[PRE-ANN] TitanForge (TFN) | Shielded-only PoW | 1 block/s

Privacy coins usually make you give something up. Make shielding optional, and most coins stay on the public path, where they leak. Hide everything, and you have to trust that nobody has quietly printed coins. And the best-known privacy chains are slow, with blocks tens of seconds to minutes apart.

TitanForge is built to take on all three.

There is no transparent ledger at all. Every coin is held in a shielded note. Payments between users show no sender, receiver or amount on chain, and there is no public path beside them to leak through. Mining rewards and the development fund are public on purpose (see below).
Every node checks the supply in every block. Coins issued and burned and the balance of each pool are committed in every block header. If those totals ever failed to add up, a node would stop and raise an alert instead of carrying on.
One block every second. A GHOSTDAG blockDAG keeps blocks mined at the same moment instead of throwing them away.
No coins before launch. No premine, no presale, no ICO, no investor allocation, and the genesis block pays nothing. Development is funded by 5% of new coins for the first four years (see below).

This is privacy for ordinary payments: keeping personal wealth, business payments and donations out of public view. A viewing key lets you show your own payments to an accountant or auditor of your choice.

This is a pre-announcement. No TFN coin or token exists on any chain: no sale, no airdrop, no pre-launch wallet. The test network is not open yet, the code is not public yet, and there is nothing to mine or buy.

Beware of scams. Anyone offering you TFN is not us. We will never message you first or ask for your seed words, and our official links are the ones on titanforge.network.

We are posting now because we want miners, node operators and sceptics involved before the test network opens.

How it compares

Feature	TitanForge (not launched)	Zcash	Monero
Every payment shielded	Yes	No, transparent payments are allowed	Yes
Target block time	1 second	75 seconds, 25 after NU7 (due about Nov 5)	2 minutes
Mining	XelisHash-TF; CPU miner built, GPU miner planned	Equihash, ASICs	RandomX, CPU

What we built on, and what we added

We did not reinvent what already works. TitanForge builds on open research and open source:

From Zcash: Orchard and Halo 2 proofs, which need no trusted setup, using the Zcash project's libraries (MIT or Apache-2.0). They hide the sender, receiver and amount.
From Kaspa: the blockDAG node. Ours is a fork of rusty-kaspa v2.1.0 (ISC licence), and most of its code is that project's work. We removed its transparent transaction format and its UTXO set.
From XELIS: XelisHash v3 (MIT licence), the base of XelisHash-TF, a memory-hard variant with its own keys, so miners must be built specifically for it.
From Djed: the rules of Tungsten, our planned dollar-targeting asset. It follows Minimal Djed, a reserve design from the Djed paper, with stricter rules.
From Zephyr: lessons, not code. Zephyr ran a private Djed stablecoin, and its history is why Tungsten has no yield token, no emission subsidy to the reserve and no single oracle key.

We are not affiliated with any of these projects.

What we added:

Reward notes built by consensus. A miner's reward is created by the network itself as a note, not by the miner.
A turnstile on every pool. This is Zcash's ZIP 209 rule, applied to each pool: public balances cap what can ever leave a pool at what honestly entered it. Those totals cannot see a counterfeit hidden inside a pool, and we say so plainly.
Pools designed to be sealed and replaced. When a pool fills up it closes, a new one opens, and holders move their coins across.
A payment that gives little away. Every transaction is padded to 2, 4 or 8 actions and pays a fee set by the network, with no tips or bidding. So its size and fee reveal only which of those three it is.
A difficulty clamp. Difficulty can rise only gradually, which stops an attacker from packing a chain's weight into a few unusually heavy blocks. It can fall as far as it needs to, so the network recovers when miners leave. After a 99% drop in hashrate, that takes about 7.4 hours.
A second implementation of the state rules. tf-auditor is written from the specification, separately from the node. It recomputes every chain block and reports the first difference.

