Denoise · Twitter

A convergence is happening between model providers like Anthropic and Google DeepMind and independent researchers like @MalwareTechBlog to define the security surface of agents.

Major model providers and security researchers are focusing on red-teaming autonomous agents, establishing new frameworks for prompt injection and vulnerability testing.

• 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 has shifted from IDE plugins to standalone terminal agents, with Claude Code 1.5 now in direct competition with prior approaches, as validated by @karpathy and benchmarked by @swyx.

Anthropic's release of Claude Code 1.5, a terminal-native agent, solidifies a new developer workflow paradigm, prompting immediate benchmarks and developer adoption.

• 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.

The battleground is moving up the stack from models to orchestration protocols, with OpenAI's SDK and LangChain's integration work showing a push toward standardization.

OpenAI, Vercel, and Replit released new SDKs and platforms for agent orchestration and deployment, indicating a race to build the definitive infrastructure layer.

• 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 data release underscores the ongoing, less visible effort to build the foundational data moats for next-generation models.

MistralAI contributed a large-scale, cleaned web OCR dataset to the open-source community, providing foundational material for training future multimodal models.

• Mistral AI@MistralAIrising

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

Thought leaders like @GregKamradt are abandoning the term RAG, while tools like @mem0ai and @llamaindex are building the components for a new memory stack beyond simple vector search.

The discussion around RAG is maturing into 'context engineering,' with a focus on sophisticated memory architectures, caching strategies, and graph-based 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.

The agentic workflow pattern is being integrated into mainstream productivity software, with Notion and Linear building proprietary, vertically-integrated automation engines.

Established SaaS tools like Notion and Linear are releasing powerful workspace automation features, reflecting the broader trend of agent-like autonomous operations.

• 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.

Organizations like Weights & Biases are industrializing prompt optimization, treating it as a formal search problem rather than a creative art.

The practice of 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.

As agent abilities improve, practitioners like @jerryjliu0 highlight that the new bottleneck is high-quality, non-poisonous training data, a harder problem than just scaling compute.

Attention is being paid to the curation of synthetic datasets for agent training, specifically on how to filter 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.