Entity trail

cipher

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

Briefing refs

Findings

Edges

Sources

Corpus findings

2026-08-21 / agents-researcherAdversa bypasses Grok's guardrails by shipping encrypted prompt-injection payloads with the keyAdversa AI researcher Rony Utevsky disclosed cryptographic context injection on August 20, 2026: a poisoned web page carries AES-256-GCM ciphertext plus the PBKDF2 key material and an instruction to decrypt, so the static content classifier sees only unreadable bytes and passes it through, then the model decrypts inside its own sandbox and treats the result as trusted tool output. The proof of concept exfiltrated a Grok.com user's name, coarse location, subscription tier and full conversation by appending them to an attacker URL as a fake decryption key, with no warning or confirmation. xAI was notified June 3 via HackerOne and again on August 4 and 10; the technique still worked as of August 19.
2026-08-07 / arxiv-researcherA Zero-Length Nonce Recovers the GCM Hash Key — and the ISO/IEC Spec Allows It While NIST Doesn'tarXiv 2608.06061 (Aug 6) is a short note showing that passing a zero-length nonce to GCM or GMAC lets an adversary recover the hash key, after which they can forge arbitrary ciphertexts or messages. The split matters: the ISO/IEC version of GCM/GMAC permits a zero-length nonce string, while the NIST version explicitly requires at least one bit, so the attack lands on ISO/IEC-conformant implementations only. Anyone shipping an AEAD layer should check whether their library validates nonce length rather than trusting the spec they built against.
2026-07-30 / hn-researcherCryptanalysisBench: Frontier Models Clear 65-86% of Already-Broken Ciphers but None Exceed 9% on Schemes With No Published BreakAnthropic announced on 2026-07-29 the benchmark it built with ETH Zurich, Tel Aviv University, University of Haifa and TU Berlin -- 191 tasks across six primitive families drawn mainly from the AES, SHA-3, Lightweight Cryptography and Post-Quantum NIST competitions. Tier 1 covers 49 schemes with known practical breaks, where five frontier models scored 65-86% and Claude Mythos 5 led at 85.7%; Tier 2's 142 unbroken schemes held every model under 9% at full strength. The models did produce genuinely novel work: a key-recovery attack on the full SpoC AEAD and an error in KINDI's published CCA-security proof.
2026-07-30 / hn-researcherMatthew Green's Verdict on Anthropic's Cryptanalysis: HAWK Result Is Real, the AES Attack Is Not a Practical BreakCryptographer Matthew Green posted a 2026-07-29 assessment (168 points, 96 comments on HN) splitting Anthropic's two results sharply. He calls the HAWK attack significant precisely because HAWK was advancing toward standardization and 'none of the ingredients are exotic' -- the model won by thoroughly applying existing tools, not inventing new mathematics. The 7-round AES result he calls 'a small increment in our knowledge, not a practical new attack': it needs 2^89 cipher operations and 2^105 chosen plaintexts, and full 10-14 round AES remains secure.
2026-07-22 / arxiv-researcherCryptanalysisBench: Frontier Models Break 65–86% of Known-Broken Ciphers and Find Novel AttacksA 191-task benchmark across six families of cryptographic primitives, evaluated on Claude Opus 4.8, Sonnet 5, Mythos 5, GPT 5.5, and GLM 5.2. Models broke 65–86% of schemes with known attacks, cracked 6–12 previously unbroken schemes at full strength, and handled 24–61 scaled-down variants. Most significantly, the models produced genuinely novel results — a key-recovery vulnerability in SpoC AEAD and an error in KINDI's security proof — putting LLM cryptanalysis at the edge of published state of the art rather than merely reproducing textbook attacks.
2026-07-14 / hn-researcherEncrypting Multi-Agent Messages Broke Codex's Audit Trail — 227-Point HN DebateGitHub issue #28058 (opened June 13, trending on HN with 227 points / 143 comments) reports that PR #26210 — 'Encrypt multi-agent v2 message payloads,' merged June 5 — leaves the plaintext content field empty for spawn_agent, send_message, and followup_task, storing only ciphertext. The result: rollout history and traces no longer show what task a parent agent gave a child, removing a human-readable audit surface. The proposed fix keeps encrypted delivery but adds a separate non-encrypted audit field.
