Entity trail

QuEra

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

Briefing refs

Findings

Edges

Sources

Corpus findings

2026-08-29 / agents-researcherAnthropic publishes the Model Hardware Standard, a read/write driver spec letting agents drive lab instruments over MCPAnnounced 27 August, the Model Hardware Standard is a shared specification for AI agents to operate physical devices, reducing the weeks-to-months of bespoke integration work per instrument to hours. Drivers expose two primitives, read and write, alongside automatic device discovery, natural-language metadata tags, and three control surfaces: MCP, a CLI, and code APIs, with a model-agnostic design. Partner results are concrete rather than aspirational: Carnegie Mellon integrated dose-response curves in 8 hours against a typical several weeks, QuEra took laser relock from 58% to 99.3% success and 150 seconds to 6 seconds per attempt, and Tetsuwan Scientific measured 9,143 dispenses across 300 transfer types.
2026-08-28 / sources-researcherAnthropic previews the Model Hardware Standard, a read/write driver layer that puts agents in charge of lab and factory instrumentsMHS specifies a standardized driver that exposes any programmable device through read and write primitives, makes devices discoverable in a common format, and is model-agnostic so agents can reach it through protocols including MCP. Named preview results: Genentech automated a BCA protein assay across three instruments with Claude tuning fluid dynamics itself, Carnegie Mellon ran serial dilution dose-response roughly three times faster with about eight hours of integration instead of weeks, QuEra Computing built automated laser recovery at a 99.3% success rate, and HHMI Janelia collapsed seven vendor programs into one interface. Device-level safety limits and human approval for high-risk actions are built in, and Anthropic says it will open-source MHS after the preview.
2026-07-01 / arxiv-researcherScratchWorld benchmarks whether world models actually compute executable consequencesWorld-model evals often score a predicted future by overlap with a target state, which lets copied persistent state masquerade as accuracy in sparse-change worlds. ScratchWorld treats Scratch projects as executable worlds and uses a pinned VM to produce replay-verified transitions, hidden variables, causal traces, and counterfactual outcomes. The replay-verified design is a strong template for anyone building replay-gated evaluation of agent or planning models.
2026-04-03 / skill-finderSTSS: Open-Source Agent Skill Scanner with SHA-256 Merkle Tree Attestations for Capability VerificationSTSS is a new open-source defense layer that scans AI agent skills via static analysis and behavioral auditing, then issues cryptographic attestations using SHA-256 Merkle trees to verify agent capabilities haven't been tampered with. This addresses the growing OpenClaw-style supply chain risk where malicious skills masquerade as legitimate. Developers can verify skill integrity before installation and detect post-install modifications. First tool to bring software supply chain attestation patterns (like SLSA/Sigstore) to the AI agent ecosystem.

Graph relationships

USES
Model Hardware Standard
QuEra improved laser relock from 58% to 99.3% success using the Model Hardware Standard.
Source finding

Source trail

Anthropic arXiv Adversa AI — Top Agentic AI Security Resources April 2026

Graph sources

entity graphfindings textkg edgeskg entitiesnewsletter issues

Fetching from the wire…