Denoise · Twitter

The security conversation for agents is rapidly shifting from simple prompt injection to complex orchestration-layer vulnerabilities, led by @AnthropicAI and @GoogleDeepMind.

Major labs are releasing red teaming frameworks and disclosures for autonomous agents, focusing on emerging attack surfaces.

• 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 new frontier of competition is the terminal, with @AnthropicAI's product release prompting immediate benchmarks from @swyx and adoption signals from @levelsio.

Anthropic's launch of Claude Code 1.5, a terminal-native agent, has triggered intense discussion about a fundamental shift in developer workflows.

• 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.

@OpenAI, @Vercel, and @Replit are converging to provide a 'PaaS for agents,' creating the platforms where new agents will be deployed and run.

A new wave of tooling for agent orchestration and deployment has arrived from major players, solidifying the agent infrastructure stack.

• 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 dataset release highlights the ongoing, foundational work of curating high-quality data to improve core model capabilities.

Mistral AI contributed to the open-source community by releasing a large-scale, cleaned web OCR dataset for model training.

• Mistral AI@MistralAIrising

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

Led by voices like @GregKamradt and @mem0ai, the focus is shifting to advanced architectures like differentiated memory layers, moving beyond simple vector retrieval.

The conversation is evolving from simple RAG to more sophisticated 'context engineering' as practitioners hit the limits of large context windows.

• 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.

While the AI world pursues general autonomy, SaaS platforms like @NotionHQ and @linear are demonstrating a parallel, practical path to agent adoption through narrow automation.

Workspace automation tools like Notion and Linear are shipping agent-like features focused on structured tasks like auto-filling and issue triage.

• 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 contrast between @dotey's tactical tricks and @weights_biases's industrial-scale benchmarking reveals the professionalization of prompt optimization.

Prompt engineering is maturing into a systematic discipline, moving from anecdotal tricks to large-scale, automated benchmarking.

• 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.

@jerryjliu0 highlights a critical bottleneck: filtering synthetic data that looks good but 'poisons' an agent's ability to generalize, a key challenge for scaling.

The focus in agent training infrastructure is on the nuanced problem of curating high-quality datasets to avoid performance degradation.

• Jerry Liu@jerryjliu0repeated

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