Denoise · Twitter

The security conversation is rapidly shifting from chatbot prompt injection to orchestration-level exploits, with @AnthropicAI and @GoogleDeepMind setting the terms of the new field.

Major labs are publicly disclosing agent vulnerabilities and releasing red-teaming frameworks for this new class of threats.

• 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 battle for developer mindshare is moving from the IDE to the terminal, as @AnthropicAI's agent-centric approach directly challenges the @OpenAI/Copilot model.

Anthropic's launch of Claude Code 1.5, a terminal-native agent, has triggered a wave of adoption and competitive benchmarking.

• 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 convergence on agent orchestration primitives is underway, with @OpenAI pushing a new SDK and @LangChainAI providing interoperability for @AnthropicAI's MCP protocol.

OpenAI and Anthropic are releasing competing SDKs for agent orchestration, while platforms like Vercel and Replit ship supporting deployment infrastructure.

• 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 major labs focus on agents, @MistralAI's dataset release provides foundational infrastructure for the open-source community to advance multimodal capabilities.

Mistral AI released a large, open OCR dataset for training multimodal models, a significant contribution in an otherwise quiet category.

• Mistral AI@MistralAIrising

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

Practitioners like @GregKamradt are declaring RAG insufficient, while tools like @mem0ai and @llamaindex are building beyond simple vector search, suggesting the problem space has evolved.

The discourse is shifting from basic RAG to 'context engineering', exploring failure modes in massive context windows and more complex memory architectures.

• 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 orchestration patterns seen in @NotionHQ and @linear's new features suggest a convergence between mainstream SaaS automation and the agent-based workflows emerging from the AI space.

Workspace automation features are being released by productivity tools like Notion and Linear, mirroring agentic workflow patterns.

• 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 work by @weights_biases to benchmark 40k prompt variants signals the industrialization of prompt engineering, treating it as a formal optimization problem.

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

• 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.

@jerryjliu0's thread highlights a critical bottleneck in agent development: ensuring the quality of synthetic training data to avoid model degradation.

The conversation this cycle focused on the nuances of dataset curation for training agents, specifically on filtering synthetic data.

• Jerry Liu@jerryjliu0repeated

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