TAO Roundtable: LeadPoet’s Gavin Zaentz Shares The Recipe For Winning On Bittensor

TAO Roundtable’s third episode brought LeadPoet (SN71)’s co-founder, Gavin Zaentz, on to answer a cruial question: what actually goes into a Bittensor subnet worth building?

The conversation walked through LeadPoet’s own thesis (sales infrastructure over the crowded coding agent space) and used it as the working example for a broader playbook on subnet design, validation, and long-term positioning.

Going through the conversation makes one realize that the subnets worth acquiring a slot for share three properties: they solve a valuable problem, their validation mechanism holds up against adversarial miners, and their team can communicate a coherent vision the market takes seriously.

What Makes a Subnet Worth Building?

The conversation moved between LeadPoet’s specific model and the general principles behind subnet design, with the strongest points falling across product thesis, validation architecture, and the market discipline the new emission rules now enforce.

1. VALIDATION IS THE CRUX OF THE WHOLE THING: Before acquiring a slot, set up the subnet on testnet and run adversarial miners against your validation mechanism using Claude or Codex to try every way of gaming the system.

If miners can extract rewards without producing value, the subnet does not work in practice, and nothing else in the design matters.

2. THE SUBNET HAS TO SOLVE A VALUABLE PROBLEM: Problems within R&D (Research and Development) scope like drug discovery, or self-contained systems accruing value like compute or inference marketplaces. Emission runway is not the point.

If the subnet does not produce something worth paying for outside the network, no incentive design saves it.

3. THE TEAM HAS TO COMMUNICATE A COHERENT VISION: Marketing, podcasts, comms, and clear articulation of what the subnet is doing. The market is the source of truth for which subnets survive or die under the new emission rules, and a subnet without a legible story ends up on the deregistration list even if the underlying tech is real.

4. MINERS ARE ECONOMIC ACTORS, NOT COLLABORATORS: They will feed low-quality outputs if the validation mechanism has any leakage, because their goal is getting paid, not producing value.

Build the system so they can only get paid by actually adding value, and assume adversarial behavior from day one.

How Bittensor Rebuilt Its Core Economics Across Three Releases

5. THE NEW EMISSIONS STRUCTURE FIXES A MISALIGNED INCENTIVE: Previously, allocating all emissions created miner sell pressure that pushed subnet value down, so subnets were incentivized to burn instead of build.

The new structure with chain buybacks and burn-penalized emissions means subnets can now allocate their full emission to real work, which aligns the incentive with the design.

6. CATEGORY SELECTION MATTERS AS MUCH AS EXECUTION: For LeadPoet, sales was chosen as a niche because coding agents are dominated by Claude, Cursor, and OpenAI, which makes infrastructure differentiation impossible. Picking an underserved category with real pain points beats fighting for a category that already has three well-funded incumbents.

7. MEASURING SUBNET SUCCESS IS CASE-BY-CASE, NOT A SINGLE METRIC: For agent or model subnets, benchmark improvement matters, but only if the benchmark is robust and the model is not overfit to it. For commodity subnets, look at demand for the commodity and the margin between miner payout and revenue. Miner count and price are misleading in isolation.

8. THE INFRASTRUCTURE PLAY IS WHAT COMPOUNDS: LeadPoet’s agent product is the meat and potatoes, but the moonshot is a sales-focused LLM sold to Apollo, ZoomInfo, Salesforce, HubSpot, and Clay.

Building infrastructure that other companies plug into is what compounds beyond a single product, and Bittensor’s miner-directed training data is what makes that infrastructure possible at all.

Const Explains Root Reborn: Everything You Need To Know

9. ROOT REBORN IS REASONABLE BUT OVERCOMPLICATED: Routing power through subnet exposure is conceptually sound, but validator voting adds a layer dynamic $TAO (dTAO) could already handle through market allocation.

Letting the market allocate through dTAO rather than through validator delegation would achieve the same outcome with less structural risk.

10. MAKING MONEY ON BITTENSOR IS BROADER THAN EMISSIONS: Mining and validating are the obvious paths. The overlooked one is white-labeling subnet outputs (Chutes, Lium, Targon) and reselling them as a service under your own brand. A one-click miner path is also coming that lets anyone participate with just a Chutes or OpenRouter API key.

The Filters the New Rules Enforce

The strongest takeaway is that the perfect Bittensor subnet is not defined by any single technical trait but by whether it passes the filter the new emission rules now enforce. A subnet that solves a real problem, validates its miners honestly, and communicates a coherent vision will accumulate market confidence and chain emission at the same time, while a subnet that misses any of those three fails at the incentive layer before the technology ever gets tested.

LeadPoet’s own model happens to hit all three, which is why the subnet already has enterprise customers, a second product version shipping this month, and a defensible long-term moonshot in the sales LLM. For anyone acquiring a slot in the current environment, the advice is the same one the market is already delivering through the deregistration list: build something valuable, validate it honestly, and tell people what you are doing clearly enough that they can believe it.