Denoise · Twitter

The focus of agent security is shifting from simple prompt injections to vulnerabilities in the orchestration layer, a concern voiced by both labs like @AnthropicAI and practitioners like @MalwareTechBlog.

Frontier model labs are now openly discussing agent security, publishing red-teaming frameworks and responsible disclosures for complex jailbreaks.

• 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 new front in the AI coding assistant war is opening in the terminal, with @AnthropicAI's Claude Code being benchmarked by voices like @swyx against established players.

The launch of Anthropic's terminal-native coding agent sparked intense discussion about a fundamental shift in developer workflows, away from traditional IDEs.

• 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 protocols is underway, with @OpenAI defining primitives and frameworks like @LangChainAI providing compatibility layers for this emerging standard.

Major players like OpenAI released foundational SDKs for agent orchestration while platforms like Vercel and Replit shipped corresponding deployment solutions.

• 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.

@MistralAI continues its strategy of strengthening the open-source ecosystem by providing foundational artifacts, like this 100M-row OCR dataset, for community model training.

The sole signal in this category was a large-scale open dataset release for web OCR from Mistral AI.

• Mistral AI@MistralAIrising

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

The community is fragmenting the monolithic concept of 'memory,' with @GregKamradt advocating for context engineering and tools like @mem0ai building specialized memory layers beyond simple vector search.

The conversation evolved from simple RAG to 'context engineering,' focusing on sophisticated caching, retrieval strategies, and multi-layered agent memory.

• 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.

Agent-like automation is being integrated into mainstream SaaS. @NotionHQ and @linear are abstracting AI capabilities into specific product features, hiding the complexity of orchestration shown by tools like @temporalio.

Workspace automation tools from Notion and Linear are shipping agent-like features that auto-triage issues and automate data 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 becoming industrialized. While @dotey shares practical tips, @weights_biases is treating it as a formal search problem, benchmarking thousands of variants to find an efficient frontier.

The focus is shifting from sharing individual prompt 'tricks' to systematic, large-scale benchmarking of system prompts to find optimal 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.

@jerryjliu0 highlights a critical bottleneck in the agent development pipeline: the difficulty of filtering synthetic data that appears valid but ultimately harms model generalization.

A key discussion point was the challenge of curating high-quality synthetic data specifically for training agents to avoid performance degradation.

• Jerry Liu@jerryjliu0repeated

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