Denoise · Twitter

The security conversation is shifting from model vulnerabilities to the orchestration layer, where agents interact with tools and external systems.

Major labs are focusing on security for autonomous agents, publishing responsible disclosures and red-teaming frameworks for this new attack surface.

• 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 to the terminal, with Anthropic's Claude Code directly challenging the Cursor and Copilot model.

Anthropic's release of Claude Code 1.5, a terminal-native agent, dominates the conversation, with benchmarks and early adopters validating a workflow shift away from traditional IDEs.

• 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 convergence is happening around agent deployment, with LLM providers (OpenAI) and cloud platforms (Vercel) building out the necessary infrastructure in parallel.

Infrastructure for deploying and managing agents is being released by major platforms, including an SDK from OpenAI and runtime environments from Vercel and Replit.

• 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.

MistralAI continues to strengthen the open-source ecosystem by releasing foundational datasets that enable direct competition with closed, proprietary models.

MistralAI released a large-scale, cleaned web OCR dataset, providing a key open-source resource for training multimodal models.

• Mistral AI@MistralAIrising

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

As context windows grow, platforms like LlamaIndex and Mem0ai are pioneering more sophisticated memory solutions beyond simple vector retrieval for agents.

The discourse moves from simple RAG to 'context engineering', exploring complex memory architectures, caching failure modes, and frameworks for managing context.

• 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.

The pattern of embedding small, specialized agents into existing SaaS products is accelerating, with Notion and Linear automating routine knowledge work.

Workspace automation tools like Notion and Linear are releasing agent-like features for task management, such as auto-filling tables and triaging issues.

• 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.

Platforms like Weights & Biases are industrializing prompt engineering, treating system prompts as a critical hyperparameter to be optimized through extensive testing.

Prompt optimization is scaling up from individual tricks to large-scale, systematic benchmarking across multiple models to find the most effective system prompts.

• 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.

Data quality, particularly for synthetic data used in agent training, is emerging as a critical and subtle MLOps problem that can poison model performance.

A key challenge in training agents is the curation of high-quality datasets, specifically filtering synthetic data to avoid negative impacts on model generalization.

• Jerry Liu@jerryjliu0repeated

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