Welcome to issue 206 of the IPFS newsletter, the first for this year. Just as spring starts for most of us in the northern hemisphere and nature begins to bloom, we have a lot of exciting updates.

In this issue we cover improvements to IPFS Mainnet making self-hosting practical at scale, Matadisco, a new data discovery network built on AT Protocol, guides and a case study on scientific data on IPFS, composable web apps built on content addressing, CIDs that finally match across implementations, and more!

🌟 Highlights

🛣️ IPFS Mainnet: 97% fewer DHT lookups and 10x faster DHT provides

IPFS Mainnet, the public network that IPFS nodes join by default, has seen significant improvements that make self-hosting more practical, especially for larger datasets. A trio of improvements to the Amino DHT has reduced the number of DHT lookups required to find content by 97% and sped up content publication by over 10x:

• Provide Sweep batches DHT provide operations into smooth, predictable sweeps instead of spiky bursts, letting resource-constrained nodes advertise large pin sets without overwhelming their connections. Shipped in Kubo 0.39. (Shipyard)
• Optimistic Provide complements Provide Sweep and reduces content publication time by over one order of magnitude while simultaneously reducing network overhead by 40%. The technique uses statistical network-size estimation to terminate walks early. Shipped as default in Kubo 0.39. (ProbeLab)
• Delegated routing caching in Someguy v0.7 adds a cached address book and active peer probing, eliminating ~83% of extra peer lookups and reducing P95 latency by 29% for browsers and light clients that can't run a full DHT. (Shipyard)

These are the fruits of years of research and implementation work across multiple teams including Shipyard, ProbeLab, and the IPFS Foundation.

🪩 Matadisco: An Open, Decentralized Network for Data Discovery

Open data is everywhere, but finding it is still painfully fragmented across siloed portals. Matadisco is a new open network for data discovery built on AT Protocol. The idea is simple: anyone can build producers that publish lightweight records to point to metadata records, and anyone can build consumers, i.e., specialized discovery interfaces that consume matadata records from the ATProto firehose or PDSs. The schema is deliberately minimal and works with any metadata standard (STAC, DataCite, IIIF, RSS).

For example, the matadisco-viewer streams Copernicus Sentinel-2 satellite imagery from matadisco records via the Bluesky Jetstream relay. For the full rationale, read the IPFS Foundation blog post by Michelle Lee and Volker's technical companion post with implementation details and code. (IPFS Foundation)

🧰 The DASL Toolbox Expands: Web Tiles, Video, and Perceptual Fingerprints

The DASL (Data-Addressed Structures & Links) spec family keeps growing. Three new specs landed since last issue:

• Web Tiles are composable web apps that can safely run in arbitrary contexts by restricting network access. Content is declared in a MASL manifest and can be stored on AT Protocol or packaged as CAR files. Try the tile browser or read the spec. (IPFS Foundation)
• BDASL extends DASL CIDs with BLAKE3 hashing for large files, enabling streaming verification and verified HTTP range requests — so you can fetch and verify arbitrary byte ranges from petabyte-scale datasets without you (or anyone in between) having to buffer the whole thing.
• Perceptual Fingerprints (PFP) define structured identifiers for perceptual hashing. Unlike CIDs that identify exact byte sequences, PFPs identify similar media — useful for content moderation and deduplication across decentralized systems. (Hypha Co-op)

And extending DASL into video: S2PA and MUXL tackle the problem that the same video through ffmpeg twice produces different bytes (and different CIDs). MUXL constrains MP4 to a canonical form so identical video always produces the same CID, while S2PA adds signing with secp256k1 and DID-based identity for video provenance. (Streamplace)

Reproducible CIDs Across the Stack

UnixFS, the default format for deriving CIDs, is intentionally flexible: chunk size, DAG layout, width, and other parameters can all be tuned, resulting in different CIDs for the same input data. That flexibility is useful for optimizing access patterns and deduplication, but it leads to confusion when the same file produces different CIDs across implementations.

IPIP-0499 fixes this friction point by introducing named CID profiles:

• unixfs-v1-2025: CIDv1, sha2-256, 1 MiB fixed-size chunks, balanced layout, 1024 links per node, raw leaves, block-bytes HAMT estimation
• unixfs-v0-2015: best-effort reproduction of legacy CIDv0 defaults from Kubo 0.39 and earlier go-ipfs versions.

In practice, given a profile name, CID derivation is now deterministic across implementations.

Support is already live in Kubo 0.40, Boxo 0.37, and Helia/unixfs 7.0.3.

