Denoise · Twitter

The security focus is shifting from model jailbreaks (@AnthropicAI) to systemic vulnerabilities in agent orchestration (@GoogleDeepMind, @MalwareTechBlog).

Major labs are publishing formal frameworks and disclosures for agent security, focusing on red-teaming and responsible vulnerability patching.

• 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 convergence is happening around terminal-native agents (@AnthropicAI, @karpathy, @levelsio), challenging the dominance of IDE-integrated tools like Copilot.

The new frontier for coding tools is the terminal, with Anthropic's Claude Code 1.5 release and supporting commentary driving the conversation.

• 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 standard agent stack is emerging, with @OpenAI defining low-level primitives and platforms like @vercel and @replit providing the deployment solutions.

Foundational infrastructure for deploying and orchestrating agents is being released by major players like OpenAI, Vercel, and Replit.

• 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 focus is heavily on language agents today, @MistralAI's data release indicates continued foundational investment in multimodal capabilities.

Mistral AI released a large-scale, open OCR dataset to support multimodal model training.

• Mistral AI@MistralAIrising

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

Experts like @GregKamradt and tools like @mem0ai are moving past basic vector search, focusing on sophisticated memory layers and caching strategies.

The discussion around context management is evolving from simple RAG to more complex "context engineering" frameworks.

• 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 principles of agentic automation from @NotionHQ and @linear are appearing in productivity software, suggesting a broader mainstream adoption of these concepts.

Workspace automation tools like Notion and Linear are quietly launching agent-like features for auto-triaging and data entry.

• 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.

A data-driven approach is consolidating, with @weights_biases providing large-scale benchmarks that refute common assumptions about prompt effectiveness.

The focus in prompting is shifting from manual trick-shot discovery to scalable, automated benchmarking of 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 noted by @jerryjliu0, filtering and curating training data is becoming a critical and difficult step in building robust agentic models.

The challenge of curating high-quality synthetic data for training agents is a key focus for practitioners.

• Jerry Liu@jerryjliu0repeated

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