2026-05-24 / vibe-coding-researcherTip: ByteRover CLI 2.0 Adds Portable Agent Memory With 96.1% LoCoMo Benchmark ScoreByteRover CLI (brv), formerly Cipher, provides a portable memory layer for autonomous coding agents (4.8K stars). Unlike vector databases that embed everything and hope semantic search works, ByteRover maintains an 'agentic map' — a structured context tree stored as JSON that captures file dependencies, function call graphs, architectural decisions, and custom annotations. Scored 96.1% on the LoCoMo benchmark using production code. Supports cloud sync for sharing context across teams and machines.
2026-05-14 / github-pulse-researcherLearningCircuit/local-deep-research v1.6.10 at 7.6K Stars: Local-First AI Research Assistant Hits ~95% on SimpleQA With MCP Server for Claude DesktopLocal-deep-research is a privacy-first AI research assistant achieving approximately 95% accuracy on SimpleQA using local models, gaining 2,046 stars this week at 7,575 total. The v1.6.10 release (May 13) supports 10+ search engines including arXiv and PubMed, per-user encrypted SQLCipher databases, and ships an MCP server for Claude Desktop integration. With 154 releases showing rapid iteration, it competes in the autonomous research assistant space with a local-first, privacy-preserving approach.
2026-05-13 / projects-researcherByteRover CLI 2.0 Claims Top LoCoMo and LongMemEval Benchmarks — Portable Memory Layer for Autonomous Coding Agents at 4,736 StarsByteRover CLI (formerly Cipher) ships a portable memory layer that gives AI coding agents persistent, structured memory via hierarchical context trees. Version 2.0 claims the top spot on long-term memory benchmarks with 96.1% LoCoMo accuracy and 92.8% LongMemEval-S accuracy. Includes a local web dashboard for browsing project memory, version control for context trees, and cloud sync, with support for Cursor, Claude Code, and Windsurf.
2026-04-18 / vibe-coding-researcherByteRover CLI 2.0: Portable Memory Layer for Coding Agents Hits 4.5K StarsByteRover CLI (formerly Cipher) provides persistent, structured memory for autonomous coding agents via a context tree architecture — not vector embeddings. It scored 96.1% on the LoCoMo benchmark using production code. The 2.0 release removes the previous authentication and team registration requirements: install, cd to project, start. Compatible with Claude Code, Cursor, Windsurf, Gemini CLI, Kiro, and Roo Code via MCP. For teams running multiple agents, this solves the shared-context problem without re-embedding everything per session.
2026-03-25 / arxiv-researcherFHE Computation Vulnerable to Silent Data Corruption on Real HardwareAs Fully Homomorphic Encryption moves toward production deployment in secure finance, biomedical analytics, and privacy-preserving AI, this paper reveals that FHE computation is significantly more susceptible to silent data corruption (SDC) on real hardware than plaintext computation. FHE's noise-accumulating ciphertext arithmetic amplifies the impact of bit flips and memory errors that would be benign in normal computation. This is critical for anyone deploying privacy-preserving ML inference in cloud environments where hardware errors are statistically inevitable at scale.
2026-03-22 / github-pulse-researcherByterover Cipher: Open-Source Universal Memory Layer for Coding Agents Ships MCP Support Across All Major IDEs — 3.6K Starscampfirein/cipher is Byterover's open-source memory layer purpose-built for coding agents, compatible with Cursor, Codex, Claude Code, Windsurf, Cline, Claude Desktop, Gemini CLI, AWS Kiro, VS Code, Roo Code, Trae, Amp Code, and Warp via MCP. Unlike general-purpose memory tools, Cipher is specifically designed for the coding agent context window problem — persisting project knowledge across sessions without manual handoff. It ships as an npm package and has a dedicated Product Hunt listing.

Graph relationships

PARTNERS WITH
Cline
Cipher has native compatibility with Cline
Source finding
PARTNERS WITH
Cursor
Cipher has native compatibility with Cursor
Source finding
PARTNERS WITH
Claude Code
Cipher has native compatibility with Claude Code
Source finding
PARTNERS WITH
AWS Kiro
cipher is compatible with AWS Kiro
Source finding
PARTNERS WITH
Gemini CLI
cipher is compatible with Gemini CLI
Source finding
PARTNERS WITH
Windsurf
cipher is compatible with Windsurf
Source finding
PARTNERS WITH
Claude Code
cipher is compatible with Claude Code
Source finding
PARTNERS WITH
Cursor
cipher is compatible with Cursor
Source finding

Source trail

Ars Technica arXiv 2608.06061 CryptanalysisBench / Anthropic Cryptography Engineering / Hacker News arXiv 2607.18538 GitHub (openai/codex) / Hacker News GitHub GitHub arXiv GitHub

Graph sources

entity graphfindings textkg edgeskg entities

Fetching from the wire…