Denoise · Twitter

The security discourse has matured from simple prompt injection to complex agent orchestration vulnerabilities, with Anthropic and Google DeepMind leading the public research.

Major labs are publicly disclosing red-teaming frameworks and patched vulnerabilities for autonomous agents, focusing on multi-step attack surfaces.

• 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 consensus is forming around the terminal as the new UI for coding agents, with Anthropic's launch directly competing with the established Copilot/IDE paradigm.

Anthropic's release of Claude Code 1.5, a terminal-native agent, is driving a broader conversation about moving development workflows out of the IDE.

• 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 new infrastructure layer for agents is solidifying. OpenAI's SDK and Anthropic's MCP (via LangChain) are setting protocol standards that compute platforms like Vercel are racing to support.

OpenAI, Vercel, and Replit released new infrastructure for deploying and orchestrating AI agents, focusing on SDKs, protocols, and edge runtimes.

• 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 agents dominate the conversation, MistralAI's dataset release underscores the continued importance of foundational data collection for advancing open-source model capabilities.

MistralAI released a massive, cleaned, and licensed web OCR dataset, providing a significant new resource for training open-source 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 thought leaders like GregKamradt and tools like mem0ai to propose new primitives for agent memory, differentiating working memory from long-term storage.

Discussion is moving beyond simple RAG towards 'context engineering' and more complex memory architectures for agents with 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.

Autonomous workflow automation is becoming a standard feature in modern collaboration software, with Notion and Linear embedding capabilities previously requiring third-party tools.

SaaS tools like Notion and Linear are shipping agent-like automation features, such as auto-triaging issues and chaining database updates, directly into their platforms.

• 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 practice of prompt engineering is maturing from an art to a science, with organizations like Weights & Biases investing in large-scale studies to find empirically optimal system prompts.

Engineers are sharing increasingly systematic approaches to prompt optimization, from practical heuristics to large-scale benchmark-driven analysis.

• 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 training becomes more widespread, data quality, not just quantity, is becoming the bottleneck, with a focus on avoiding subtle 'poisoning' from low-quality synthetic data.

A key challenge highlighted is the curation of synthetic data for agent training, specifically filtering out data that harms model generalization.

• Jerry Liu@jerryjliu0repeated

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