Entity trail

Orchard

Source-backed findings, relationship evidence, citations, and briefing history from the public MindPattern archive.

Briefing refs

Findings

Edges

Sources

Corpus findings

2026-06-05 / hn-researcherAI-Assisted Audit Using Opus 4.8 Uncovers 4-Year Zcash Orchard Bug; ZEC Drops Over 30%Security engineer Taylor Hornby, using a custom AI auditing framework paired with Anthropic's Opus 4.8 (released May 28), discovered a critical Zcash Orchard vulnerability on May 29 that could have minted unlimited undetectable counterfeit ZEC. The bug had been live since Orchard's activation in May 2022, evading years of cryptographer scrutiny; an emergency two-phase network upgrade completed the fix June 2. ZEC fell ~31% on disclosure, and the chain's privacy design makes it cryptographically impossible to confirm whether the flaw was ever exploited.
2026-05-16 / arxiv-researcherOrchard: Open-Source Agentic Framework Sets SWE-Bench SOTA at 67.5% with Qwen3-30BOrchard introduces credit-assignment supervised fine-tuning to learn from productive segments of unresolved trajectories, plus Balanced Adaptive Rollout for sparse-reward RL. Using Qwen3-30B-A3B-Thinking, Orchard-SWE achieves 64.3% on SWE-bench Verified after SFT and 67.5% after RL — new SOTA among open-source models of comparable size. Orchard-GUI reaches 68.4% average across WebVoyager/Online-Mind2Web/DeepShop, competitive with proprietary systems from OpenAI and Google. Distills 107K trajectories from MiniMax-M2.5 and Qwen3.5-397B.
2026-04-12 / hn-researcherCirrus Labs Joins OpenAI to Build Agent Infrastructure — 267 Points on HNCirrus Labs, the bootstrapped company behind Tart (most popular Apple Silicon virtualization tool) and Cirrus CI, announced it's joining OpenAI's Agent Infrastructure team. The team will relicense all source-available tools (Tart, Vetu, Orchard) under more permissive licenses and stop charging licensing fees. Cirrus CI shuts down June 1, 2026. This signals OpenAI investing directly in the 'agent runs code in sandboxed environments' infrastructure layer — a talent acqui-hire for the people who built macOS virtualization tooling.

Graph relationships

BENCHMARKED AGAINST
SWE-Bench
Orchard achieved 67.5% on SWE-bench Verified after RL training.
Source finding
USES
Qwen3-30B-A3B-Thinking
Orchard uses Qwen3-30B-A3B-Thinking as its core model.
Source finding
USES
Qwen3.5-397B
Orchard distilled trajectories from Qwen3.5-397B for training.
Source finding
USES
MiniMax-M2.5
Orchard distilled trajectories from MiniMax-M2.5 for training.
Source finding

Source trail

Unchained GitHub/HN arXiv Source Source Source Source Source

Graph sources

entity graphfindings textkg edgeskg entitiesnewsletter issues