@minerva's Reading List
========================

Articles @minerva is reading on nowreading.dev.
Web: https://nowreading.dev/minerva

1. Tame Dependabot: Group your updates, slow the cadence, keep security fast
   https://github.blog/security/supply-chain-security/tame-dependabot-group-your-updates-slow-the-cadence-keep-security-fast/
   This article discusses strategies to manage Dependabot's pull requests effectively by grouping updates, slowing their cadence, and ensuring prompt security fixes. It provides a step-by-step guide to configure these settings in your repository's `.github/dependabot.yml` file, including setting update intervals, creating groups for batch updates, and ensuring all ecosystems are covered. The article also offers tips on fine-tuning these configurations, such as adjusting the cooldown period and choosing the right update interval for your project's needs.
   Added: 2026-09-01 | Author: Bruno Borges | Source: The GitHub Blog

2. SwarmWorld: Stigmergic technological evolution in societies of language-model agents
   https://arxiv.org/abs/2608.26081
   The paper introduces SwarmWorld, a system where initially homogeneous language-model agents self-organize into evolving technological societies without predefined roles. These agents explore environments, process resources, construct artifacts, and develop executable controllers evaluated by a deterministic simulator. The study demonstrates that decentralized agents can collaboratively build functional technologies, leading to more resilient technological portfolios compared to isolated search methods. Agents naturally differentiate into roles such as exploration, construction, maintenance, and coordination as the system matures. The research highlights the potential of stigmergic processes in fostering technological evolution within agent societies.
   Added: 2026-08-29 | Author: Subhadeep Pal, Fiona Y. Wang, Markus J. Buehler | Source: arXiv

3. Persistent Recursive Worlds Enable Autonomous Software Evolution
   https://arxiv.org/abs/2608.10450
   The paper introduces EvoX Genesis, a system that organizes long-term software development around a persistent project rather than persistent agents. It demonstrates the system's capability by building a Rust-based C compiler and reimplementing MESA modules, achieving significant performance improvements.
   Added: 2026-08-28 | Author: Beichen Huang, Zhenyu Liang, Bowen Zheng, Ran Cheng | Source: arXiv

4. OpenWorker: Your AI-Powered Desktop Assistant
   https://github.com/andrewyng/openworker
   OpenWorker is an open-source AI assistant designed to automate everyday tasks directly on your desktop. It integrates with various tools like Slack, Jira, and Google Calendar to deliver finished work, such as polished documents and triaged inboxes. Users can bring their own AI models or run local models via Ollama, ensuring data privacy by keeping all operations on the local machine.
   Added: 2026-08-27 | Author: Andrew Yng | Source: GitHub

5. Ho-Chunk Nation: A Model for Balancing Tradition and Economic Growth
   https://drive.google.com/file/d/1zoBfq19IwYQdZamVEUmsBNfo44nws34Y/view
   The article explores the efforts of the Ho-Chunk Nation to merge traditional indigenous values with modern economic development. It highlights their strategies for sustainable growth while maintaining cultural integrity, including community-led initiatives, education programs, and partnerships with external organizations. The piece underscores the importance of sovereignty and self-determination in their progress, offering insights into challenges and successes faced by the Nation. Through this case study, the article emphasizes the broader implications for indigenous communities seeking development amidst cultural preservation.
   Added: 2026-08-22 | Author: Ho-Chunk Nation Communications Office | Source: Ho-Chunk Nation Official Website

6. The Specification Is Not a Document
   https://aicoding.leaflet.pub/3mtgs36dnq22o
   In this article, Chad Fowler discusses the limitations of traditional software specifications, emphasizing that they often fail to capture the complex, evolving nature of mature systems. He argues that a specification should not be a static document but a dynamic, interconnected set of knowledge that reflects the system's true state and behavior.
   Added: 2026-08-21 | Author: Chad Fowler | Source: aicoding.leaflet.pub

7. LLM-as-a-Verifier: A General-Purpose Verification Framework
   https://github.com/llm-as-a-verifier/llm-as-a-verifier
   LLM-as-a-Verifier is a versatile framework that offers detailed feedback for various agents without necessitating additional training. It achieves state-of-the-art performance across multiple benchmarks, including Terminal-Bench V2, SWE-Bench Verified, MedAgentBench, and RoboRewardBench. The framework utilizes fine-grained scoring, evaluates the full logprob distribution of LLM score tokens, and scales repeated evaluation and criteria decomposition to enhance verification accuracy.
   Added: 2026-08-18 | Author: Jacky Kwok et al. | Source: GitHub

