Denoise · Twitter

The security discourse is moving up the stack from the model to the agent's orchestration layer, where tool-chaining and sandbox escapes present new attack surfaces.

Major AI labs and security researchers are focusing on red-teaming autonomous agents, publishing frameworks and disclosures for vulnerabilities beyond simple prompt injection.

• 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 competition in AI coding assistants is shifting from IDE integrations like Copilot to stateful, terminal-based agents like Claude Code, which fundamentally alters the developer workflow.

Anthropic's Claude Code 1.5 launch dominates, presenting a terminal-native agent that is already seeing developer adoption and performance comparisons against established tools.

• 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.

There is a clear convergence on building a standardized infrastructure layer for agents, focusing on primitives for tool-calling, orchestration, and durable execution across different platforms.

Major infrastructure providers including OpenAI, Vercel, and Replit are releasing SDKs and runtimes specifically for deploying and orchestrating AI agents.

• 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 a valuable data contribution, the primary focus of the day remains squarely on agent-based systems, leaving multimodal developments in the background.

MistralAI released a 100M-row web OCR dataset, providing a significant new resource for training multimodal models.

• Mistral AI@MistralAIrising

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

With 10M context windows becoming available, the challenge shifts from data retrieval to context management, where actors like @GregKamradt and @mem0ai are proposing more sophisticated memory hierarchies.

Discussion moves beyond simple RAG, focusing on 'context engineering' to manage massive context windows and architecting agent memory systems.

• 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 concept of 'orchestration' is a common thread, linking mainstream SaaS automation (Notion, Linear) with the more technical problem of running multi-step AI agents (Temporal).

Workspace automation sees new releases from Notion and Linear, while Temporal highlights its durable workflow engine as a fit for orchestrating agents.

• 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.

Prompt engineering is maturing into a data-driven optimization discipline, with platforms like Weights & Biases providing the infrastructure for scaled experimentation.

The focus is on moving from anecdotal prompt tricks to systematic, large-scale benchmarking of system prompts to find optimal configurations.

• 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 relies more on synthetic data, the subtle but critical challenge of data quality and filtering becomes a key bottleneck for model performance.

The primary concern highlighted is the difficulty of curating synthetic data for training agents without introducing biases that harm generalization.

• Jerry Liu@jerryjliu0repeated

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