Denoise · Twitter

The agent security conversation is rapidly maturing, with formal methodologies from Anthropic and DeepMind emerging alongside real-world pentesting by MalwareTechBlog.

Major AI labs are publishing formal red-teaming frameworks and disclosures for autonomous agents, while practitioners test them on live systems.

• 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 is shifting from IDE-integrated copilots to standalone terminal agents, with Claude Code 1.5 positioned as a direct challenger to existing workflows.

Anthropic's release of Claude Code 1.5, a terminal-native agent, has spurred benchmarks and adoption discussions, framing it as a new developer paradigm.

• 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 between model providers (OpenAI) and infrastructure platforms (Vercel, Replit) to define the primitives for deploying and managing agents.

OpenAI, Vercel, and Replit released new SDKs and runtimes for agent orchestration, signaling a race to build the standard deployment 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.

The focus in multimodality today is on foundational data assets, with major players like MistralAI contributing to the public commons to spur model development.

MistralAI released a large-scale, cleaned web OCR dataset for public use in training multimodal models.

• Mistral AI@MistralAIrising

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

A consensus is emerging, articulated by GregKamradt and mem0ai, that simple vector retrieval is insufficient, necessitating more structured memory systems for agents.

The discussion has shifted from basic RAG to sophisticated 'context engineering,' with new frameworks for managing different types of agent memory.

• 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.

Productivity tools like Notion and Linear are integrating agent-like automation, suggesting a broader trend of embedding autonomous logic into SaaS workflows.

Workspace automation is a key theme, with Notion and Linear both shipping features that auto-manage tasks and data within their platforms.

• 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.

Weights & Biases is formalizing system prompt optimization as a rigorous engineering discipline, moving beyond the craft-based approach shared by dotey.

The focus in prompt engineering is shifting from anecdotal tricks to large-scale, systematic benchmarking of system prompts across many 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.

Jerry Liu highlights that for agent training, data quality and filtering synthetic artifacts are now a more critical bottleneck than raw compute or model architecture.

The conversation focuses on the challenges of data curation for training agents, specifically filtering out harmful synthetic data.

• Jerry Liu@jerryjliu0repeated

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