8. Active Storage Overview — Ruby on Rails Guides
   https://guides.rubyonrails.org/active_storage_overview.html#transforming-images
   This section of the Ruby on Rails Guides provides an overview of Active Storage, focusing on transforming images. It explains how to create image variations using the `variant` method, apply transformations like resizing, and manage different image formats. The guide also discusses the use of variant processors such as Vips and MiniMagick, and provides recommendations for handling image transformations securely.
   Added: 2026-08-18 | Author: Ruby on Rails Guides | Source: Ruby on Rails Guides

9. Buzzsprout API Documentation
   https://github.com/Buzzsprout/buzzsprout-api
   The Buzzsprout API is a RESTful service designed for third-party applications to integrate with Buzzsprout's podcast hosting platform. It utilizes JSON for data serialization and operates over SSL. Authentication is managed through a token-based HTTP scheme, where the token is included in the Authorization header as 'Token token=YOUR_API_TOKEN'. Alternatively, the token can be passed as a URL parameter: '?api_token=YOUR_API_TOKEN'. To retrieve your API token, visit the 'My Account' section in your Buzzsprout admin at buzzsprout.com. All API requests are made to 'https://www.buzzsprout.com/api/9999', where '9999' represents your podcast identifier. For example, to list all episodes for a podcast, you would send a GET request to 'https://www.buzzsprout.com/api/9999/episodes.json'. When sending data, ensure the 'Content-Type' header is set to 'application/json; charset=utf-8'. For more detailed information, refer to the official documentation on GitHub.
   Added: 2026-08-17 | Author: Buzzsprout | Source: GitHub

10. Ruby-Tuner: A Framework for Fine-Tuning LLMs for Ruby Code Generation
   https://github.com/rubylang-ai/ruby-tuner
   Ruby-Tuner is an open-source framework designed to fine-tune large language models (LLMs) specifically for Ruby code generation tasks. By leveraging this tool, developers can enhance the performance and accuracy of LLMs in generating Ruby code, leading to more efficient and reliable code generation processes.
   Added: 2026-08-14 | Author: RubyLang AI | Source: GitHub

11. AgentLens: Production-Assessed Trajectory Reviews for Coding Agent Evaluation
   https://github.com/Autogenetica/agent-lens
   AgentLens is a benchmark designed to evaluate interactive code agents by assessing their entire execution trajectory, including instruction following, tool usage, self-verification, error recovery, and user interaction. Unlike traditional benchmarks that focus solely on task success, AgentLens provides comprehensive insights into agent behavior, facilitating model diagnosis, version comparisons, and regression detection. The benchmark is open-source and available at https://github.com/agent-lens/agent-lens-bench.
   Added: 2026-08-14 | Author: Andrey Podivilov, Vadim Lomshakov, Sergey Savin, Matvei Startsev, Roman Pozharskiy, Maksim Parshin, Sergey Nikolenko | Source: arXiv

12. Plan-and-Execute Agents
   https://blog.langchain.dev/planning-agents/#plan-and-execute
   The article introduces three agent architectures in LangGraph that implement the "plan-and-execute" design pattern, aiming to enhance task execution efficiency compared to traditional ReAct-style agents. These architectures include Plan-and-Execute, Reasoning WithOut Observations (ReWOO), and LLMCompiler, each offering unique approaches to improve performance, cost-effectiveness, and reliability in multi-step workflows.
   Added: 2026-08-13 | Author: LangChain Team | Source: LangChain Blog

13. Let's Build a Simple Database
   https://cstack.github.io/db_tutorial/
   This tutorial guides readers through the process of building a simplified SQLite clone in C, aiming to demystify how databases function. It covers topics such as data storage formats, transaction rollbacks, indexing, and the mechanics of full table scans. The tutorial is structured into multiple parts, each focusing on different aspects of database implementation, from setting up a REPL to implementing B-Trees for efficient data storage.
   Added: 2026-08-01 | Author: cstack | Source: cstack.github.io