IPFS for Geospatial and Scientific Data

New docs and a compelling case study show IPFS gaining traction in scientific data workflows:

• ORCESTRA case study: When the ORCESTRA atmospheric science campaign's planned server infrastructure got stuck in customs in Barbados, researchers improvised with a Raspberry Pi running Kubo and laptops sharing data peer-to-peer. What started as an emergency fix became their core data infrastructure, publishing Zarr arrays with CID-based verification across 20+ institutions. A great example of IPFS solving real problems and enabling global scientific collaboration.
• Landscape guide: A new starting-point reference for scientific teams evaluating IPFS, covering architectural patterns, content routing approaches, and the current state of tooling.
• Publishing geospatial Zarr data: A step-by-step guide covering data preparation, chunk sizing, IPFS upload with the new unixfs-v1-2025 profile, and content discovery via CIDs, IPNS, or DNSLink.

CIDs and ATProtocol

CIDs, originally conceived as part of IPFS, play a foundational role in ATProtocol's data model. Nick Gerakines from the GitHub engineering team wrote a two-part deep dive: Part 1 covers how ATProto uses CIDs and why this enables content verification without server trust. Part 2 digs into how AT-URIs relate to CIDs via MVCC-style versioning, Merkle Search Trees, and CAR file transport. A solid developer reference for the intersection of these two ecosystems.

✨ Ecosystem Spotlights

radicle-artifact is a new Collaborative Object for Radicle, the peer-to-peer code collaboration stack built on Git. It closes the "last mile" gap between verified source code and verified artifacts — letting developers distribute release artifacts addressed by (B)DASL CIDs in a transport-agnostic way, with iroh-blobs support built into the radicle-artifact CLI. Built by Radicle contributor and IPFS developer advocate, Daniel Norman.

Omnipin is a CLI tool for decentralized website deployment — it pins content to multiple IPFS (and Swarm) providers in one command and updates ENS names automatically via Gnosis Safe. (v1rtl)

ATFile is a command-line tool written entirely in Bash for storing and retrieving files on ATProto. It supports encrypted uploads via GnuPG, file streaming, and multiple PDS implementations.

browser-source bridges browser file APIs to Helia — use <input type="file">, drag-and-drop, or showOpenFilePicker() to add files directly to IPFS in the browser. (@tabcat)

Helia-adventure is a self-guided workshop for learning IPFS building blocks with JavaScript: CIDs, UnixFS, CAR files, and routers.

Simple Page lets you publish websites on ENS names using markdown, with content stored on IPFS and history notarized on-chain. The editor ships with the site itself, so even if the project disappears, your site won't. (@oed)

A new Helia private network example shows how to create an isolated IPFS swarm using pre-shared keys.

☁️ Services and Providers

Filebase Sites is a new product for publishing static websites on IPFS with stable IPNS addresses and custom domains — solving the "every update changes your CID" problem without the operational overhead of managing keys and gateways yourself.

Storacha 🔄 Fil One

Storacha recently announced the next phase in their evolution — the technical team is joining the Filecoin Foundation. The goal is to integrate their open source technology to accelerate the development of Fil One, a new S3-compatible Filecoin storage product that provides a unified Filecoin client experience.

As part of the evolution, the Storacha developer upload and IPFS service will sadly be shutting down on May 31, 2026; so make sure to migrate to a different service, or use the opportunity to give self-hosting a try with all the recent improvements to IPFS Desktop and Kubo.

📐 Protocol and Standards

Several specification milestones this issue:

• UnixFS — the spec defining how files and directories map onto content-addressed blocks in IPFS — covering chunking, directory sharding, metadata, and path resolution — is now ratified. If you've ever imported a file into IPFS, this is the format it used under the hood. @lidel
• Kademlia DHT — the routing spec for how IPFS nodes discover peers and locate content through the Amino DHT — is also ratified. It covers provider records, IPNS value storage, and the lookup procedures that make ipfs get work without a central index. @guillaumemichel
• IPIP-0476 adds GET /routing/v1/dht/closest/peers/{key} to the Delegated Routing API, letting browser nodes and other lightweight clients discover nearby peers via a simple HTTP call instead of running a full DHT. Useful for finding circuit relay servers without the overhead. @achingbrain and @lidel
• The CID specification is now an open draft in the IPFS specs repo, standardizing scope, binary encoding, string representation, and decoding procedures for CIDv1. Every implementation already does this, but until now there was no single authoritative reference. @darobin

