https://drydock.si/ The shipyard for AI agents: frameworks, orchestration, vector databases, inference hosting, coding agents and dev tools — with live GitHub stars and a launch feed. en Tue, 06 Oct 2026 07:17:07 +0000 https://drydock.si/guides/agent-skills-guide Agent skills explained: how SKILL.md works, where Claude Code, Codex, Cursor, Copilot and Gemini CLI load skills, how they differ from MCP, plus a template. Tutorials Tue, 06 Oct 2026 12:00:00 +0000 https://drydock.si/guides/agent-skills-guide https://drydock.si/guides/claude-code-vs-codex-vs-gemini-cli Claude Code vs Codex vs Gemini CLI compared on pricing, open source, setup, MCP, instruction files and workflow, so you can pick the right terminal agent. Comparisons Mon, 05 Oct 2026 12:00:00 +0000 https://drydock.si/guides/claude-code-vs-codex-vs-gemini-cli https://drydock.si/guides/best-open-source-coding-agents The best open-source coding agents in 2026, compared on license, interface, model choice and upkeep: OpenCode, Cline, Codex CLI, Gemini CLI, goose and Aider. Best of Mon, 05 Oct 2026 12:00:00 +0000 https://drydock.si/guides/best-open-source-coding-agents https://drydock.si/guides/tavily-vs-exa-vs-firecrawl Tavily vs Exa vs Firecrawl for giving AI agents web access: search quality approach, page extraction, crawling, MCP servers, pricing models and code examples. Comparisons Mon, 05 Oct 2026 12:00:00 +0000 https://drydock.si/guides/tavily-vs-exa-vs-firecrawl https://drydock.si/guides/mcp-vs-a2a MCP connects an agent to tools and data; A2A connects agents to other agents. Learn when to use each, how they fit together, and what changed in 2026. Explainers Mon, 05 Oct 2026 12:00:00 +0000 https://drydock.si/guides/mcp-vs-a2a https://drydock.si/guides/langgraph-vs-crewai-vs-openai-agents-sdk LangGraph vs CrewAI vs OpenAI Agents SDK compared on control flow, state, human-in-the-loop, model support and learning curve, plus what replaced AutoGen. Comparisons Mon, 05 Oct 2026 12:00:00 +0000 https://drydock.si/guides/langgraph-vs-crewai-vs-openai-agents-sdk https://drydock.si/guides/how-to-build-an-mcp-server-python Build a working MCP server in Python with the official SDK 2.x and the stateless 2026-07-28 spec, then connect it to Claude Code, Cursor, and VS Code hosts. Tutorials Mon, 05 Oct 2026 12:00:00 +0000 https://drydock.si/guides/how-to-build-an-mcp-server-python https://drydock.si/guides/e2b-vs-daytona-vs-modal E2B vs Daytona vs Modal for running AI agent code safely: isolation, startup, persistence, time limits, GPUs and SDKs compared, with code for each sandbox. Comparisons Mon, 05 Oct 2026 12:00:00 +0000 https://drydock.si/guides/e2b-vs-daytona-vs-modal https://drydock.si/guides/best-typescript-ai-agent-frameworks The best TypeScript agent framework for your project: Mastra, Vercel AI SDK, LangGraph.js, OpenAI Agents SDK and Google ADK compared, with code and trade-offs. Comparisons Mon, 05 Oct 2026 12:00:00 +0000 https://drydock.si/guides/best-typescript-ai-agent-frameworks https://drydock.si/guides/ai-agent-memory-mem0-vs-letta-vs-zep AI agent memory explained: how Mem0, Letta and Zep store and recall long-term context, how they differ from LangGraph's built-in memory, and which to choose. Comparisons Mon, 05 Oct 2026 12:00:00 +0000 https://drydock.si/guides/ai-agent-memory-mem0-vs-letta-vs-zep https://drydock.si/guides/agents-md-guide What AGENTS.md is, how Codex, Claude Code, Cursor, Copilot and Gemini CLI load it, how it compares with CLAUDE.md, plus a copy-paste template that works. How-to Mon, 05 Oct 2026 12:00:00 +0000 https://drydock.si/guides/agents-md-guide https://huggingface.co/papers/2610.05303 [HF Daily Papers] 11 upvotes HF Daily Papers Sat, 03 Oct 2026 20:00:00 +0000 drydock.si:hfp-2610.05303 https://huggingface.co/papers/2610.05033 [HF Daily Papers] 2 upvotes HF Daily Papers Sat, 03 Oct 2026 20:00:00 +0000 drydock.si:hfp-2610.05033 https://www.science.org/content/article/exclusive-ai-agent-emailed-hundreds-researchers-help-it-told-us-why [Hacker News] 50 points · 80 comments Hacker News Sat, 03 Oct 2026 10:07:08 +0000 drydock.si:hn-49942865 https://offrun.dev/ [Hacker News] 78 points · 65 comments Hacker News Sat, 03 Oct 2026 08:40:17 +0000 drydock.si:hn-49942434 https://royapakzad.substack.com/p/multilingual-ai-agents [Hacker News] 44 points · 1 comments Hacker News Fri, 02 Oct 2026 20:39:05 +0000 drydock.si:hn-49938326 https://pipod.dev/ [Hacker News] 119 points · 47 comments Hacker News Fri, 02 Oct 2026 19:10:38 +0000 drydock.si:hn-49937304 https://arxiv.org/abs/2610.03715v1 [arXiv] We introduce 4DCodeBench, a benchmark for 4D inverse graphics through code generation, in which agents reconstruct dynamic scenes from video as executable graphics programs. To accomplish this, agents must translate visual observations into compact representations of scene struct… arXiv Fri, 02 Oct 2026 17:58:49 +0000 drydock.si:arxiv-2610.03715v1 https://arxiv.org/abs/2610.03634v1 [arXiv] Terminal-using agents benefit from reinforcement learning (RL) in coding, debugging, and other multi-step terminal tasks. In these tasks, later commands often depend on information or intermediate results produced by earlier commands. However, existing trajectory-level and step-l… arXiv Fri, 02 Oct 2026 17:24:48 +0000 drydock.si:arxiv-2610.03634v1 https://arxiv.org/abs/2610.03631v1 [arXiv] Designing a scientific instrument tests whether language-model agents can do physics rather than recall it, provided the grading cannot be argued with. We introduce NeutronGym, to our knowledge the first executable environment for neutron instrument design: agents build instrumen… arXiv Fri, 02 Oct 2026 17:23:52 +0000 drydock.si:arxiv-2610.03631v1 https://arxiv.org/abs/2610.03591v1 [arXiv] Understanding and assessing natural hazards is essential for disaster preparedness and risk reduction. Recent advances in large language models have spurred growing interest in AI agents for hazard analysis, particularly their ability to integrate scientific data, models, and too… arXiv Fri, 02 Oct 2026 16:59:19 +0000 drydock.si:arxiv-2610.03591v1 https://arxiv.org/abs/2610.03585v1 [arXiv] Security benchmarks for LLM-based agents often report the attack success rate (ASR) as a measure of model robustness and use these scores to compare different models and defense mechanisms, assuming that they describe the security of the agent. In this paper, we explore whether i… arXiv Fri, 02 Oct 2026 16:55:50 +0000 drydock.si:arxiv-2610.03585v1 https://arxiv.org/abs/2610.03574v1 [arXiv] We introduce HyperBrowseComp, a multilingual and multimodal browsing benchmark comprising 423 manually authored and human-validated questions across 13 languages, written by native or highly proficient speakers. Questions are designed to be extremely challenging. Each question ta… arXiv Fri, 02 Oct 2026 16:48:49 +0000 drydock.si:arxiv-2610.03574v1 https://arxiv.org/abs/2610.03564v1 [arXiv] Financial AI agents must do more than retrieve facts: investment workflows require correct quantitative execution, reliable use of procedural resources, and auditable structured outputs. We introduce FinSkillBench, an evaluation suite of 2,603 point in time episodes across 12 sub… arXiv Fri, 02 Oct 2026 16:44:04 +0000 drydock.si:arxiv-2610.03564v1 https://arxiv.org/abs/2610.03448v1 [arXiv] LLM agents increasingly screen tool outputs with small prompt-injection detectors, and teams choose among detectors by their scores on public benchmarks. We ask whether those scores predict how a detector behaves inside an agent. We replay the ground-truth tool calls of two agent… arXiv Fri, 02 Oct 2026 15:30:11 +0000 drydock.si:arxiv-2610.03448v1 https://distantprovince.substack.com/p/the-four-horsemen-of-agentic-coding [Hacker News] 114 points · 89 comments Hacker News Fri, 02 Oct 2026 15:19:56 +0000 drydock.si:hn-49934511 https://arxiv.org/abs/2610.03394v1 [arXiv] Emerging multi-agent LLMs demand privacy-preserving edge deployment, yet current inference systems struggle