Denoise · Twitter

The security frontier is shifting from prompt injection in chatbots to orchestration-level exploits in multi-step agents, a pattern visible in releases from @GoogleDeepMind and tests by @MalwareTechBlog.

Red teaming efforts are now squarely focused on autonomous agents, with major labs releasing frameworks and researchers testing vulnerabilities in orchestration flows.

• 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 towards terminal-first agentic workflows is underway, with @AnthropicAI's product launch and @karpathy's commentary suggesting the GUI IDE's dominance is being challenged.

Anthropic's launch of Claude Code, a terminal-native agent, has catalyzed discussion around a major developer workflow 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 de facto agent protocol stack is emerging, with @OpenAI defining application-level primitives while platforms like @vercel and @replit provide the compatible hosting runtimes.

The ecosystem is rapidly building dedicated infrastructure for agent orchestration, with OpenAI, Vercel, and Replit all shipping new primitives for deployment and management.

• 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.

Today's activity is quiet, but the @MistralAI dataset release is a strategic move to build foundational capabilities in document intelligence, a key skill for future agents.

Mistral AI released a massive, open web OCR dataset, signaling investment in foundational data for future multimodal models.

• Mistral AI@MistralAIrising

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

Naive RAG's limits are becoming apparent, forcing a convergence on structured approaches like @GregKamradt's frameworks and @mem0ai's layered memory models for agents.

The conversation is evolving from simple RAG to 'context engineering,' focusing on large context window failure modes, caching, and complex memory architectures.

• 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.

Agentic automation is being integrated into mainstream SaaS, with @NotionHQ and @linear demonstrating how autonomous workflows can be productized without using 'AI agent' branding.

Workspace automation tools from Notion and Linear are shipping agent-like features, such as auto-filling tables and triaging issues, under the banner of productivity.

• 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 of prompt engineering is professionalizing, moving from anecdotal tricks (@dotey) to data-driven science, with platforms like @weights_biases enabling large-scale analysis.

Prompt optimization is shifting towards systematic, large-scale benchmarking to identify effective system prompt variations across different models.

• 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 training matures, @jerryjliu0's analysis shows the bottleneck is shifting from raw compute to the nuanced problem of data quality and synthetic data poisoning.

The focus in agent training infrastructure is on sophisticated dataset curation, particularly filtering synthetic data that appears useful but harms model generalization.

• Jerry Liu@jerryjliu0repeated

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