33: ASMap, nix, and nix-nostr

Julien (guest) Jul 26, 2026 1:35:09

"There is no real address book for the internet."

Julien & Gigi take a walk.

Julien came to Sovereign Engineering through a strange path: Nix, Bitcoin Core, ASMap, Nostr, and a growing sense that packaging, routing, identity, and deployment are all part of the same sovereignty problem. The first part of the walk goes deep into ASMap: why Bitcoin Core cares which autonomous systems its peers live in, how eclipse attacks turn peer discovery into a security question, and why mapping the internet is harder than it looks.

The trade-off is familiar: IP addresses are useful but misleading. A single entity can control many non-contiguous ranges, routing data is messy, BGP was not designed as a cryptographic truth machine, and even better inputs like RPKI only cover part of the world. So Kartograf builds from several imperfect sources, and the asmap-data process turns map generation into a collaborative run with hashes, review, and a paper trail.

From there the conversation moves into Nostr and the social layer of infrastructure. Julien describes finding Nostr through the simple feeling that the primitives were right, using Amethyst and Zapstore, thinking about Git on Nostr, and watching FIPS make old assumptions about DNS, IP addresses, NAT, and transport layers feel suddenly optional.

The last thread is nix-nostr, or more precisely Julien's early nostrpkgs work: a Nix package set and deployment path for Nostr tools. The inspiration is nix-bitcoin, but the target is wider: relays, FIPS nodes, git servers, Blossom drives, Cashu mints, and other services that should be easy to compose, reproduce, and ship. If agents are going to run infrastructure for us, declarative systems start to look less like taste and more like leverage.

In this dialogue:

how Julien found a path into Bitcoin Core work through BTC++, Nix, and a conversation with Fabian Jahr
why ASMap exists as a Bitcoin Core feature, and what it tries to improve about peer bucketing
how eclipse attacks, Sybil attacks, AWS, Hetzner, and autonomous systems change the threat model for peer discovery
why IP address prefixes are only a heuristic for control
how regional registries, autonomous systems, BGP, RPKI, IRR, and RIPE RIS fit into the map-making problem
why no one can build a perfect ASMap, and why that does not make the project useless
how Kartograf aggregates imperfect data into a usable ASMap
why collaborative runs give the process a reproducible paper trail instead of relying on one person's machine
how ASMap files are now managed in the bitcoin-core/asmap-data repository
why Julien would still like more users to generate fresh maps for themselves
how Nostr first clicked as the right mix of clients, relays, keys, and trade-offs
why Zapstore makes Android distribution feel native to Nostr
why replacing GitHub means replacing hosting, search, issues, pull requests, CI, releases, and social discovery
how Git-on-Nostr, agents, and mobile-first workflows point away from old repository interfaces
why voice, text, and agent interfaces are starting to blur
how FIPS changes the address and transport conversation by routing through Nostr identities
why a FIPS route for Bitcoin Core peers would be worth testing
why Julien sees Nix as sovereign infrastructure, not just a package manager
how reproducible builds, dependency graphs, hashes, and declarative systems connect Nix back to Bitcoin instincts
why nix-bitcoin is the obvious reference point for packaging Nostr infrastructure
what nostrpkgs is trying to do for Nostr tools, binary caches, and deployment recipes
why Nix is unusually friendly to agents because the desired machine state is written down as text
how demo day, Napplets, FIPS, Archipelago, and Sovereign Engineering changed Julien's first-week view of what can be built
why the right way to reach younger users is to build tools they love before explaining why the architecture matters

Related dialogues:

People mentioned:

Projects & tech mentioned:

Recorded at 959,604.

← All Episodes

33: ASMap, nix, and nix-nostr

Applications are now closed