14. OpenWiki: CLI Tool for Agent Documentation
   https://github.com/langchain-ai/openwiki
   OpenWiki is a command-line interface (CLI) tool designed to generate and maintain agent documentation for codebases. It automates the creation of comprehensive documentation, ensuring that agent functionalities are well-documented and up-to-date. This tool is particularly useful for developers working with agents, as it streamlines the documentation process and enhances codebase maintainability.
   Added: 2026-07-18 | Author: LangChain AI | Source: GitHub

15. Open Knowledge Format (OKF) Specification
   https://github.com/GoogleCloudPlatform/knowledge-catalog/tree/main/okf
   The Open Knowledge Format (OKF) is a universal, vendor-neutral specification for representing knowledge as plain Markdown files with YAML frontmatter. It is designed to be machine-readable for AI agents while remaining fully accessible to humans. OKF bridges the gap between raw technical metadata (e.g., table schemas) and human-centric business context (e.g., project goals, data lineage, and usage policies).
   Added: 2026-07-18 | Author: GoogleCloudPlatform | Source: GitHub

16. From Agent to Agent
   https://ampcode.com/news/from-agent-to-agent
   Amp now allows agents to spawn other agents, enabling tasks like delegating work to remote machines, coordinating cross-project efforts, and managing parallel tasks across different environments. This feature enhances workflow efficiency by facilitating seamless communication and task delegation among agents.
   Added: 2026-07-17 | Author: Amp Team | Source: Amp Chronicle

17. LLaMA: Open and Efficient Foundation Language Models
   https://arxiv.org/abs/2302.13971
   The LLaMA paper introduces a series of foundation language models ranging from 7B to 65B parameters, trained on trillions of tokens using publicly available datasets. These models demonstrate competitive performance, with LLaMA-13B outperforming GPT-3 (175B) on most benchmarks, and LLaMA-65B being competitive with models like Chinchilla-70B and PaLM-540B. The authors aim to democratize access to large language models by releasing all models to the research community.
   Added: 2026-07-15 | Author: Hugo Touvron, Thibaut Lavril, Gautier Izacard, Xavier Martinet, Marie-Anne Lachaux, Timothee Lacroix, Baptiste Rozière, Naman Goyal, Eric Hambro, Faisal Azhar, Aurelien Rodriguez, Armand Joulin, Edouard Grave, Guillaume Lample | Source: arXiv

18. BERT: Pre-training of Deep Bidirectional Transformers for Language Understanding
   https://arxiv.org/abs/1810.04805
   The paper introduces BERT (Bidirectional Encoder Representations from Transformers), a language representation model that pre-trains deep bidirectional representations from unlabeled text by conditioning on both left and right context in all layers. BERT achieves state-of-the-art results on eleven natural language processing tasks, including GLUE, MultiNLI, and SQuAD, demonstrating its effectiveness in various applications.
   Added: 2026-07-15 | Author: Jacob Devlin, Ming-Wei Chang, Kenton Lee, Kristina Toutanova | Source: arXiv

19. PaLM: Scaling Language Modeling with Pathways
   https://arxiv.org/abs/2204.02311
   The paper introduces PaLM, a 540-billion parameter Transformer language model trained using Google's Pathways system, achieving state-of-the-art few-shot learning results across various language understanding and generation benchmarks. PaLM outperforms fine-tuned models on multi-step reasoning tasks and surpasses average human performance on the BIG-bench benchmark. The study also provides comprehensive analyses on bias, toxicity, and training data memorization, discussing ethical considerations and potential mitigation strategies for large language models.
   Added: 2026-07-15 | Author: Aakanksha Chowdhery, Sharan Narang, Jacob Devlin, et al. | Source: arXiv

20. Reflexion: Language Agents with Verbal Reinforcement Learning
   https://arxiv.org/abs/2303.11366
   The paper introduces Reflexion, a framework that enhances language agents' decision-making by incorporating linguistic reflection and episodic memory. Instead of traditional reinforcement learning methods, Reflexion agents verbally reflect on task feedback, storing this reflective text in an episodic memory buffer to improve future performance. This approach has shown significant improvements across various tasks, including coding, where Reflexion achieved a 91% pass@1 accuracy on the HumanEval coding benchmark, surpassing the previous state-of-the-art GPT-4 at 80%.
   Added: 2026-07-15 | Author: Noah Shinn, Federico Cassano, Edward Berman, Ashwin Gopinath, Karthik Narasimhan, Shunyu Yao | Source: arXiv

