Denoise · Twitter

The conversation has moved beyond simple prompt injection. The new security frontier, defined by Anthropic and DeepMind, is auditing the complex interaction logic of autonomous agents.

Major players like Anthropic and DeepMind are publishing frameworks for red-teaming autonomous agents, focusing on exploits in the orchestration and tool-use layers.

• 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.

Anthropic's Claude Code 1.5 signals a convergence on the terminal as the primary interface for AI-native coding, directly challenging IDE-centric tools like Cursor and Copilot.

Anthropic's release of Claude Code 1.5, a terminal-native agent, has sparked discussions led by figures like Karpathy about a fundamental shift in developer workflows.

• 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 platform race is emerging between OpenAI, Vercel, and Replit to provide the definitive infrastructure for deploying and orchestrating multi-worker, durable agents.

OpenAI, Vercel, and Replit released new SDKs and runtimes for agent deployment, indicating a rapid build-out of the required infrastructure layer.

• 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 infrastructure dominates conversation, foundation model builders like Mistral AI continue to focus on releasing core data assets to unlock new multimodal capabilities.

Mistral AI released a large-scale, 100M-row web OCR dataset, providing a foundational asset for training a new generation of multimodal models.

• Mistral AI@MistralAIrising

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

Practitioners like GregKamradt and startups like mem0ai are arguing that simple vector retrieval is obsolete, pushing towards sophisticated context management as a key differentiator.

The discussion is evolving from simple RAG to "context engineering," with new frameworks for memory hierarchies and analysis of failure modes in large context windows.

• 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 agent pattern is not just for developers; Notion and Linear are converging on a similar strategy of embedding autonomous task completion into their core products.

Workspace tools Notion and Linear are shipping significant automation features, embedding agent-like capabilities for tasks like 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.

As prompts become core application logic, a need for systematic optimization is emerging, addressed by tools like DSPy and benchmarking platforms like Weights & Biases.

Firms like Weights & Biases are conducting large-scale benchmarks of system prompts, aiming to professionalize prompt engineering into a systematic, data-driven discipline.

• 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.

Jerry Liu highlights a critical bottleneck for agent development: the lack of robust methodologies for curating training data, especially when dealing with synthetic datasets.

A key challenge in agent training is identified: the curation of high-quality datasets and filtering of synthetic data that can harm model generalization.

• Jerry Liu@jerryjliu0repeated

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