Denoise · Twitter

The conversation is shifting from model-level safety (e.g., refusals) to system-level security, focusing on how agents interact with tools and external systems.

Major labs like Anthropic and DeepMind are publicly sharing research on agent vulnerabilities and red-teaming frameworks, while practitioners test them in the wild.

• 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 focus is now on the developer experience of self-contained, terminal-based agents (Claude Code) rather than IDE-integrated assistants (Copilot), signaling a potential platform shift.

Anthropic's launch of Claude Code 1.5, a terminal-native agent, dominates the conversation, with commentary from @karpathy and benchmarks from @swyx reinforcing the shift 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 consensus is forming around the need for a standardized orchestration layer for agents, with OpenAI's SDK and LangChain's protocol integrations representing two different approaches to solving the same problem.

OpenAI, LangChain, Vercel, and Replit are all releasing infrastructure and SDKs for deploying and orchestrating multi-step, multi-worker 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.

While the agent conversation focuses on reasoning, foundation model players like Mistral AI continue to commoditize the data layer, enabling others to build specialized perception models.

Mistral AI released a large-scale, 100M-row web OCR dataset for training multimodal models.

• Mistral AI@MistralAIrising

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

A fault line is emerging between simply expanding the context window (@reach_vb) and developing smarter memory systems (@mem0ai, @GregKamradt) that differentiate memory types and retrieval strategies.

With massive context windows becoming available, the discussion is shifting from basic RAG to "context engineering," exploring complex caching, memory hierarchies, and knowledge graph retrieval.

• 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.

SaaS platforms like Notion and Linear are building agent-like automation features directly into their products, competing with general-purpose agents by leveraging deep domain context.

This category shows a trend towards workspace automation, with Notion and Linear launching features for auto-filling and auto-triaging, mirroring the task automation goals of AI agents.

• 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 is maturing from craft to science, with tools like Weights & Biases enabling quantitative analysis of prompt variants, treating system prompts as a key hyperparameter.

The focus in prompt engineering is moving towards systematic, large-scale benchmarking of system prompts rather than relying on anecdotal "tricks."

• 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 focus in agent training data is on quality over quantity, recognizing that poorly curated synthetic data can actively degrade performance, a key challenge for scaling agent capabilities.

@jerryjliu0 discusses the nuances of dataset curation for training agents, specifically how to filter out synthetic data that harms generalization.

• Jerry Liu@jerryjliu0repeated

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