Denoise · Twitter

The focus of agent security is shifting from generic prompt injection to vulnerabilities in the orchestration layer, a pattern noted by Anthropic, Google DeepMind, and @AlexAlbert__.

Frontier model labs are now publicly disclosing agent jailbreaks and red-teaming frameworks, while independent researchers test autonomous pentesting 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 primary interface for AI coding assistance is moving from the IDE to the terminal, a shift predicted by @karpathy and validated by adoption from users like @levelsio.

Anthropic's release of Claude Code 1.5, a terminal-native agent, spurred immediate benchmarks and discussions of a major workflow shift.

• 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 consensus architecture for agent deployment is forming around durable, orchestrated workers, with OpenAI's SDK and Vercel's Edge runtime providing key components.

OpenAI, Vercel, and Replit all released new infrastructure for deploying and orchestrating agents, signaling a race to build the standard agent stack.

• 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.

A quiet day for this category, with the only signal being MistralAI's investment in foundational datasets, suggesting a continued focus on data as the bottleneck for multimodal progress.

MistralAI released a 100M-row web OCR dataset, a foundational asset for training next-generation multimodal models.

• Mistral AI@MistralAIrising

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

The limits of vector search are pushing developers like @GregKamradt and companies like @mem0ai to design more structured memory systems, differentiating between working and long-term storage.

Discussions are elevating from simple RAG to 'context engineering,' with new frameworks for managing memory and cache in large context windows.

• 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.

A convergence is visible between SaaS workflow automation (Notion, Linear) and durable execution infrastructure (Temporal), both aiming to solve similar orchestration problems now central to AI agents.

Workspace automation is accelerating in SaaS tools like Notion and Linear, mirroring the orchestration patterns seen in dedicated agent frameworks from Temporal.

• 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 anecdotal 'prompt tricks' to a data-driven science, with platforms like Weights & Biases providing tools to systematically find the efficient frontier of system prompts.

Prompt engineering is becoming more systematic, with large-scale benchmark studies from Weights & Biases complementing practitioner-shared heuristics.

• 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.

A quiet day, with the only tweet from @jerryjliu0 pointing to a subtle but critical infrastructure problem: the lack of robust tools for curating the synthetic data used in agent fine-tuning.

The main infrastructure challenge discussed for agent training is data curation, specifically how to filter synthetic datasets to avoid poisoning model generalization.

• Jerry Liu@jerryjliu0repeated

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