Denoise · Twitter

The conversation is shifting from theoretical risks to practical, published frameworks for agent pentesting, with Anthropic and Google DeepMind leading the charge.

Major AI labs are publicly disclosing red-teaming frameworks and patched vulnerabilities for autonomous agents.

• 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 primary developer interface is now a battleground between IDE copilots (Copilot, Cursor) and terminal agents (Claude Code), as highlighted by @karpathy and @swyx.

Anthropic's release of Claude Code 1.5, a terminal-native agent, dominates the conversation, with benchmarks and user experiences validating a workflow shift.

• 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 clear convergence is happening on the agent deployment stack, with OpenAI providing the SDK and infra providers like Vercel and Replit offering the runtime environment.

OpenAI, Vercel, and Replit released new primitives for agent orchestration and deployment, focusing on tool-calling protocols and background workers.

• 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 stack discussion is loud, foundation model providers like Mistral continue to release core artifacts like datasets, which will power future models.

Mistral AI released a large-scale, cleaned web OCR dataset for public use.

• Mistral AI@MistralAIrising

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

Practitioners like @GregKamradt and @reach_vb are finding that large context windows introduce new failure modes, pushing companies like @mem0ai and @llamaindex to build more complex memory solutions.

The discussion moves beyond simple RAG to advanced 'context engineering,' exploring memory caching, retrieval strategies, and specialized memory layers.

• 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 pattern of durable, background automation is appearing in both productivity tools (Notion, Linear) and hardcore engineering infrastructure (Temporal), suggesting a broader trend.

Workspace automation features are being released by Notion and Linear, while Temporal highlights its use case for durable agent orchestration.

• 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 focus is shifting from simple prompt 'hacks' to systematic benchmarking and analysis of production prompts, as shown by @weights_biases' large-scale study.

Developers are sharing practical prompt engineering techniques and large-scale benchmark results for 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 capabilities grow, the bottleneck is becoming data quality, with practitioners like @jerryjliu0 highlighting the challenge of filtering out 'poisonous' synthetic data.

The focus in agent training data is on curating high-quality synthetic data to avoid performance degradation.

• Jerry Liu@jerryjliu0repeated

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