The Quiet Death of Open Source at Loophole Labs
In January 2025, Loophole Labs told Hacker News that "all core components are open source." Nine months later, every one of those projects was archived. No announcement. No blog post. No goodbye.
This is a story about a startup, some AGPL code, and what happens when the money comes calling.
Who is Loophole Labs?
Loophole Labs is a New York-based startup founded in 2019 by Shivansh Vij. They've raised $13.4M from investors including Cervin Ventures, Materialized View Capital, Deep Acre, and Haystack. Their pitch: rethinking cloud infrastructure from first principles.
The company built genuinely interesting technology for live migration of Kubernetes workloads. Their open source projects included:
- Drafter - A compute primitive for live migration (AGPL-3.0)
- Silo - A storage primitive for migrating data while in use (AGPL-3.0)
- Scale - A WebAssembly plugin framework (Apache-2.0, 632 stars)
The technology was real. You could snapshot a running Kubernetes pod, migrate it across cloud providers, and restore it elsewhere with active network connections preserved. That's not vaporware.
The Timeline
Here's how the pivot unfolded:
- June 2024 - Last Scale release (WebAssembly framework abandoned)
- January 8, 2025 - HN post: "all core components are open source"
- June 1, 2025 - New funding from Materialized View Capital
- July 23, 2025 - Last Drafter release (v0.7.4)
- August 7, 2025 - Last Silo release (v0.2.21)
- September 30, 2025 - Silo archived
- October 15, 2025 - Drafter archived
- October 17, 2025 - Scale archived
The June funding sits right in the middle of this timeline. Draw your own conclusions.
The January 2025 HN post where they emphasized open source
What Replaced It?
Architect. A closed-source commercial product that hibernates Kubernetes workloads and wakes them instantly. It's built on the same technology as Drafter and Silo, but there's no public repo. It's in "early access" with a waitlist.
The product promises to run any Kubernetes workload on spot instances with zero downtime and 75%+ cost savings. No code changes required. The technology is compelling, but the source code is no longer ours to examine.
The Silence
What struck me most was the absence of any public conversation:
- No blog post explaining the archival
- No GitHub discussion about deprecation
- No HN thread about the change
- No tweets, no Bluesky posts
- The GitHub issues are all just dependabot bumps
The repos were quietly archived without explanation. The AGPL code remains available for forking, but Loophole Labs has moved on.
Why AGPL Matters
The AGPL license was an interesting choice. It's copyleft that requires anyone using the software over a network to release their modifications. Cloud providers avoid AGPL like the plague. That's kind of the point.
But here's the tension: AGPL protects community contributions, yet it also makes monetization harder. Companies won't pay for support when they can't use the software in the first place. And if you're taking VC money, "harder to monetize" eventually becomes "need to change strategy."
A Pattern We've Seen Before
This isn't unique to Loophole Labs. The pattern is well-worn:
- Build open source to gain adoption and credibility
- Raise venture capital
- Realize open source doesn't convert to revenue fast enough
- Archive the open source, ship the commercial product
HashiCorp did a version of this. Elastic did a version of this. Redis did a version of this. The details differ, but the arc is familiar.
What We Lost
The Drafter codebase was genuinely interesting. It used Linux USERFAULTFD for memory tracking, Firecracker for microVM isolation, and custom PVM patches for nested virtualization. This was low-level systems work that few teams attempt.
The code is still there, frozen in time
You can still fork it. The AGPL code is there. But without active maintenance, it becomes a snapshot of what was rather than a living project. The knowledge embedded in that codebase disperses.
What We Can Learn
If you're building on open source infrastructure from a VC-backed startup, here's the uncomfortable truth: the license is a snapshot, not a promise. AGPL today doesn't mean AGPL tomorrow. The codebase you depend on might be archived next quarter.
This isn't cynicism. It's pattern recognition. When you see:
- A small team with ambitious scope
- Venture funding
- Copyleft licensing that limits commercial adoption
- A pivot from one technology area to another
You should plan for the possibility that the open source version has an expiration date.
The Cooperative Alternative
There's another way to build software. Instead of venture capital that demands exponential growth, we could fund projects through cooperative models where users are stakeholders. Instead of AGPL as a defensive measure, we could use licenses that genuinely invite participation.
The challenge is that cooperative funding is slower and smaller. It doesn't build $13.4M war chests. But it also doesn't create the pressure that leads to quiet archival notices and abandoned issue trackers.
Loophole Labs made the choice that made sense for their investors. The question for the rest of us is whether we want to keep playing that game, or build something different.