TestingCatalog
Public MindPattern findings, entities, and graph evidence that cite this source.
Findings
8
All-time hits
8
High value
2
Last seen
2026-05-24
Connected entities
Related findings
- 2026-05-24 / TOOLSMythos 1 Model Strings Spotted in Claude Code — Anthropic's Security AI Moving Toward Developer ToolsReddit users discovered product strings referencing 'Claude Mythos' model access inside Claude Code and Claude Security builds, suggesting Anthropic is preparing to integrate its vulnerability-finding Mythos system into developer-facing workflows. TestingCatalog confirmed the strings reference access to the Claude Mythos model in both tools. If shipped, this would embed Anthropic's security scanning — which has already found 10,000+ critical vulnerabilities across 50 partners — directly into the coding workflow.
- 2026-05-07 / RESEARCHClaude Orbit Leaked: Anthropic's Proactive Briefing Assistant for CoworkCode references to 'Orbit' surfaced across Claude web and mobile builds — a proactive, opt-in briefing system that auto-generates insights from Gmail, Slack, GitHub, Calendar, Drive, and Figma without prompting. Users can deploy 'Orbit apps' pinned to their workspace. This is Anthropic's answer to OpenAI's ChatGPT Pulse (shipped September 2025), differentiated by explicit GitHub and Figma connectors targeting developer/creative workflows.
- 2026-05-02 / RESEARCHManus Launches Cloud Computer: Always-On Persistent Agent Hosting Without CodeManus shipped Cloud Computer, giving users persistent Ubuntu Linux machines (SSH and web terminal) that retain state across sessions — a major upgrade from ephemeral sandboxes. Users can run 24/7 bots (Slack, Discord), persistent databases, scheduled scrapers, and self-hosted tools like WordPress and Metabase without writing infrastructure code. Available across Basic, Standard, and Advanced tiers. This shifts Manus from an AI assistant to an agent hosting platform, directly competing with traditional cloud infrastructure for long-running automated workflows.
- 2026-04-14 / AGENTSOpenAI Codex Scratchpad: Parallel Agent Task Execution from TODO List View Coming to Desktop AppOpenAI is building a Scratchpad feature for Codex that allows users to trigger multiple Codex agents in parallel from a TODO list view within a unified desktop app merging ChatGPT, Atlas browser, and coding tools. Heartbeat references in the codebase suggest managed agents that run in the background, check in periodically, and execute multi-step workflows without constant user input. This directly parallels Anthropic's Epitaxy coordinator mode, signaling both companies are converging on multi-agent desktop orchestration as the next UX paradigm for coding assistants.
- 2026-04-14 / DISPATCHClaude Code 'Epitaxy' — Anthropic Testing Full IDE Vision with Multi-Repo, Coordinator Mode, and Parallel Sub-AgentsAnthropic is finalizing a major upgrade to Claude Code codenamed 'Epitaxy' that transforms it from an assistant into a full development environment. The update introduces a Cowork-style layout with dedicated panels for Plan, Tasks (sub-agent tracking), and Diffs — plus the ability to work across multiple repositories simultaneously. A new Coordinator Mode would let Claude act as an orchestrator delegating implementation across parallel sub-agents. Multiple sources (TestingCatalog, TimesOfAI) report both Anthropic and OpenAI are preparing to release desktop app updates next week.
- 2026-04-12 / MARKETSNotion Tests Computer Use for AI Agents — Custom VM Environments, Model Selector, and Claude Code IntegrationNotion is testing a 'Computer' feature for its custom AI agents that adds managed VM environments where agents can install dependencies, execute code, and operate browsers. The feature includes a Computer Use model selector, trusted URLs configuration, and custom scripted workflows. Combined with Notion's existing 20+ minute multi-step agent capabilities and state-of-the-art memory system, this positions Notion as a potential full development and automation environment — not just a doc tool. If shipped, a Notion agent could research, write, code, test, and deploy without leaving the workspace.
- 2026-04-04 / VOICESAnthropic Secretly Testing Conway: An Always-On Agent With Extensions, Webhooks, and Chrome UseLeaked code reveals Anthropic is building Conway, a standalone Claude agent environment featuring persistent instances, custom extensions (.cnw.zip packages), webhook-based triggers from external services, and Chrome browser integration. Conway introduces its own UI with Search, Chat, and System panels, plus a Manage area for installing third-party tools and context handlers. The architecture suggests Anthropic is moving toward always-on agents that remain active beyond a single session and respond to external events autonomously — a significant strategic shift from their current chat-first model.
- 2026-03-31 / VOICESAnthropic Operon: Biology Research Agent with CRISPR Design and RNA Analysis Surfaces in Claude DesktopTestingCatalog discovered 'Operon,' a specialized biology research workspace that appeared in Claude Desktop on March 27, sitting alongside Chat, Code, and Cowork modes. Operon supports CRISPR screen design, single-cell RNA analysis, phylogenetic trees, and protein language models — targeting computational biologists, pharma R&D teams, and genomics researchers. Named after the gene cluster concept in molecular biology, Operon appears to be in internal testing with public access expected within 1-3 months.