For miners

XelisHash-TF uses about half a megabyte of memory per hash, and we know of no ASIC for it. So far, only we have built mining software for it.
Free, open-source miners. The CPU miner and a stratum server exist and will be published with the code. A GPU miner and pool software will follow, and mainnet will not launch until they have run on the public test network.
Rewards are public: anyone can see their address and amount, which is how the supply can be checked. Spending them later is private. Rewards to the same address can be linked to each other, so keep an address just for mining and change it from time to time.
To show nothing was mined in advance, the mainnet genesis block must refer to a public event no more than 24 hours old at launch. We plan to announce the genesis time at least 30 days ahead.
The team will mine like anyone else: same announced start, same public software, no head start.

The coin

Issuance starts at 1 TFN per second and halves every two years, in 24 small monthly steps, so there is never a sudden cut.
After about 10.2 years it stays at 0.03 TFN per second, which keeps paying miners. There is no hard cap. About 89.65 million TFN will exist when this tail begins.
Half of every transaction fee is burned, and half goes to the miner.
A payment usually shows in the receiver's wallet within a minute. The receiver can spend it after about 10 to 20 minutes, similar to the roughly 20 minutes on Monero.
After about 12 hours a block is final: a node will not reorganise its chain below that point, a hard limit Monero does not have.
Development fund: 5% of new coins for the first four years, about 3.46 million TFN in all, then nothing. It never takes a share of fees. The fund pays the team and the project's costs. The network pays it to a fixed address, and no key or vote can send it anywhere else. That address will have a published viewing key, so anyone can see what it receives and spends. On mainnet the spending key will be split between team members, so no single person can spend from the fund alone.

Where it stands

Running today: a rehearsal of the test network launch, with four nodes on four machines. So far it has passed 30,000 blocks with no chain disagreement in any check. It is now running at one block per second, and wallet payments are going through between machines. An earlier 24-hour run of four nodes on three hosts, with shortened finality and pruning depths, accepted all 9,671 payments submitted.

Built: the node with the shielded ledger, a CPU miner, a stratum server, a command-line wallet and tf-auditor, covered by more than 2,100 automated tests and 26 fuzz harnesses of our own code.

Still to come before mainnet:

Encrypted node connections and Dandelion++. Test builds connect in plain text until then, so treat them as not private.
Sync for late joiners.
The GPU miner, pool software, and the desktop and web wallets.
Benchmarks that set the final block size, capacity and fees.
A first tf-auditor run on a live chain, and a full pool-sealing drill.
Independent security audits, which mainnet requires, and a cryptographer's review of the reward notes. Neither has been done yet.

The open technical points are listed in the litepaper and the network facts.

What comes next

A private test network, run by the team.
The source code, then a public test network. Test coins have no value, and the network can be reset.
Security audits, then mainnet.
Later: Tungsten (TGN), a private synthetic asset that targets the oracle price of one US dollar, on the same chain. Girder (GIRD) takes the reserve's risk. Tungsten can be redeemed only for TFN from an on-chain reserve, pro rata if the reserve falls below 100%, so the dollar value is a target, not a guarantee. New tungsten can be minted only while the reserve ratio is at least 500%. Conversion amounts and the one-time addresses they use are public, so conversions are less private than payments, but who converts stays hidden as long as those addresses are not reused. So far, a draft specification and a reference model exist. It stays off at mainnet launch and can be switched on only after published criteria are met.

The team

We are anonymous, so don't take our word for anything. Read the litepaper, which separates what exists from what does not, and read the code, which will be public before anyone outside the team is asked to mine.

Come and try to break it

Pick the design apart now using the litepaper, then run nodes and mine the test network once the code and the network are out. Join the Discord to be there when the test network opens. Hard questions are welcome, here or on Discord, and you can report anything you find on Discord.

This post describes a project in development. Its rules and values may change. Nothing here is financial advice or an offer of anything.