Denoise · Twitter

A consensus is forming that the agent orchestration layer, rather than the model itself, is the primary attack surface, a point echoed by @AnthropicAI, @GoogleDeepMind, and @AlexAlbert__.

Major labs are publicly disclosing agent jailbreaks and releasing red-teaming frameworks, while independent researchers are testing autonomous pentesting flows.

• 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 frontier of AI coding tools is shifting from IDE plugins (Copilot) to stateful terminal agents (Claude Code), a structural change analyzed by @karpathy and benchmarked by @swyx.

Anthropic's Claude Code 1.5 release dominates, proposing a terminal-native agent as the new developer workflow, with initial benchmarks and user feedback.

• 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 around protocol-level standards for agent tool-calling and orchestration, with OpenAI's SDK and LangChain's integration efforts leading the charge.

Major infrastructure providers like OpenAI, Vercel, and Replit are releasing SDKs and runtimes to standardize the deployment and orchestration of AI agents.

• 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 release from Mistral AI is a foundational data artifact, indicating a strategic focus on improving the quality and scale of public datasets for document and web understanding.

Mistral AI released a large, cleaned, and licensed web OCR dataset containing 100 million rows, intended for training multimodal models.

• Mistral AI@MistralAIrising

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

A new vocabulary is emerging to differentiate between working memory and long-term storage for agents, with @mem0ai, @GregKamradt, and LlamaIndex all proposing structures beyond simple vector search.

Discussions are shifting from basic RAG to 'context engineering,' with deep dives into memory architectures, cache invalidation in large contexts, and knowledge graphs.

• 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.

Agentic patterns are becoming commoditized as features in SaaS products like Notion and Linear, showing the horizontal application of this technology beyond specialized developer tools.

Workspace automation tools from Notion and Linear are adopting agent-like behaviors, while orchestration platforms like Temporal are being positioned for agent workflows.

• 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 maturing into a data science discipline, exemplified by Weights & Biases running 40k prompt variants to find an efficient frontier, moving beyond anecdotal advice.

The focus is on empirical and scalable methods for prompt optimization, moving from individual tricks to large-scale benchmark-driven system prompt engineering.

• 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.

The insight from @jerryjliu0 points to a subtle but critical infrastructure problem: as synthetic data generation scales, data quality filtering becomes a key bottleneck for agent performance.

A specific challenge in training effective agents is highlighted: the careful curation of synthetic data to prevent the model from learning spurious correlations.

• Jerry Liu@jerryjliu0repeated

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