave return? I can think of a few catalysts.

Subnet speculation

As Mr. Pango brilliantly explained in his recent piece, speculation, on its own, is a use case. Perhaps the use case when it comes to crypto.

Subnet tokens going live opens the playing field.

Think cancer research means fuck-all compared to graph optimization problem solving? Or do you think there will be a ton of demand for AI-driven market signals?

You can now bet on this.

Buying subnet tokens is like bidding on AI startups. Buying TAO is like buying an AI ETF, composed of AI startups.

There is no other venue where you can speculate on which AI use cases are more valuable than others. AI is the most hype-inducing narrative of our lifetime and now we are given a tool that lets us gamble on which AI products are most exciting... And you’re bearish?

Internal forces - crypto x AI hype

While a subnet making a breakthrough and smashing incumbent benchmarks would make some nerds happy and could even make the evening news, I don’t think this would cause an overnight flood of new capital.

Nevertheless, it would force AI researchers and ML engineers to take a deeper look into Bittensor, which would be a huge boost in fundamentals.

Subnets that fall into this category:

• Model development: SN4 (LLM response time), SN9 (Pre-training), SN21 (Multimodal), SN29 (Collaborative Training), SN35 (Logical models), SN37 (Fine-Tuning)
• ML advancements: SN15 (LLM benchmarking), SN33 (Data Structuring), SN38 (Distributed Training), SN39 (Edge computing), SN40 (Chunking)
• Verifiability: SN2 (Proof-of-Compute)
• Academic Problems: SN25 (Protein folding), SN31 (Neural architecture search), SN43 (Graph problems)

If any of these unseat current centralized leaders or offer new insights for ML researchers, this is long-term bullish for the Bittensor network. The AI community is small, tight-knit, and competitive. Show them a new tool they can use to gain an edge, and they will jump at the opportunity.

What would bring instant hype is a viral, user-facing application that leverages the power of crypto rails (or, more succinctly, crypto-empowered AI tech). Imagine a project on the scale of TikTok that uses open-source intelligence in the background. Several subnets are currently working on technology that is a single bright idea away from a killer app:

• DeepFake detection: BitMind (SN34)
• Role Playing LLMs: Dippy (SN11)
• Trading signals: Proprietary Trading Network (SN8)
• Generative AI: text (SN4, SN19, SN18), images (SN19, SN26), speech (SN3), 3D (SN17)
• Prediction markets: Sports betting (SN30, SN41, SN44), General prediction markets (SN6)
• AI-powered assistant/agent: Apex (SN1), BitAgent (SN20)

AI has already produced overnight unicorns in Web2. What makes you think Web3 is any different?

Bittensor empowers subnets to take a lot of shots on goal. All it takes is a bright-eyed entrepreneur to look at the raw intelligence being produced and saying “hey, what if you took this and used it to…” and Bob’s your uncle.

External forces - centralized AI hype, hate, or regulation

Looking beyond the realm of crypto for a moment, traditional AI companies make news headlines on the daily.

And unlike within our little corner of the world, there’s not much your average market participant can speculate on to align with their theses. Sure, they could chuck another thousand dollars into the money pit of NVDA, but how much smoke is left in that cigar?

The big players, like OpenAI and Anthropic, are all private companies. And unless you are already part of the VC class, you’re S.O.L.

But what can you speculate on? AI tokens.

So, what could reignite this desperation in traditional markets, where investors are willing to venture into the dark world of our magic internet money?

• New foundational models like GPT-5 or SORA unlocking new use cases for users
• Job concerns: reports that “AI is taking our jobs”
• Regulation concerns: laws like California’s SB 1047 and Europe’s AI Act, which hinder the development of AI companies within these areas  
• Centralization concerns: reports that tech companies are selling sensitive data to train AI

No matter what the spark turns out to be, one thing is certain - AI adoption is on a clear path: up. And whether centralized AI updates make people bullish or terrified, decentralized AI wins.

Comps and conclusions

Alright, lads, you’ve made it to the end. As a bit of a reward, let’s talk about everyone’s favorite topic - moon math. Just how high can the numbers on the screen go?

