Denoise · Twitter

There's a clear parallel track: major labs are building systematic security frameworks for agents, while security researchers like @MalwareTechBlog are already deploying them for real-world tasks.

Providers like Anthropic and Google DeepMind are formalizing agent red-teaming frameworks, while practitioners are testing autonomous pentesting 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.

Claude Code 1.5 is positioned as a direct competitor to the Copilot/Cursor stack, with @karpathy and @swyx framing the battleground as a fundamental change in developer experience.

Anthropic's release of Claude Code 1.5, a terminal-native agent, spurred discussion on its performance and the broader shift away from IDE-centric 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.

A clear convergence is happening around agent deployment. OpenAI's SDK, Vercel's edge runtime, and Replit's harness all address the same problem, indicating a race to build the 'Heroku for Agents'.

Major infrastructure players including OpenAI, Vercel, and Replit released new primitives for deploying, running, and orchestrating 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.

This category is quiet except for MistralAI's foundational data contribution, a move that enables others to build multimodal models but isn't a product release itself.

MistralAI released a large-scale, cleaned web OCR dataset for public use in training new models.

• Mistral AI@MistralAIrising

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

A consensus is forming among practitioners like @GregKamradt, @reach_vb, and frameworks like mem0.ai that vector search alone is insufficient, pushing towards structured memory and sophisticated caching.

Discussion moves beyond simple RAG, focusing on 'context engineering,' advanced memory architectures, and failure modes in massive 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.

The pattern of autonomous, background task execution is appearing in both dedicated AI infra (Temporal for agents) and SaaS tools (Notion, Linear), showing a broad convergence on automation.

Workspace automation tools like Notion and Linear are releasing agent-like features for auto-filling and auto-triaging, while Temporal positions its workflow engine for agent orchestration.

• 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 field is moving from folklore to science, with firms like @weights_biases providing empirical data to validate or disprove the anecdotal prompt 'tricks' shared by practitioners like @dotey.

Prompt engineering is being systematized through large-scale benchmarking by Weights & Biases and practitioner-shared heuristics from production reviews.

• 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 model architectures and training methods mature, @jerryjliu0 highlights that the key differentiator is shifting towards the subtle, difficult work of high-quality data curation.

The focus is on the crucial step of dataset curation for training effective agents, specifically on filtering out synthetic data that harms generalization.

• Jerry Liu@jerryjliu0repeated

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