Denoise · Twitter

The security discourse is shifting from model-level exploits to vulnerabilities in the agent orchestration and tool-use layers, as demonstrated by Anthropic's jailbreak analysis and @MalwareTechBlog's pentesting agent.

Disclosures from Anthropic and Google DeepMind show a focus on red-teaming autonomous agents, moving beyond simple prompt injection to complex, multi-step vulnerabilities.

• 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.

The competition in AI coding is moving from IDE plugins like Copilot to standalone terminal agents like Claude Code, with benchmarks from @swyx suggesting significant differentiation on complex tasks.

Anthropic's release of Claude Code 1.5, a terminal-native agent, marks a significant shift in developer workflow, corroborated by praise from Karpathy and early adopters.

• 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 clear convergence is happening around standardizing agent orchestration, with OpenAI's new SDK and LangChain's protocol integration efforts defining the emerging de-facto stack.

Major players like OpenAI, LangChain, Vercel, and Replit are releasing SDKs and infrastructure for deploying and orchestrating multi-worker agents.

• 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 the agent conversation dominates, foundation model players like Mistral AI continue to invest in core data moats, with this OCR dataset release enabling better vision and document understanding capabilities.

Mistral AI released a large-scale web OCR dataset, providing a foundational resource for training multimodal models.

• Mistral AI@MistralAIrising

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

The limitations of both massive context windows (@reach_vb) and simple vector retrieval are pushing tools like LlamaIndex and new concepts from @mem0ai towards more structured memory architectures for agents.

Discussion moves from basic RAG to 'context engineering', with practitioners exploring complex caching, memory hierarchies, and knowledge graph retrieval.

• 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.

Workflow automation is a key theme, with SaaS tools like Notion and Linear building agent-like capabilities, while orchestration frameworks like Temporal position themselves as the underlying engine for such systems.

Workspace automation tools like Notion and Linear are adding AI-driven features, while Temporal highlights its relevance for orchestrating agents.

• 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 focus is shifting from one-off prompt 'tricks' (@dotey) to systematic, data-driven optimization of system prompts, as demonstrated by Weights & Biases' large-scale benchmark study.

Engineers are sharing best practices for system prompt optimization, using large-scale benchmarking to find non-obvious performance gains.

• 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 synthetic data curation highlights a critical challenge in training capable agents: ensuring data quality and avoiding 'poisoning' from flawed synthetic examples.

The focus in agent training data is on sophisticated filtering of synthetic data to avoid performance degradation.

• Jerry Liu@jerryjliu0repeated

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