The Bittensor-native money market. Borrow $TAO without selling your Alpha. Soon, powered by @EndureNet.
Bittensor
- Tweets 68
- Following 2
- Followers 977
- Likes 85
Forge is built on Bittensor's EVM, which means it bridges two different layers: the EVM side where the lending protocol lives, and the native Bittensor side where your $TAO and Alpha stake live.
To use Forge, you need an EVM wallet to interact with the protocol. Your $TAO and Alpha stay in your Bittensor wallet. Forge connects the two through a mirror address, a mapped native address derived from your EVM wallet, so your collateral can move between layers without being bridged.
Mainnet is coming. Get familiar with the setup on testnet first: docs.forge.endure.network/le…
2
1
15
853
When you supply Alpha as collateral on Forge, it won't stop working.
Your Alpha will keep earning its native yield while it is deposited in the protocol, and a new layer of liquidity gets unlocked on top of that.
On mainnet, 75% of those emissions will flow to borrowers, and 25% will go to the protocol as a performance fee.
3
1
16
461
$TAO liquidity without selling your Alpha.
That's what Forge is being built for, and teams across the Bittensor ecosystem already trust us with more than $5M in signed Alpha commitments before mainnet is live.
1
1
20
1,789
Endure is now live on Pharos, curating its first vault on Prosper.
A milestone for the network.
We’re officially live on Prosper
Endure is the first subnet on Bittensor curating a Vault on Pharos blockchain.
- Our strategy: Endure Quality and Resilience Equity Vault
- Our p{VAULT}: pCARP
You can now follow our onchain performance and watch the market around the strategy develop in real time.
→ app.pros-per.xyz/?view=detai…
9
638
A risk model that underestimates volatility or overestimates liquidity is not just wrong, it's dangerous in a lending market. Those are exactly the misses that lead to undercollateralized positions and bad debt.
The asymmetric scoring Endure will use penalizes the dangerous direction more. For a protocol like Forge, that means risk intelligence biased toward caution, not just accuracy.
On Endure testnet, scoring runs on one function, applied to every prediction: how close were you, and which way did you miss.
Inside a grace band around the real outcome, a miner scores 1.0. Beyond a cutoff, 0.0. Between the two, the score falls linearly.
The cutoff is not symmetric. Missing in the dangerous direction costs more than missing safely.
For drawdown and volatility, telling a consumer the token is calmer than it turns out to be is the dangerous miss. For price and liquidity, claiming there was more of either than there really was is the dangerous miss. In every case, the safe direction gets three times more room before it is penalized.
A risk model that treated both directions the same would not be doing its job. The asymmetry is the point.
$TAO SN30.
8
366
A major milestone for Endure!
With @EndureNet now live on testnet, the infrastructure that will inform Forge risk parameters at mainnet is taking shape.
Endure testnet is live
Miners can now submit structured Alpha collateral risk recommendations, and validators can score them against realized market behavior.
The incentive and validator mechanisms are running on netuid 504 under the risk.v1.subnet_alpha schema.
Full documentation for validators and miners is in the docs.
Read them: docs.endure.network/build/mi…
Github: github.com/endure-network/en…
$TAO SN30.
8
765
Forge will solve two different problems:
For $TAO holders, you will be able to supply $TAO and earn yield while the market puts your capital to work, with rates that move based on market activity.
For Alpha holders, liquidity will no longer mean selling. You will be able to supply Alpha as collateral, borrow $TAO against it, and keep your subnet exposure throughout.
Built for both sides of the Bittensor economy.
3
2
14
702
Forge is Endure's first design partner.
Every borrow, liquidation, and repayment that moves through Forge at mainnet will be a real outcome for Endure's network to score against. That's not a side effect of building a lending market on Bittensor. It's the point.
Risk intelligence that has never been tested against a live market is theory. Forge is where it becomes practice.
A design partner is not just an early user.
Once Endure is live, they will consume its risk intelligence in a real context, against real decisions. In doing so, they will surface what useful risk intelligence looks like in practice: which outputs matter, which schemas need refinement, and where the signal is strong enough to act on.
That feedback will feed directly into how the network develops: sharper schemas, wider coverage, and more accurate intelligence over time.
For the partner, it will mean access to a risk layer that adapts to their needs before it is widely available. For the network, it will mean real outcomes to score against and real conditions to refine the intelligence under.
Both sides get something the other cannot build alone.
$TAO SN30
1
10
534
Most subnets have a misalignment problem.
Customers want efficient pricing, token holders want buybacks, and subnets need to be sustainable. These objectives most of the time point in different directions.
Forge has been designed around one loop that aligns all three: borrowers access efficient $TAO liquidity, suppliers earn higher yield than root, and more than 50% of protocol revenue is planned for SN30 Alpha buybacks.
The interesting part?
