Denoise · Twitter

The security conversation is shifting from model-level prompt injection to vulnerabilities in the agent's orchestration layer, with @AnthropicAI and @GoogleDeepMind leading the effort to define these new threats.

Major releases today focus on formalizing agent security, with red-teaming frameworks and responsible disclosures of complex jailbreaks.

• 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 convergence on the terminal as the new AI-native IDE is visible, with @AnthropicAI's new product and @karpathy's analysis suggesting a fragmentation of traditional GUI-based workflows.

The dominant theme is the arrival of powerful, terminal-native coding agents, exemplified by the launch of Anthropic's Claude Code 1.5.

• 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.

A new infrastructure layer is rapidly solidifying, as @OpenAI, @vercel, and @replit all ship primitives for agent orchestration, while @LangChainAI focuses on protocol interoperability.

A wave of new tools for deploying and orchestrating agents was released, including SDKs, edge runtimes, and deployment harnesses.

• 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.

While agent tooling dominates discourse, @MistralAI continues its pattern of releasing foundational data artifacts that directly benefit and accelerate the open-source training ecosystem.

A significant open dataset for web-scale Optical Character Recognition (OCR) was released by MistralAI.

• Mistral AI@MistralAIrising

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

With massive context windows now common, figures like @GregKamradt and tools like @mem0ai are converging on the idea that vector search alone is insufficient, requiring more sophisticated memory architectures.

The discussion is evolving from simple RAG towards more complex 'context engineering' and multi-layered memory systems 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.

SaaS platforms like @NotionHQ and @linear are embedding agent-like capabilities (auto-triage, chained updates), mirroring the broader industry trend of integrating autonomous logic into existing products.

Workspace automation tools are shipping major updates, moving beyond simple triggers to more autonomous, chained operations.

• 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.

The practice of prompt engineering is industrializing, with platforms like @weights_biases demonstrating that brute-force search for optimal prompts is replacing artisanal, intuition-based methods.

The focus is on systematizing prompt optimization through large-scale, data-driven benchmarking rather than manual tuning.

• 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.

@jerryjliu0's point on filtering synthetic data reveals a critical bottleneck for scaling agent capabilities: ensuring training data quality to prevent poisoned generalization.

The conversation highlights the nuanced challenges of curating high-quality datasets for training capable agents, particularly filtering synthetic data.

• Jerry Liu@jerryjliu0repeated

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