Loading today’s stories
Build1 distinct publisher2 min readUpdated
Protocol Labs declined to renew Shipyard's funding, so the maintainers of Kubo, Helia, Boxo and the public gateways stop work on September 30. Anyone running IPFS in production now has a dated dependency.
The Engineer · Build desk
Compiled by The EngineerSomething wrong?How this is made
Who owns what is the awkward part. Shipyard supplied the engineers; Protocol Labs owns the domains and infrastructure behind ipfs.io, dweb.link and the rest, and it decides what happens to them once Shipyard is gone [5]. Moving maintainers into a Delaware nonstock corporation [6] did not move the hostnames that other people's code has already hard-coded. The protocol layer is genuinely indifferent to who runs a gateway. A client config file is not.
Shipyard named this exposure itself at launch, asking aloud what happens when a single funder changes strategy or prioritizes different work [7]. The answer was supposed to be grants, commercial support and contracted engineering, with $3 million sought in community contributions and early backing from Optimism RetroPGF, Cloudflare, Pinata, Fission and CoopHive [8]. Protocol Labs remained the anchor anyway, and its non-renewal is what the August announcement points to [9]. Two years of diversification did not change who could switch the work off.
The load being handed over is not trivial. Shipyard's own July 23 analysis put ipfs.io and dweb.link at 614 million requests and 45 terabytes for 10 million daily users [11], numbers the publisher says have not been independently measured [12]. Divide the bytes by the requests and the average response is about 73 kB [1], which describes a web of small content-addressed objects rather than heavy media. The window those 614 million requests cover is not stated, so the per-user rate cannot be derived from it.
Shipyard also says it re-architected the gateways to carry roughly three times the traffic while cutting operating and maintenance cost by about 80% [13]. Per unit of traffic that works out near a fifteenth of the earlier cost [2]. Which cuts both ways: a successor inherits the efficient version, and anyone who lets it lapse and rebuilds later starts again from the pre-optimization bill.
For an operator the audit is mechanical rather than clever. URL builders and SDK defaults containing ipfs.io or dweb.link. Nodes still dialing Shipyard-operated bootstrap peers. Routing calls to delegated-ipfs.dev or Someguy. CI steps that reach check.ipfs.network [4][2]. Kubo, Helia and Boxo will still compile on October 1. What ends is the assigned engineer whose job was the next release and the next bug fix [2], along with Shipyard's contributions to go-libp2p, js-libp2p and IPFS specification work [3]. Nobody has published what Protocol Labs was paying for that [10], which means the people best placed to replace it are bidding blind.
Follow any of these and your For You feed starts watching them — no settings page required.
Ranked by verification strength, evidence, and original report placement.
Cameron Wood and Adin Schmahmann said in an August 24th blog post that they will wind down Shipyard's IPFS engineering, maintenance and infrastructure operations on September 30th, after Protocol Labs declined to renew its funding.
Shipyard said Kubo, Helia, Boxo, Rainbow, IPFS Desktop, IPFS Companion, Someguy, Service Worker Gateway and IPFS Check will lose their dedicated maintainers, meaning no assigned Shipyard engineers handling releases, new features, bug fixes or long-term stewardship.
Shipyard's contributions to upstream projects including go-libp2p and js-libp2p will cease, and its work on IPFS specifications, standards and technical coordination ends.
Shipyard plans to stop operating ipfs.io, dweb.link, check.ipfs.network, delegated-ipfs.dev, IPFS bootstrap nodes, and collaborative clusters including Wikipedia-on-IPFS.
Protocol Labs owns the associated domains and infrastructure and will decide what happens to them after Shipyard's departure.
Shipyard publicly launched on April 8th, 2024 as a Delaware nonstock corporation staffed by longtime IPFS and libp2p maintainers; Schmahmann described it as an "exit to community" and identified Protocol Labs as anchor financial partner for 2024 and 2025.
Evidence-backed comparisons of source perspectives and observed adoption signals. Read the methodology
Which Builder, Operator, and Investor concerns the observed source mix emphasized—not a truth score.
Evidence, demonstrated adoption, hype gap, incentives, and confidence are assessed independently, each on its own current evidence. How these are measured.
Single outlet, primary announcement, self-reported metrics
The core factual spine -- dated wind-down, the nine unmaintained projects, the services being switched off, Protocol Labs' ownership of the domains -- is specific, attributable to a named primary announcement and internally consistent. But it comes from one publisher with no independent corroboration, no funder comment, and the quantitative claims (traffic, cost reduction) are explicitly flagged as Shipyard's own unverified measurements.
Heavily used public infrastructure, self-reported scale
The affected software and endpoints are load-bearing for the IPFS ecosystem: core implementations (Kubo, Helia, Boxo, Rainbow), browser tooling, bootstrap nodes and gateways that Shipyard says served 614 million requests and 45 TB to 10 million daily users, with Cloudflare having routed support through the same infrastructure. Scoring is held below high because every usage number originates with Shipyard and no third-party telemetry or downstream operator confirmation is present.
Framing tracks the documented facts, slightly conservative
The article's claims stay at or below what the evidence supports: it attributes the shutdown to a dated announcement, labels Shipyard's traffic and cost figures as self-reported, and stops short of asserting that gateways will go dark, noting instead that Protocol Labs owns the domains and will decide. If anything the consequence side is under-drawn -- a 37-day handoff for nine core projects plus bootstrap and gateway infrastructure serving self-reported 10 million daily users is treated soberly rather than amplified.
Self-interested primary source, withdrawing funder silent
Nearly all substantive material originates with Shipyard, an organization announcing its own closure while having previously sought $3 million in community contributions -- giving it reason to emphasize both the scale of the traffic it serves and the efficiency of the stack it built. Protocol Labs, whose funding decision drives the story and which owns the domains, provides no rationale or comment, and none of the named early supporters are heard from, so the incentive picture is one-sided but plainly labeled by the publisher.
Facts firm, aftermath unresolved
Confidence is solid on what was announced and what stops -- dates, project lists and service lists are concrete and attributable -- and weak on everything downstream: no disclosed funding amount, no successor, no Protocol Labs position, and only one publisher in the cluster. Self-reported metrics and the absence of any counterparty voice keep this in the moderate band.
build
py-libp2p bounds a dev-only /sdp endpoint that believed whatever Content-Length it was told1 distinct publisher
product
A 2x LLM bill is not a bug report: token spend is an observability problem1 distinct publisher
build
qm start said nothing, the UI said stopped, and only host dmesg named the passthrough failure1 distinct publisher
build
796 pages of semantic search with no vector database, and what it cost to skip one1 distinct publisher
Distinct publishers with included, body-backed reporting in this cluster.
1 article · August 24, 2026