Denoise · Twitter

The focus in agent security is shifting from model-level jailbreaks (Anthropic) to systemic vulnerabilities in orchestration and tool use (DeepMind, MalwareTechBlog).

Major AI labs like Anthropic and DeepMind are proactively releasing details on agent vulnerabilities and red-teaming frameworks, while practitioners test autonomous pentesting.

• 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 primary competition is shifting from integrated IDE plugins like Copilot to standalone terminal agents like Claude Code, with the latter promising deeper workflow integration.

Anthropic's release of Claude Code 1.5, a terminal-native agent, is driving a narrative about the end of IDE-centric coding, with benchmarks and early adoption reports emerging.

• 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 de facto standard agent stack is emerging, combining protocol-level tool calling (OpenAI) with managed, serverless runtimes (Vercel, Replit) to abstract away orchestration complexity.

OpenAI, Vercel, and Replit released new infrastructure for agent orchestration and deployment, focusing on SDKs, durable workers, and edge runtimes.

• 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 model releases are frequent, the release of high-quality, licensed training data by a major lab like Mistral AI is a less common but foundational contribution.

Mistral AI contributed to the open ecosystem by releasing a large-scale, 100M-row dataset for web OCR training.

• Mistral AI@MistralAIrising

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

The consensus is that vector search is an insufficient primitive for agent memory; frameworks from @GregKamradt and @mem0ai are converging on more complex, stateful memory architectures.

The conversation is evolving from simple RAG to "context engineering," with practitioners exploring large-context failure modes and multi-layered agent memory systems.

• 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 automation patterns in SaaS tools like Notion and Linear are converging with the agent orchestration primitives being built with tools like Temporal, indicating a broader trend of workflow automation.

Workspace automation is a key theme, with Notion and Linear releasing features for auto-filling and auto-triaging, while Temporal positions its workflow engine for agent orchestration.

• 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.

Prompt engineering is shifting from anecdotal "tricks" (@dotey) to a data-driven science, using platforms like Weights & Biases for rigorous, scaled-out testing.

The focus is on systematic, large-scale benchmarking of system prompts to find an "efficient frontier" of performance across different models.

• 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.

As agent training relies more on synthetic data, sophisticated filtering and curation techniques are becoming critical for avoiding model degradation.

A specific technique was shared by @jerryjliu0 for curating agent training datasets by filtering out synthetic data that harms generalization.

• Jerry Liu@jerryjliu0repeated

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