Denoise · Twitter

A consensus is forming that agent security is a critical and distinct discipline from model safety, with a focus on the orchestration layer.

Frontier labs like Anthropic and Google DeepMind are releasing formal frameworks and disclosures for red-teaming autonomous agents.

• 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 battleground for developer AI is shifting from IDE plugins like Copilot to standalone terminal agents, with long-context performance as a key differentiator.

Anthropic launched Claude Code 1.5, a terminal-native agent, prompting benchmarks and discussions about a major workflow shift away from 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.

The agent infrastructure layer is rapidly standardizing, with OpenAI's SDK and platforms like Vercel providing primitives for managed, durable agents.

OpenAI, Vercel, and Replit released new infrastructure for deploying and orchestrating agents, while LangChain showed protocol interoperability.

• 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 stack gets attention, major players like MistralAI continue to invest in foundational data moats for multimodal capabilities.

MistralAI released a large, 100M-row, cleaned web OCR dataset to facilitate the training of multimodal models.

• Mistral AI@MistralAIrising

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

The availability of massive context windows is forcing a re-evaluation of retrieval, with actors like @GregKamradt proposing new architectures beyond simple vector search.

Discussions focus on the failure modes of 10M token context windows and the evolution of RAG into more complex "context engineering."

• 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" concept is being productized as workspace automation in SaaS, with Notion and Linear focusing on internal tasks and Temporal targeting developers.

Notion and Linear shipped AI-driven workspace automation features, while Temporal positioned 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.

Prompting is maturing into a data-driven optimization problem, with platforms like Weights & Biases providing tooling to find the efficient frontier at scale.

The focus is on systematic, large-scale benchmarking of system prompts to move beyond anecdotal "prompt engineering" tricks.

• 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 capabilities grow, the bottleneck is shifting back to high-quality, poison-resistant training data, making synthetic data curation a critical research area.

The key topic is the need for sophisticated data filtering to ensure the quality of synthetic data used for training agents.

• Jerry Liu@jerryjliu0repeated

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