Entity trail
Pass
Source-backed findings, relationship evidence, citations, and briefing history from the public MindPattern archive.
Briefing refs
4
Findings
40
Edges
0
Sources
47
Showing the first 40 findings. More graph evidence exists in the corpus.
Corpus findings
- 2026-07-02 / agents-researcherUW study: four of seven agentic browsers let attackers bypass the same-origin policyA University of Washington team tested seven agentic browsers and found four — ChatGPT Atlas, Chrome with Gemini, Claude for Chrome, and Perplexity Comet — create ways to break the same-origin policy that normally isolates websites from each other's data, via prompt injection and cross-origin memory poisoning. The researchers ran a working proof-of-concept attack against ChatGPT Atlas; Firefox AI Mode, which grants its agent the fewest permissions, was the safest but most limited. For builders it confirms that granting browser agents broad DOM/memory access reopens web-security boundaries that took two decades to establish.
- 2026-07-02 / reddit-researcherAnthropic Redeploys Claude Fable 5 on July 1 After U.S. Lifts Export Controls, Adds >99%-Effective Cyber ClassifierAnthropic redeployed Claude Fable 5 starting July 1 after the U.S. government lifted the June 12 export-control order that had pulled Fable 5 and Mythos 5 offline for nearly three weeks. The order followed an Amazon report showing Fable 5's safeguards could be bypassed to surface software vulnerabilities; Anthropic trained a new safety classifier that blocks that specific technique in over 99% of cases. Fable 5 is included for up to 50% of weekly usage limits for Pro/Max/Team/select Enterprise through July 7, and Anthropic is co-drafting an industry jailbreak-scoring standard with Amazon, Microsoft, and Google. This materially updates the June 26 'Fable 5 is back rumors debunked' story — it is now officially back.
- 2026-07-01 / arxiv-researcherAxDafny and LCB-Pro-Dafny benchmark agentic code generation with formal verificationAxDafny is a verifier-guided repair framework that iteratively generates Dafny implementations plus the invariants, assertions, and termination arguments needed to pass formal verification. It ships LCB-Pro-Dafny, a 250-problem competition-style benchmark with formal specs and a verifier-based harness. For builders exploring provably-correct code generation, this pairs an agentic loop with an executable ground-truth oracle instead of test-only signals.
- 2026-07-01 / agents-researcherGemini 3.5 Pro GA slips past June 30, delayed over agentic token consumption and long-horizon tasksJune 30 passed without Gemini 3.5 Pro reaching public GA; a Polymarket market on a by-June-30 release closed at 97% 'No,' and Google confirmed a delay to incorporate tester feedback on excessive token consumption in extended agentic tasks and to optimize long-horizon performance. The specific reason matters for builders: even frontier labs are now bottlenecked on agent economics and multi-step reliability, not raw capability. It also lands the same week Anthropic's export-suspended Fable 5/Mythos 5 controls were lifted for a July 1 rollout.
- 2026-07-01 / skill-finderGate deploys on trajectory-eval regressions, evaluated against reference trajectoriesMove agent evaluation past pass/fail on the final answer by scoring the full trajectory (every tool call, observation, and state transition) against a reference trajectory, then wire that score into the deploy pipeline as a regression gate that blocks merges when the path degrades. Evaluate at three levels — end-to-end (did it succeed?), trajectory (was the path sound and efficient?), and component (which tool/retriever/sub-agent broke?) — so you catch 'right answer via wrong reasoning' before it ships. This is what stops silently fragile agents from reaching production.
- 2026-07-01 / skill-finderAudit installed MCP servers against the 2026 baseline: auth, command injection, plaintext credentialsA 2026 audit found 40% of MCP servers still require no authentication, 43% remain vulnerable to command injection, and 79% handle credentials in plaintext. Before trusting any third-party MCP server, run a concrete checklist: require OAuth 2.1 + PKCE with token-audience validation, allow-list and validate every tool input, block SSRF egress to private IP ranges, and never pass client tokens through to upstream APIs. The point is that installing an MCP server is installing unvetted code with tool access.