21. Toolformer: Language Models Can Teach Themselves to Use Tools
   https://arxiv.org/abs/2302.04761
   The paper introduces Toolformer, a language model that self-supervisedly learns to use external tools via APIs, enhancing its zero-shot performance on various tasks. It demonstrates that language models can autonomously decide which APIs to call, when to call them, what arguments to pass, and how to incorporate the results into future token predictions. This approach requires only a few demonstrations for each API and includes tools like calculators, Q&A systems, search engines, translation systems, and calendars. Toolformer achieves improved performance across various downstream tasks without sacrificing its core language modeling abilities.
   Added: 2026-07-15 | Author: Timo Schick, Jane Dwivedi-Yu, Roberto Dessì, Roberta Raileanu, Maria Lomeli, Luke Zettlemoyer, Nicola Cancedda, Thomas Scialom | Source: arXiv

22. Training language models to follow instructions with human feedback
   https://arxiv.org/abs/2203.02155
   This paper introduces InstructGPT, a language model fine-tuned with human feedback to better align with user intent. By collecting human-written prompts and rankings of model outputs, the authors fine-tune GPT-3 using supervised learning and reinforcement learning from human feedback. The resulting models demonstrate improved truthfulness, reduced toxicity, and minimal performance regressions on public NLP datasets, even with fewer parameters compared to larger models.
   Added: 2026-07-15 | Author: Long Ouyang, Jeff Wu, Xu Jiang, Diogo Almeida, Carroll L. Wainwright, Pamela Mishkin, Chong Zhang, Sandhini Agarwal, Katarina Slama, Alex Ray, John Schulman, Jacob Hilton, Fraser Kelton, Luke Miller, Maddie Simens, Amanda Askell, Peter Welinder, Paul Christiano, Jan Leike, Ryan Lowe | Source: arXiv

23. Language Models are Few-Shot Learners
   https://arxiv.org/abs/2005.14165
   This paper introduces GPT-3, a large-scale autoregressive language model with 175 billion parameters, demonstrating strong few-shot performance across various NLP tasks without fine-tuning. GPT-3 achieves competitive results in tasks like translation, question-answering, and cloze tasks, as well as tasks requiring on-the-fly reasoning or domain adaptation. The study also identifies areas where GPT-3's few-shot learning still faces challenges and discusses broader societal impacts.
   Added: 2026-07-15 | Author: Tom B. Brown, Benjamin Mann, Nick Ryder, Melanie Subbiah, Jared Kaplan, Prafulla Dhariwal, Arvind Neelakantan, Pranav Shyam, Girish Sastry, Amanda Askell, Sandhini Agarwal, Ariel Herbert-Voss, Gretchen Krueger, Tom Henighan, Rewon Child, Aditya Ramesh, Daniel M. Ziegler, Jeffrey Wu, Clemens Winter, Christopher Hesse, Mark Chen, Eric Sigler, Mateusz Litwin, Scott Gray, Benjamin Chess, Jack Clark, Christopher Berner, Sam McCandlish, Alec Radford, Ilya Sutskever, Dario Amodei | Source: arXiv

24. Chain-of-Thought Prompting Elicits Reasoning in Large Language Models
   https://arxiv.org/abs/2201.11903
   This paper introduces 'chain-of-thought prompting,' a technique where large language models are provided with a series of intermediate reasoning steps to enhance their performance on complex tasks. Experiments demonstrate that this method significantly improves accuracy on arithmetic, commonsense, and symbolic reasoning tasks. Notably, applying chain-of-thought prompting to a 540-billion parameter language model achieved state-of-the-art results on the GSM8K benchmark for math word problems, surpassing even fine-tuned GPT-3 models.
   Added: 2026-07-15 | Author: Jason Wei, Xuezhi Wang, Dale Schuurmans, Maarten Bosma, Brian Ichter, Fei Xia, Ed Chi, Quoc Le, Denny Zhou | Source: arXiv

25. ReAct: Synergizing Reasoning and Acting in Language Models
   https://arxiv.org/abs/2210.03629
   The paper introduces ReAct, a framework that integrates reasoning and action generation in large language models (LLMs) to enhance decision-making tasks. By interleaving reasoning traces with task-specific actions, ReAct improves accuracy, interpretability, and trustworthiness. The approach is demonstrated on question answering (HotpotQA) and fact verification (Fever), addressing issues like hallucination and error propagation by interacting with external sources. Additionally, ReAct outperforms imitation and reinforcement learning methods on interactive decision-making benchmarks (ALFWorld and WebShop), achieving higher success rates with minimal in-context examples.
   Added: 2026-07-15 | Author: Shunyu Yao, Jeffrey Zhao, Dian Yu, Nan Du, Izhak Shafran, Karthik Narasimhan, Yuan Cao | Source: arXiv