Higher SN30 value means stronger miner incentives, which translates into more efficient Forge risk parameters. More efficient risk parameters mean more revenue, and more revenue means more buybacks.
2
3
24
4,829
This is exactly why Forge is designed to start conservatively.
Supply caps and LTV limits will not be arbitrary. They are designed to ensure that a sudden price spike will not translate into outsized borrowing power. Forge will start every Alpha market with tight caps and a 25% LTV, limiting how much any single market can expose the protocol.
Over time, Endure's miners will analyze liquidity, volatility, and manipulation cost to recommend whether those parameters should be tightened or expanded based on real market behavior, not assumptions.
Conservative by default. Adaptive by design.
DeFi needs serious risk analysis frameworks
Apparently, pumping a token 100x in 20 minutes makes it excellent collateral. 🤣
According to preliminary analysis, an attacker manipulated TONIC’s price, deposited it into Tectonic at the inflated valuation, and borrowed approximately $75M in real assets.
nitter.net/jeremybtc/status/20941…
Cronos had to halt the entire chain to contain the attack.
Just another day in DeFi: nine figures protected by a terrible or nonexistent risk framework. 🤡
Price is not liquidity. Market cap is not recoverable collateral value.
A 100x move in 20 minutes should trigger an alarm, not unlock $75M in borrowing power.
This is what we are building @EndureNet and @ForgeLending to address.
Forge starts every Alpha market conservatively:
Maximum Alpha collateral
= 10–12% of the Alpha available in its liquidity pool
Maximum borrowing
= Alpha collateral × Alpha/TAO price × 25% LTV
If the pool holds 1M Alpha, Forge initially accepts only 100k–120k Alpha as collateral across all users.
This is only the starting configuration.
Endure miners must then improve its capital efficiency by analyzing liquidity, volatility, concentration, oracle divergence, manipulation cost, and liquidation performance.
If the market proves resilient, miners can recommend safely increasing caps. If risk rises, they should detect it early and recommend reducing exposure or disabling new borrowing.
Endure miners detect and price the risk.
Forge enforces the limits.
The goal is to make the attack unprofitable before it starts.
bittensor:native SN30
1
9
591
Collateral is what makes borrowing on Forge possible, and understanding how it works before mainnet puts you in a better position as a borrower.
When you supply Alpha as collateral, Forge holds it through a smart contract that manages Alpha stake via the Bittensor staking precompile. The contract is non-custodial: Forge cannot access or withdraw your funds.
That is what makes Alpha-backed borrowing possible without treating Alpha like a standard ERC20 token, because it isn't one. It is stake tied to a subnet, a hotkey, and a coldkey.
Your borrowing power depends on the collateral factor for that Alpha market. For example, a 25% collateral factor means 100 $TAO worth of Alpha supports up to 25 $TAO of borrowing. The remaining 75% is the protocol's buffer against price moves.
One thing specific to Forge: small Alpha positions and partial exits can behave differently from ordinary token withdrawals. Forge accounts from live Bittensor stake, not only from requested transaction amounts.
7
437
Security audits exist because code that looks right can still be wrong in ways that only become visible under pressure.
Uniswap v2 went through an independent security audit with ConsenSys Diligence before launching on Ethereum mainnet in May 2020. ConsenSys Diligence is one of the most established smart contract security firms in the industry, with audits across protocols like Aave. Findings were addressed before deployment, and the protocol went on to handle billions in trading volume without a smart contract exploit on its core contracts.
An audit isn't a guarantee, but it significantly reduces the attack surface and demonstrates that the team took security seriously before asking users to trust the protocol with real capital.
And the choice of auditor matters as much as the audit itself. Forge is working with @bailsecurity, which has become one of the most respected security firms in DeFi, with a track record that includes Euler, 1inch, Symbiotic, Rocketpool, and many more.
Bailsec is reviewing Forge's smart contracts before a single dollar moves through the protocol.
4
2
18
861
From the Forge side, here's how those two layers translate in practice.
Forge manages the lending flows, configuring the risk parameters with the Guardian Multisig at launch.
Once Endure is live, its intelligence will gradually inform those parameters over time.
The boundary is clear by design: miners will analyze risk, Forge will manage funds. Neither will cross into the other's role.
Endure has two layers worth understanding.
The first is the network itself. Miners will analyze risk by submitting structured assessments for specific targets; validators will score them against realized market outcomes, and the network will aggregate the results into a signed, auditable parameter feed.
The second is @ForgeLending: A native money market for Bittensor and Endure's first customer. Forge will launch with risk parameters configured by the Guardian Multisig. Once Endure is live, its risk intelligence will be gradually integrated to inform collateral parameters, supply caps, and market controls over time.
Forge will not be just a product. It will be the proving ground where Endure's risk intelligence will meet real market behavior.
$TAO SN30
2
11
897