🔧 Code and Tools

🚢 Releases

• Kubo v0.40.0 — Reproducible CID Profiles, connectivity diagnostics, improved ipfs dag stat output, improved content providing after startup, new command to publish signed IPNS records without key, and more.
• Service Worker Gateway v3.1.5 — Major internals overhaul: simplified UX (no more config page), _redirects support for SPAs, ipfs:// protocol handler registration, and significantly faster page loads. Now live at inbrowser.link.
• iroh 0.97.0 — Adds pluggable transport support, opening the door to Bluetooth, WebRTC, and beyond. Also ships noq, iroh's own QUIC implementation, replacing the previous dependency. 1.0 is on the near horizon.
• Someguy v0.11.1 — includes support for GET /routing/v1/dht/closest/peers/{key} per IPIP-476
• Boxo v0.37.0
• Rainbow v1.22.0
• ipfs-check v0.9.1
• IPFS Cluster v1.1.5
• IPFS Desktop v0.48.0
• @helia/verified-fetch v7.1.0

📚 Articles and Tutorials

• Content Addressing 2025 Year in Review — Robin Berjon's comprehensive look at the shift toward modularity, the DASL spec family, ecosystem wins, and the state of content addressing entering 2026.
• 📘 Content Addressing Data Sets — A new docs guide comparing three binary formats for content-addressing file collections: UnixFS, iroh collections, and MASL/DRISL. Covers the tradeoffs between flexibility, determinism, and overhead to help you pick the right approach.
• Finding Your Files in Decentralized Storage Just Got Easier — Starling Lab introduces the CAR Content Locator, a browser-based tool for tracking individual files within CAR archives across Filecoin and IPFS. Built for institutional needs (USC Digital Repository, 1.2+ PiB of archives).
• How to Migrate IPFS Websites from Fleek — A practical migration guide from Shipyard after Fleek discontinued its hosting service. Covers how the IPFS project migrated 15+ websites to a modular stack: GitHub Pages, Kubo, IPFS Cluster, and DNSLink.
• IPLD 2025 Review: From Monoliths to Modules — Volker Mische covers the migration of Rust libipld to modular crates, Python DAG-CBOR getting ~2x faster, and DASL as a simpler on-ramp for new projects.
• Shipyard 2025: Bringing IPFS Home — Shipyard's year-in-review covering seven Kubo releases, DHT improvements, AutoTLS, and browser-based retrieval.

🎓 Research Papers

• Large-Scale Measurement of NAT Traversal for the Decentralized Web: A Case Study of DCUtR in IPFS — published April 14, 2026, co-authored by Dennis Trautwein from the ProbeLab team.

🌐 Community & Events

Recent Talks

Several IPFS-related talks at FOSDEM 2026 (Feb 1, Brussels):

• 🎥 DASL Your Protocols! — Robin Berjon tours the DASL spec family and how to reuse it in your own protocols.
• 🎥 Local-First P2P with OrbitDB — Nico Krause on building offline-first apps with js-libp2p, IPFS, and OrbitDB, including a working todo PWA demo.
• 🎥 Bringing Decentralization to Your Doorstep — Mosh Lee on five years of getting IPFS into browsers, including the three-year effort to land Ed25519 in WebCrypto.

IPFS at Eurosky Live Berlin — Robin Berjon on ATProto's relationship to IPFS and the importance of independent AT Protocol infrastructure in Europe.

ATmosphereConf 2026 (Mar 26-29, Vancouver) — The global AT Protocol community conference featured several IPFS-adjacent talks:

• Rewilding the Internet with ATProto — Arushi Bandi on moving beyond replicating existing platforms and rethinking what the social internet is for.
• Matadata! Publishing Scientific Data Straight to AT — Robin Berjon on publishing and retrieving scientific metadata directly on AT Protocol.
• WebTiles Showcase — Ted Han and Robin Berjon demonstrate composable web apps built on DASL.
• Feature / Product / Business — Mosh Lee on frameworks for sustainable ATProto projects.

Upcoming Events

• CID Congress: An ongoing series of working sessions focused on making CIDs more deterministic and compatible across implementations is meeting next on Thursday, May 21 2026.
• Neocypherpunk Summit by Web3Privacy (June 14, Berlin) — A one-day event on human rights, open-source, decentralized tech, and practical activism. Juan Benet, the creator of IPFS, and Robin Berjon from the IPFS Foundation will both be speaking.
• IPFS (JS & GO) Triage meets Tuesday, April 21 2026
• Helia (IPFS & JS) Working Group meets Thursday, April 23 2026

Check out the full IPFS community calendar for more events.

Want to contribute to the newsletter or share your feedback? We'd love to hear from you! Submit your suggestions and feedback by simply replying to this email and join the IPFS forums.

If you made it this far, thanks for reading!