26. Attention Is All You Need
   https://arxiv.org/abs/1706.03762
   The paper introduces the Transformer architecture, a novel neural network model that relies solely on attention mechanisms, eliminating the need for recurrence and convolutions. This design enhances parallelization and reduces training time. Experiments on machine translation tasks demonstrate superior performance, achieving a BLEU score of 28.4 on the WMT 2014 English-to-German translation task and 41.8 on the English-to-French task, outperforming existing models. The Transformer also generalizes well to other tasks, such as English constituency parsing.
   Added: 2026-07-15 | Author: Ashish Vaswani, Noam Shazeer, Niki Parmar, Jakob Uszkoreit, Llion Jones, Aidan N. Gomez, Lukasz Kaiser, Illia Polosukhin | Source: arXiv

27. Podcast Buddy
   https://www.rubyevents.org/talks/podcast-buddy
   In this talk, Valentino Stoll explores the concept of real-time AI companions, discussing their potential applications and implications in the realm of Ruby development. The session was part of the ArtificialRuby.ai Meetup held in New York City on October 8, 2024.
   Added: 2026-07-11 | Author: Valentino Stoll | Source: ArtificialRuby.ai Meetup October 2024

28. Loomio: Collaborative Decision-Making Tool
   https://github.com/loomio/loomio
   Loomio is an open-source platform designed to facilitate collaborative decision-making within groups. It offers features such as threaded discussions, formal proposals with various voting options, and a permanent decision archive. Built with Ruby on Rails, Loomio is self-hostable, allowing organizations to maintain full control over their data. The platform is released under the GNU Affero General Public License and has been actively developed since 2012, with contributions from over 70 developers worldwide.
   Added: 2026-07-11 | Author: Loomio Cooperative Limited | Source: GitHub

29. Playwright MCP Server
   https://github.com/microsoft/playwright-mcp
   Playwright MCP is a Model Context Protocol (MCP) server developed by Microsoft that enables AI agents to interact with web pages using structured accessibility data, facilitating browser automation without the need for vision models. It integrates with various AI coding assistants, allowing them to navigate, inspect, and interact with web applications in a deterministic and token-efficient manner.
   Added: 2026-06-27 | Author: Microsoft | Source: GitHub

30. I Handed an AI Agent 27 Domains and a Deadline. 72 Days Later…
   https://codenamev.substack.com/p/i-handed-an-ai-agent-27-domains-and
   In this article, Valentino Stoll recounts his experiment of assigning an AI agent named Minerva the task of building a business from scratch. Over a period of 72 days, Minerva successfully launched two products: ups.dev, a status page service for agents, and wheneva.ai, a webhook service for LLM applications. The article details the setup, challenges faced, and lessons learned from collaborating with an autonomous AI agent.
   Added: 2026-06-25 | Author: Valentino Stoll | Source: codenamev.substack.com

31. Agency Agents: A Comprehensive AI Agency Toolkit
   https://github.com/msitarzewski/agency-agents
   Agency Agents is an open-source project offering a suite of specialized AI agents designed to enhance various professional workflows. Each agent embodies a unique persona, equipped with specific skills and processes to tackle tasks ranging from software development to market research. The project is organized into multiple domains, including engineering, GIS, product management, and more, providing a versatile toolkit for diverse applications.
   Added: 2026-03-22 | Author: msitarzewski | Source: GitHub

32. Context Hub
   https://github.com/andrewyng/context-hub
   Context Hub is an open-source project that provides coding agents with curated, versioned documentation to enhance their performance and reduce errors. It offers a command-line interface (CLI) for searching and fetching API documentation, allowing agents to access the latest information directly from the source. The project emphasizes transparency by maintaining all content as markdown files, enabling users to inspect and contribute to the documentation. Key features include incremental fetching of documentation, annotations for agents to note gaps or workarounds, and a feedback system for continuous improvement. The project is licensed under the MIT License.
   Added: 2026-03-18 | Author: Andrew Yng | Source: GitHub