Denoise · Twitter

The release of Claude Code 1.5 establishes a direct competitor to OpenAI's legacy in code generation, with initial benchmarks by @swyx indicating performance divergence on complex, long-context tasks.

Anthropic released Claude Code 1.5, a terminal-native agent positioned as a new primary coding interface, sparking discussion on the shift away from traditional IDEs.

• Anthropic@AnthropicAIrising

  Responsible disclosure on a Claude jailbreak chain we patched last week. Full write-up including our red team timeline.

• 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.

• @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 primitives. OpenAI, Vercel, Replit, and Temporal are all independently building towards a common abstraction of durable, replayable, multi-worker agents.

Major infrastructure providers including OpenAI, Vercel, and Replit all released new tools 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.

• Vercel@vercelrising

  Edge runtime for agent workers is live. Spawn durable background agents from any serverless deployment.

• Replit@replitrising

  New agent deployment harness. One command to go from local orchestration to hosted agent worker.

• Temporal@temporaliorepeated

  Orchestrating agents with durable workflows: replayable, resumable, and multi-worker by default. Walkthrough from our infra team.

• Jerry Liu@jerryjliu0repeated

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

There's a shift from artisanal prompt crafting (@dotey) to systematic, data-driven optimization, with tools like DSPy and platforms like Weights & Biases enabling large-scale, automated benchmarking and search.

The field is moving towards industrial-scale practices, highlighted by MistralAI's 100M-row dataset release and new frameworks for automated prompt optimization from DSPy.

• Mistral AI@MistralAIrising

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

• DSPy@dspy_airising

  DSPy 3.0: prompt optimization via compile-time search over system prompt variations. Benchmarks inside.

• Notion@NotionHQrising

  Notion workspace automation is out of beta. Auto-fill tables, chained updates across databases, and a new audit log surface.

• 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.

The vocabulary is changing: @GregKamradt declares 'RAG is dead', while @mem0ai proposes layered memory models. This reflects a consensus that simple vector retrieval is insufficient for complex agentic workflows.

Discourse is evolving from simple RAG to more sophisticated 'context engineering', addressing failure modes in large context windows and proposing layered 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.

• LangChain@LangChainAIrising

  MCP protocol integration thread. How to wire existing LangGraph agents into the Anthropic Model Context Protocol server spec.

• 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.

• Jason Liu@jxnlcorepeated

  Vector db beauty contest. Ran 50k RAG queries against 6 vendors. Results inside, free.

The critical vulnerability is migrating from the model to the orchestration layer. Both Google DeepMind and @AlexAlbert__ now point to tool interaction and deployment logic as the primary areas for agent jailbreaks.

Security research is shifting focus to the unique vulnerabilities of autonomous agents, with Google DeepMind releasing a red teaming framework for this new attack surface.

• Google DeepMind@GoogleDeepMindrising

  New red team framework for prompt injection in autonomous agents. Covers cross-tool leakage, scanner evasion, and sandbox escape patterns.

• 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.

• 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.

While headlines focus on general agents, the most immediate user value comes from specialized, embedded AI. Linear's auto-triage is a prime example of a non-obvious, workflow-integrated feature that quietly delivers significant productivity gains.

AI is being integrated into established SaaS tools for narrow, high-leverage automation, such as Linear's new auto-triage feature for issues.

• Linear@linearrising

  Linear now auto-triages incoming issues. Quiet launch, but already our favorite workspace feature of the year.

• James Clear@jamesclearrepeated

  The best habit tracker is the one you actually open. Three open-source alternatives worth trying.

• Ben Andrew@BenAAndrewrepeated

  The sourdough app I shipped in 48 hours is now my most-used side project. Source and write-up linked.