As it stands today, if we add up all 190 projects listed in the Artificial Intelligence section of CoinGecko, we see that it only amounts to $32B.

All of AI x crypto can be boiled down to just 21% of OpenAI’s valuation ($157B) or 1% of Nvidia ($3T). I don’t know about you, but that doesn’t seem right.

In my mind, I see TAO easily cracking the top 10 of all cryptocurrencies (eventually top 5), which would be a rough 10x from here if we use the values from the 2021 bull run (conservative, in my mind).

Bittensor itself sits at a measly 2.6% of OpenAI and 0.1% of NVDA at their market cap of around $4B - tiny in the grand scheme of things.

With OpenAI pushing GPT-4o, o1, SORA, and robotic models, all with different objectives and specialties, we already see the cracks forming in the argument that there will be a single “God model” of AI. All these specialties, all these objectives… I don’t know, kinda just reminds me of subnets.

Plan of attack

Though I’m quite happy with my current allocation of TAO, I could see myself buying more on pullbacks >30%. If I didn’t have any exposure, I probably would just DCA until I was happy with the percentage in my portfolio.

As far as strategies, I am looking forward to playing around with dTAO tokens once they go live (probably around the end of the year). I think there will be a ton of opportunities to burn yourself, so I will only do this with a small portion of my TAO off the rip.

Subnets generally either appeal strongly to AI-focused market participants or crypto-focused market participants, which is evident if you take a look at current validator weightings.

To visualize this, let’s look at how the Opentensor Foundation (OTF) values subnets by examining their top 10.

Take a look at the subnet objectives on the left - notice anything? These are some heavily-technical subnets. And aside from a few outliers (SN11 and somewhat 19 & 4), these subnets are far more focused on hard AI research as opposed to consumer products.

From OTF’s perspective (and I’m not saying they are wrong), that’s what Bittensor should be focusing on - deep AI research. But if I had to guess, I think the market may care a bit more about subnets where there is actually a product they can play around with.

Examples come to mind like Taoshi’s trading bots (SN8), BitMind’s DeepFake detectors (SN34), betting apps like 6/30/41/44, and SocialTensor’s Make-It-A-Quote social tool (SN23). More eyes = more value.

My plan will be to look for these discrepancies, especially with newer subnets that align somewhat to the AI-building ethos, but lean heavy into the virality potential of consumer applications.

The discord will be a great way to sniff out new projects, as teams will be building in the testnet channels before setting their sights on mainnet subnet registration. And, of course, being the host of a Bittensor-themed podcast could be a nice edge. Paying attention to trends and updates in the ecosystem will pay off big time.

Regardless, these are the early days. Bittensor hasn’t even undergone its first halving cycle, the blockchain itself is still on training wheels, and there is an abundance of confusion amongst market pundits about what the project even is.

I like Bittensor because it reminds me of what first drew me into crypto. Using cryptography to cleave power away from the fuckers who only have their own interests at heart. Bitcoin gave us peer-to-peer money that doesn’t bend to the Fed. Ethereum gave us rails to create financial markets that don’t care about your nationality nor credit history.

But, AI is powerful. It’s disruptive. And we’ve only really seen the early effects.

It will, without any shadow of a doubt, shift our lives and the lives of our future generations in ways we can only imagine. So, ask yourself, do you want to be beholden to that power, or do you want to own a piece of it?

‍

‍

DISCLAIMER: The author of this report, 563, is so deep into Bittensor that he made a TAO-branded podcast. He also holds a lot of TAO. He is a… community member, at this point (pray for him). Please be aware of this bias while reviewing this report.

‍

‍

For further reading, here are some of my favorite resources:

• Running Bittensor by Collab+Currency - the article that originally TAO-pilled me. Send this to folks that want to learn more about Bittensor, but aren’t (yet) meal deal members
• Sami Kassab’s blog - great articles on many different topics

Toolkit:

• Bittensor Discord invite - weekly community calls and a channel for each subnet
• Taostats - great for live data about subnets, validator weightings, registrations, etc.
• Taopill - amazing dashboard that organizes the subnets
• TaoMarketCap - data for subnets and validators
• TaoHub - more learning about the subnets
• SubnetAlpha - more learning about the subnets