Denoise · Twitter

The field is rapidly moving from theoretical agent security to applied practice, with both Anthropic and Google DeepMind establishing public norms for vulnerability disclosure and testing.

Major labs are publicly disclosing agent jailbreaks and releasing red-teaming frameworks, while independent researchers are beginning to pentest agents in the wild.

• 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 battleground for coding assistants is shifting from IDE extensions (Copilot, Cursor) to standalone terminal agents (Claude Code), a workflow shift predicted by karpathy and benchmarked by swyx.

Anthropic's release of Claude Code 1.5, a terminal-native agent, marks a significant new direction for AI coding assistants, drawing immediate benchmarks and adoption.

• 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 stack for agent orchestration is emerging, with major infrastructure providers like OpenAI and Vercel converging on similar primitives for deploying and managing agent workers.

A wave of new infrastructure for agents has been released, including SDKs from OpenAI, edge runtimes from Vercel, and deployment tools from 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.

Even as model capabilities advance, the creation of large-scale, high-quality, open datasets by major players like Mistral AI remains a key driver for progress in the field.

Mistral AI released a massive, cleaned 100M-row web OCR dataset, providing a foundational resource for training next-generation multimodal models.

• Mistral AI@MistralAIrising

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

There's a growing consensus that simple vector search is insufficient for agent memory, leading to more complex solutions from GregKamradt, mem0ai, and LlamaIndex that treat memory as a structured system.

The discourse is evolving from 'RAG' to 'context engineering', with new frameworks for managing large context windows and multi-layered 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.

Durable execution frameworks like Temporal are being positioned as a robust way to orchestrate complex, stateful AI agents, bridging the gap between workflow automation and agent infrastructure.

Workspace automation is a recurring theme, with Notion and Linear launching features for auto-filling data and triaging issues, reflecting a broader trend in productivity software.

• 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 approach to prompt optimization is bifurcating: practitioners like dotey share curated heuristics, while platforms like Weights & Biases provide industrial-scale data to find the efficient frontier.

Prompt engineering is becoming more systematic, moving from anecdotal tricks to large-scale, empirical benchmarking of system prompts across multiple 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 architectures mature, Jerry Liu's work highlights that the critical bottleneck is shifting back to data quality, specifically filtering out 'poisonous' synthetic data that harms model performance.

The focus in agent training is on sophisticated data curation techniques to prevent generalization failure caused by seemingly high-quality synthetic data.

• Jerry Liu@jerryjliu0repeated

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