Denoise · Twitter

The security conversation is shifting from prompt injection in single models (Anthropic) to systemic vulnerabilities in agent control planes (GoogleDeepMind, MalwareTechBlog).

Major labs are now publicizing red-teaming frameworks for autonomous agents, focusing on vulnerabilities in orchestration and tool interaction.

• Anthropic@AnthropicAIrising

  Responsible disclosure on a Claude jailbreak chain we patched last week. Full write-up including our red team timeline.

• Google DeepMind@GoogleDeepMindrising

  New red team framework for prompt injection in autonomous agents. Covers cross-tool leakage, scanner evasion, and sandbox escape patterns.

• MalwareTech@MalwareTechBlogrepeated

  Autonomous agent running pentest flows against a real SaaS. First real-world run: fewer false positives than I expected on the vulnerability surface.

A clear pattern emerges: developers like @karpathy and @levelsio are advocating for terminal-based agents over existing GUI/IDE tools like Cursor, framing it as a speed and workflow advantage.

The launch of Anthropic's terminal-native Claude Code 1.5, validated by influential developers, signals a potential shift away from IDE-centric coding assistants.

• Anthropic@AnthropicAIrising

  Claude Code 1.5 is live. Terminal-native coding agent with full Claude Opus reasoning, file-ops sandbox, and session replay.

• Andrej Karpathy@karpathyrising

  The developer-experience shift from IDE to terminal agent is underrated. Coding workflows are about to look nothing like 2024.

• swyx@swyxrising

  Codex vs Claude Code terminal agent benchmarks. Pass@1 diverges more than I expected on the long-context editor tasks.

• DSPy@dspy_airising

  DSPy 3.0: prompt optimization via compile-time search over system prompt variations. Benchmarks inside.

• @levelsio@levelsiorising

  Switched my whole editor setup to Claude Code this week. Shipping faster than when I used Cursor + Copilot.

There's a convergence towards standardized agent protocols, with @OpenAI releasing an SDK and @LangChainAI building integrations for Anthropic's MCP, abstracting away the orchestration complexity.

A race is on to provide the definitive infrastructure for deploying agents, with OpenAI, Vercel, and Replit all shipping orchestration and hosting solutions.

• OpenAI@OpenAIrising

  New agent SDK: protocol-level tool calling, deployment harness, and multi-worker orchestration primitives. Docs live.

• LangChain@LangChainAIrising

  MCP protocol integration thread. How to wire existing LangGraph agents into the Anthropic Model Context Protocol server spec.

• Vercel@vercelrising

  Edge runtime for agent workers is live. Spawn durable background agents from any serverless deployment.

• Alex Albert@AlexAlbert__rising

  When your security scanner finds nothing scary on an agent deploy, check the orchestration layer again. That's usually where the jailbreak sneaks through.

• Replit@replitrising

  New agent deployment harness. One command to go from local orchestration to hosted agent worker.

A quiet day focused on foundational artifacts; MistralAI's dataset release provides a key resource for training future open-source multimodal models.

MistralAI released a large-scale, cleaned OCR dataset for public use.

• Mistral AI@MistralAIrising

  Open dataset release: 100M-row web OCR dataset. Cleaned, licensed, ready to train.

Practitioners like @GregKamradt and startups like @mem0ai are converging on the idea that a simple vector index is insufficient, pushing for more structured memory architectures.

The discussion has evolved from simple RAG to 'context engineering,' exploring tiered memory systems and sophisticated caching strategies for agents.

• Vaibhav Srivastav@reach_vbrising

  Tested the new 10M context memory window end to end. Surprising failure modes around rag retrieval cache invalidation, thread below.

• Greg Kamradt@GregKamradtrising

  RAG is dead, long live context engineering. My framework for when to cache, when to retrieve, and when to just dump memory into the prompt.

• mem0@mem0airising

  Memory layer for agents: differentiating working memory from the subconscious store. Vector index isn't enough anymore.

• LlamaIndex@llamaindexrepeated

  Knowledge graph retrieval walkthrough: when semantic vector search misses, graph hop beats it every time.

A parallel trend to developer agents is emerging in productivity tools, where @NotionHQ and @linear are building non-technical 'agents' for workspace orchestration.

Workspace automation is a key theme, with both Notion and Linear shipping features that auto-manage tasks and data, mirroring agent-like behavior in a business context.

• Notion@NotionHQrising

  Notion workspace automation is out of beta. Auto-fill tables, chained updates across databases, and a new audit log surface.

• Linear@linearrising

  Linear now auto-triages incoming issues. Quiet launch, but already our favorite workspace feature of the year.

• Temporal@temporaliorepeated

  Orchestrating agents with durable workflows: replayable, resumable, and multi-worker by default. Walkthrough from our infra team.

• James Clear@jamesclearrepeated

  The best habit tracker is the one you actually open. Three open-source alternatives worth trying.

Efforts by @weights_biases to systematically benchmark prompts reveal a more rigorous, engineering-led approach to prompt optimization, moving beyond simple craft.

The focus is shifting from anecdotal prompt 'tricks' to large-scale, data-driven benchmarking of system prompts to find optimal performance.

• dotey@doteyrising

  Five prompt tricks learned this week from reviewing 200 production prompts. Short thread.

• Weights & Biases@weights_biasesrising

  System prompt benchmarking at scale: we ran 40k variants across 6 frontier models. The efficient frontier is not where you think.

The insight from @jerryjliu0 highlights a critical bottleneck in agent development: training data quality, especially for synthetic data that can degrade generalization.

Discussion centered on the nuanced challenges of data curation for training effective agents, specifically filtering harmful synthetic data.

• Jerry Liu@jerryjliu0repeated

  Dataset curation for agent training: how we filter synthetic data that looks good but poisons generalization.