Denoise · Twitter

Consensus is forming that the most critical agent vulnerabilities are in the orchestration layer and cross-tool communication, not just simple prompt injection.

Major AI labs are focusing on red-teaming autonomous agents, with Anthropic disclosing a patched jailbreak and Google DeepMind releasing a new framework.

• 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.

A direct competition is emerging between Anthropic's terminal-first agent and OpenAI's ecosystem, with DSPy's update suggesting a deeper, programmatic approach to optimizing agent reasoning.

Anthropic's release of Claude Code 1.5, a terminal-native agent, triggered extensive discussion and early benchmarks against existing toolchains.

• 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 industry is converging on agent orchestration primitives, with OpenAI's SDK and LangChain's protocol integrations demonstrating a push towards standardization.

OpenAI, Vercel, and Replit all shipped infrastructure for deploying and orchestrating agents, solidifying the agent-as-a-platform model.

• 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 development dominates the conversation, foundational dataset releases like Mistral's are a critical, quieter enabler for the next generation of model capabilities.

Mistral AI released a massive, cleaned web OCR dataset, providing a foundational asset for training new 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 argue vector search is insufficient, pushing for memory systems that distinguish between working and long-term storage.

Discussions are shifting from basic RAG to more complex 'context engineering', exploring advanced memory architectures for agents.

• 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 is 'workspace as agent,' where tools like Notion, Linear, and Temporal are abstracting orchestration to perform autonomous tasks for users.

Workspace automation advanced with Notion and Linear launching features for 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.

The field is maturing from individual craft (dotey) to industrial-scale optimization, with Weights & Biases running 40k variants to map the efficient frontier.

The focus is shifting from anecdotal prompt tricks to systematic, large-scale benchmarking of system prompts to optimize model performance.

• 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 from LlamaIndex highlights a critical bottleneck: synthetic data that appears high-quality can poison model generalization, requiring sophisticated filtering.

The primary discussion point was the challenge of curating high-quality synthetic data for training robust agents.

• Jerry Liu@jerryjliu0repeated

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