Fedimint gets real

Context: Recorded Tuesday, 30 June. The conversation focused on Fedimint being used in practice in South Africa, rather than the protocol itself.

Guests

Key Highlights

The South African federation

• It is the second federation set up by the group; the first was a 3-of-4 test federation. The live one is a 5-of-7 guardian federation — the largest known, the first 5-of-7 running exclusively on Start9 servers and the first on iro.
• Hermann uses it "almost on a daily basis" and reported that not a single Lightning transaction has failed, despite several guardians being on mobile or wireless internet rather than fibre, and despite South African power cuts (guardians use UPS battery backups; the federation tolerates two nodes down).

iro integration

• Removes the old DNS requirement and cloud/VPS complexity; communication is end-to-end encrypted with static public keys.
• The iro library uses all available network interfaces (Wi-Fi, mobile, etc.), switching and load-balancing automatically — federations have run across three continents simultaneously (Nairobi, South Africa, US), and a guardian could theoretically run in a pickup truck over Starlink.

Setup and backups

• Setup was "a click-through process": sync the Bitcoin node (a few days to a week on poor internet), install the Fedimint package (originally sideloaded; now in the Start9 community store), exchange guardian codes via Signal chat, and confirm. Hermann: "the most complicated part was not to get confused whose code belongs to who."
• Backups are static — made once, stored on a USB stick, containing private keys. Restoring is: click restore, upload the file, "30 seconds later they are online again."
• In the test federation, one guardian accidentally deleted and uninstalled his guardian — the system kept running, and he recovered from his Start9 backup.
• If a guardian loses both machine and backup, the federation can continue as an effective 5-of-6, or guardians can configure a shutdown date and successor federation — users get an in-app notification and migrate with one click. This was used to migrate users from the 3-of-4 to the 5-of-7 federation.

Federation sizes and consensus

• Recommended sizes: 4, 7, 10, then 13 guardians (e.g. 7-of-10 next). 5-of-7 is the current sweet spot; larger federations are possible (tests up to 20), limited by Bitcoin's rules for on-chain multisig, but mainly by coordination complexity.
• Thresholds are not arbitrary: the underlying Byzantine fault-tolerant consensus algorithm sets minimum thresholds — "5-of-7 is like the lowest threshold that is technically achievable." Going lower risks guardians disagreeing on confirmed transactions.
• Latency stays essentially constant as guardian count rises; only bandwidth increases slightly.

Gateways and Lightning

• The gateway is a separate service connecting a Lightning node to a federation — custody (guardians) is split from servers (gateways). Multiple gateways can serve one federation, and one gateway can serve multiple federations.
• Trust trade-off with gateways: essentially none — the gateway is "just an extension of the Lightning network into the federation"; like any routing node, it either forwards or earns nothing. The only reliance is that it doesn't deny service, against its own fee-earning interest.
• The Fedimint project runs its own gateway to bootstrap new federations; when a federation reaches critical volume it can spin up its own. Gateways run in the cloud because Lightning nodes need high uptime and good backups.
• Guardians vote on gateways via the UI, but even a single guardian can unilaterally add a gateway to unblock a federation; clients try gateways ordered by votes.

Wallets and South African context

• Users are on Conduit, a new wallet Joscha developed specifically for South Africa (in app stores a few months), supporting MoneyBadger QR codes. Joscha lived on the first federation for three months in South Africa with no failed payments.
• Hermann: "you can probably to 90% live on Bitcoin" in South Africa and Kenya; he landed in Kenya with only a Bitcoin wallet and never touched shillings. He keeps a bank account solely for visa proof-of-funds requirements.
• Zero fees for eCash transactions between federation members — a better trade-off than onboarding everyone to a fully custodial wallet.

UTXO management (V2 wallet module)

• Hermann's federation runs the V2 generation of modules; the wallet V2 module consolidates every deposit into a single UTXO — the federation only ever holds one UTXO.
• Each user pays exactly the on-chain cost they impose, avoiding socialised fees, fee-pool exploits via dust, and management burden (zero config). The federation requires roughly next-block fee estimation — it won't let transactions sit pending 12 hours since the UTXO is a shared resource.
• Lightning transactions are effectively batched into large on-chain transactions; gateways rebalance by sending funds (e.g. via Bolt Exchange) into the federation for more eCash.

Custody vs. chat (Fedi app)

• Only custody moves to the community federation; Fedi's chat remains cloud-hosted, using the same encryption protocol as Signal. The underlying Matrix protocol could allow decentralisation later, with MLS-based options like Whitenoise mentioned as promising.

Broader philosophy

• Eric's original fear was that in hyperbitcoinisation, most people would simply keep Bitcoin with JPMorgan-style banks — Fedimint offers a pragmatic middle ground rather than a black-and-white self-custody debate.
• Eric argued many "fully decentralised" projects in the wider crypto space "end up with basically a federation of extra steps" — being honest about having a multisig unlocks features, e.g. Fedi's module letting users hold synthetic USD to hedge short-term Bitcoin exposure.
• On privacy: eCash gives "incredibly strong default privacy guarantees" to people who couldn't otherwise afford it.
• On education: "if we don't have to explain, we shouldn't" — users get eCash's benefits without needing the concept explained; eCash notes can be sent like chat messages or printed as paper vouchers for onboarding.

Final thoughts

• Hermann: Fedimint is "probably going to be one of the most important technologies in Bitcoin going forward" for creating an acceptable trust model — far preferable to onboarding people to fully centralised wallets, with a great user experience.
• Joscha/Eric: Fedimint is as much a social experiment as a technical one — "can we form the social structures around those systems that make them work for communities?"
• Eric: "Bear markets are for building. More than ever before, we have to fight for our privacy, fight for our freedom. Join us."

(Sponsor/audience-support segments at the start and end — zap leaderboards, sharing/subscription plugs — omitted as requested.)