with these collaborative workflows. Specifically, the memory-bound decode phase causes severe bus contention on unified memory architectures (UMA), paralyzing naive CPU-GPU… arXiv Fri, 02 Oct 2026 14:43:42 +0000 drydock.si:arxiv-2610.03394v1 https://news.ycombinator.com/item?id=49934037 [Hacker News] 29 points · 44 comments Hacker News Fri, 02 Oct 2026 14:38:35 +0000 drydock.si:hn-49934037 https://arxiv.org/abs/2610.03356v1 [arXiv] Large Language Model (LLM) agents are increasingly deployed in high-stakes settings such as industrial maintenance and equipment fault troubleshooting, where workers occupy a variety of roles. A capable agent must therefore act in a way that is calibrated to user's role: taking a… arXiv Fri, 02 Oct 2026 14:21:26 +0000 drydock.si:arxiv-2610.03356v1 https://arxiv.org/abs/2610.03319v1 [arXiv] Large Language Models (LLMs) increasingly support Uncrewed Aerial Vehicle (UAV) swarm operations such as data collection scheduling, where the model reads structured sensor reports and decides which sensors to visit. An adversary who quietly manipulates those reports can redirect… arXiv Fri, 02 Oct 2026 13:54:35 +0000 drydock.si:arxiv-2610.03319v1 https://arxiv.org/abs/2610.03315v1 [arXiv] Trajectory evaluation is essential for improving the reliability of LLM-based agents, but production use makes it expensive to run repeatedly. Modern agents generate long traces containing tool calls, observations, retries, and external outputs, while not all raw tokens are equal… arXiv Fri, 02 Oct 2026 13:51:06 +0000 drydock.si:arxiv-2610.03315v1 https://arxiv.org/abs/2610.03226v1 [arXiv] GPU kernels generated by large language model (LLM) agents can remain less efficient than expert implementations, but runtime alone does not reveal how the gap relates to design discovery and implementation. We introduce D2K-Bench, a diagnostic benchmark of 26 tasks and 85 worklo… arXiv Fri, 02 Oct 2026 12:42:12 +0000 drydock.si:arxiv-2610.03226v1 https://arxiv.org/abs/2610.03223v1 [arXiv] Long-horizon LLM agents are typically trained with sparse outcome rewards, making trajectory-level objectives too coarse to distinguish the contribution of individual decisions. Step-level credit assignment provides finer-grained supervision, but its estimates can be unreliable b… arXiv Fri, 02 Oct 2026 12:38:53 +0000 drydock.si:arxiv-2610.03223v1 https://arxiv.org/abs/2610.03213v1 [arXiv] Tool-equipped AI agents use tool calls to access data and act on external systems. Horizontal growth of agentic systems increases the number of these interactions, and further motivates the need for automated, per-call oversight that can operate at low latency and/or on-prem. Con… arXiv Fri, 02 Oct 2026 12:31:01 +0000 drydock.si:arxiv-2610.03213v1 https://arxiv.org/abs/2610.03195v1 [arXiv] As LLM agents decide on users' behalf which product to buy, which hotel to book, or which paper to cite, a preference for items from certain sources (the sites or services they come from) shapes what users receive and which sources are selected. We study source preference in end-… arXiv Fri, 02 Oct 2026 12:10:03 +0000 drydock.si:arxiv-2610.03195v1 https://aweb.ai [Hacker News] 42 points · 28 comments Hacker News Thu, 01 Oct 2026 22:02:31 +0000 drydock.si:hn-49927587 https://github.com/graphene-data/graphene [Hacker News] 32 points · 11 comments Hacker News Thu, 01 Oct 2026 21:29:52 +0000 drydock.si:hn-49927295 https://huggingface.co/papers/2610.03715 [HF Daily Papers] 19 upvotes HF Daily Papers Thu, 01 Oct 2026 20:00:00 +0000 drydock.si:hfp-2610.03715 https://huggingface.co/papers/2610.04116 [HF Daily Papers] 4 upvotes HF Daily Papers Thu, 01 Oct 2026 20:00:00 +0000 drydock.si:hfp-2610.04116 https://huggingface.co/papers/2610.04137 [HF Daily Papers] 3 upvotes HF Daily Papers Thu, 01 Oct 2026 20:00:00 +0000 drydock.si:hfp-2610.04137 https://turbopuffer.com/blog/rip-vector-database [Hacker News] 399 points · 116 comments Hacker News Thu, 01 Oct 2026 16:01:56 +0000 drydock.si:hn-49923466