# Into: Coldint / bootstrap Coldint registered as a subnet for open, collaborative AI model training. Its team now says that idea is behind them, and the SN29 slot is being repurposed as a place for other people to test their subnet ideas. // From training subnet to subnet incubator --- > New to Bittensor? Start here. Experienced users can skip to the full analysis. ### What is this? Coldint (Subnet 29) launched in 2024 as a Bittensor subnet for collaborative, distributed AI model training, where miners competed to improve a shared model and could also earn for publishing useful code and fixes. In a June 2026 post in its Bittensor Discord, where the channel has been renamed "bootstrap", the team said the original Coldint idea is behind them and described a new direction for the slot. The new plan, as the team describes it, is to run SN29 as a host for other people's early-stage subnet ideas. An aspiring subnet owner would run their idea as a "subsubnet" on the SN29 slot, and a fraction of the slot's mining reward would be shared among the miners working that idea. Everything in this section about the new direction comes from that single Discord announcement; it is not yet reflected in the subnet's on-chain identity or its public code. **The simple version:** Think of it like a sandbox or incubator for Bittensor subnets. Instead of paying to register your own slot to test an idea, you run it inside someone else's slot first, with real miners attached. **Centralized equivalent:** The closest analogy is a startup accelerator or a shared testing sandbox. There is no direct decentralized equivalent that we are aware of. **How it works (as described by the team, and not yet built):** - **Aspiring subsubnet owners** bring a subnet idea and validator code, and run it on the SN29 slot rather than registering a new slot of their own. - **Miners** chase the slot's mining rewards, which the team says will continue to be paid in alpha, giving the idea real participants to harden it against. --- ### Why This Matters - **The problem it solves:** Launching a Bittensor subnet is expensive and unforgiving. A new slot costs a TAO burn to register and starts a four-month immunity clock, so early bugs eat into limited runway. The team's pitch is that a builder can shake out those bugs on SN29 first, before committing capital to a real slot. - **The opportunity:** If it works as described, an incubator slot lowers the cost of trying a subnet idea: real miners, no registration burn, and feedback from an experienced team, all before raising money for a slot of one's own. - **The Bittensor advantage:** A shared slot can attach live economic incentives to an untested idea in a way a private testnet cannot, since testnet miners have nothing real at stake. - **Traction signals:** Thin, and worth stating plainly. The plan is described in a single Discord post and is not yet corroborated by code, on-chain identity, or other public sources. On-chain, SN29 currently shows a 0% share of network emissions and one active miner (TaoSwap, snapshot at publish). --- ## Full Analysis **Category:** Other (subnet incubator) | **Centralized Competitor:** Startup accelerators, private testnets Coldint's on-chain identity still reads "Coldint" and still points at the `coldint/coldint_validator` repository, the codebase for the original collaborative-training subnet. The shift described here is a stated change of direction from the team, not a change the chain or the repositories have caught up with yet. We are treating the Discord announcement as the team's own statement of intent and nothing more, because it is the only source for it. **What the team announced:** The team frames the new model around stake locking. In their words, an aspiring subsubnet owner's SN29 stake would be "safe from large unstakes" because the team intends to lock most, but not all, of the alpha, so that "only root selling and unlocked stake can push the price down." They describe this as depending on an upcoming native stake-locking feature, and the language throughout is future tense: this is a plan, not a running system. That framing matches what is verifiable on-chain today, which is that the plan has not been executed yet. As of the snapshot, the owner has locked roughly 3,166 alpha through conviction staking, which is about 0.06% of alpha supply. "Most alpha" is not locked. Worth noting on the mechanics: under Bittensor's conviction-staking design, locking stake is a commitment signal and does not change the emissions or rewards a position earns. The team's stated benefit is about price stability and protection from sudden exits, not higher yield. **The emission picture:** Under Bittensor's flow-based emission model (Taoflow, live since November 2025), a subnet's share of network emissions tracks its net staking flows smoothed over time, and a subnet with non-positive net flows receives a 0% share. SN29 currently sits at a 0% emission share (TaoSwap and TAO.app, snapshot at publish), and TAO.app shows net staking outflow over the trailing week. The team acknowledges this directly, saying SN29 emission is "now zero for a while," while claiming miners will still earn alpha mining rewards. We are reporting the team's claim as a claim; the on-chain figure we can confirm is the 0% emission share. **Market state:** At snapshot, the alpha token traded near 0.00332 TAO with a market cap around 17,500 TAO and roughly 9,087 TAO of depth in the liquidity pool (TaoSwap). Price is down about 33% over the trailing 30 days. The pool is thin, so even modest positions can move it. On public code: the validator repository `coldint/coldint_validator` shows its most recent commit on 2025-07-11, and the related `coldint/sn29` configuration repository on 2026-01-28, both confirmed via a live GitHub API check on 2026-06-13. No public repository describing the "subsubnet" or "bootstrap" model exists yet under the team's GitHub organization. None of this is presented as a judgment on the team's private work; it is simply what is public. --- ### Risk Factors - **Deregistration:** SN29 is well past its four-month network immunity and currently earns a 0% emission share. When a new subnet registers into a full network, the protocol automatically deregisters the non-immune subnet with the lowest EMA price. A subnet at 0% emissions with a low, falling price sits in the exposed group, which matters more than usual here because the incubator pitch depends on the slot continuing to exist. - **Unproven model:** The subsubnet incubator is described in a single announcement and is not yet built, documented in code, or reflected on-chain. There is no public way to verify how rewards would be split among "subminers" or how a subsubnet would run inside the slot. - **Dependence on a feature that is not live:** The team's price-protection argument rests on locking most of the alpha through a stake-locking capability it describes as upcoming. Today only a small fraction of supply is locked, so the central promise is unverified. - **Liquidity:** The pool is thin relative to market cap. Entries and exits can move the price meaningfully, and the team's own pitch concedes that unlocked stake and root selling can still push it down. --- We'll revisit SN29 with a fuller writeup once the incubator model shows up in code, on-chain, or in more than one place.