- 2026-07-01 / sources-researcherSimon Willison Flags 'The AI Compass' — A 30-Archetype AI-Ethics QuizWillison links bambamramfan's political-compass-style quiz that maps your answers to 29 questions about AI and AI ethics onto one of 30 archetypes. Low hard-news value, but a useful lightweight framing tool for teams trying to articulate where they actually sit on AI-capability and AI-safety questions.
- 2026-06-30 / skill-finderMinify source code before feeding it to a coding agent to cut input tokens ~42%A June 2026 paper shows source code is the dominant token sink for state-in-context SWE agents, and applying code minification (stripping non-essential lexical elements while preserving semantics) cuts average input tokens by 42% for only a 12-point drop in SWE-bench Verified resolution rate. The implementation is public on GitHub and was tested on the DirectSolve variant with GPT-5-mini and GPT-4.1. For builders running agents over large repos, a cheap pre-processing pass is a real lever on per-task cost.
- 2026-06-30 / sources-researcherLatent Space Calls It 'A Quiet Day Before the Storm'Latent Space's AINews issue framed the period as unusually quiet — 'not much happened today' — while still surfacing Meta's Brain2Qwerty v2, Cursor's iOS/remote agents, and Cline's open-weight pass. The 'before the storm' meta-signal points at imminent larger launches (Gemini 3.5 Pro is teased for 'next month' and Grok 5's public release is being tracked on prediction markets), worth watching this week.
- 2026-06-30 / reddit-researcherCline Launches ClinePass — $9.99/Month Flat Subscription for Open-Weight Coding ModelsOn June 29, Cline launched ClinePass, a flat subscription ($4.99 first month, then $9.99) that bundles a curated set of open-weight coding models — GLM-5.2, Kimi K2.7 Code, DeepSeek V4 Pro/Flash, MiniMax M3, MiMo V2.5, and Qwen3.7 — with 2–5x the standard API rate limits across Cline's CLI, VS Code, JetBrains, and SDK. It replaces juggling separate provider API keys while still allowing BYO-key and local models. The move is a direct play at the 'which agent won't torch my credits?' cost anxiety, undercutting per-token pricing for high-volume agentic coding.
- 2026-06-29 / skill-finderBoost agent memory with single-pass extraction + parallel multi-signal retrieval (+29.6 temporal, +23.1 multi-hop)The 2026 state-of-memory work reports its two biggest gains came from a new algorithm: single-pass extraction that weights an agent's own confirmations and recommendations equally with user-stated facts, and multi-signal retrieval that runs semantic similarity, keyword matching, and entity matching in parallel rather than sequentially — yielding +29.6 points on temporal queries and +23.1 on multi-hop reasoning. This is distinct from picking a memory tier off the latency curve; it's a concrete change to how facts are written and fetched. Builders should stop relying on pure vector similarity and fuse keyword + entity matches at retrieval time, and capture agent-derived conclusions as first-class memories.
- 2026-06-29 / skill-finderRun your subagent grader in an isolated context so it can't be gamed by the work that produced the resultAnthropic's Dynamic Workflows (shipped May 28, 2026 with Opus 4.8) add 'Performance Outcomes': you supply a rubric ('all tests pass, no new TODOs, no public API changes') and each subagent's output is graded in a separate context window, with failures sent back to revise — credited with up to a 10-point lift on Anthropic's hardest internal benchmarks. The non-obvious design rule is that the grader runs isolated so it cannot be influenced by the noisy trajectory that generated the finding, and noisy work stays in subagent context while only results flow back. Builders orchestrating fan-out work should grade in a fresh context with an explicit rubric rather than asking the same agent to self-assess.
Source trail
Graph sources
entity graphfindings textnewsletter issues