# loop engineering (blogs) — RSS Amplifier Recent posts from the 11 feeds in the RSS Amplifier directory that cover loop engineering. Page: Feed: --- ## [Your Ultimate Resource List for Learning AI Engineering in 2026](https://aiagentssimplified.substack.com/p/your-ultimate-resource-list-for-learning) _2026-08-20 · AI Agents Simplified · AI Agents Simplified_ Welcome to the 68th issue of AI Agents Simplified 🍻 ## [AI Traffic Now Converts 54% Better Than Everything Else. Many Sites Can't Be Read by It!](https://aiagentssimplified.substack.com/p/ai-traffic-now-converts-54-better) _2026-08-18 · AI Agents Simplified · AI Agents Simplified_ AI assistants are reading the web on your users' behalf. Adobe just put a number on how badly most pages handle that and it turns out "AI visibility" is a retrieval problem, not a marketing one. ## [Least-Data Supervised Fine-Tuning of Qwen2.5-0.5B-Instruct](https://cobusgreyling.substack.com/p/least-data-supervised-fine-tuning) _2026-08-18 · Cobus Greyling · Cobus Greyling on LLMs, NLU, NLP, chatbots & voicebots_ It is the smallest model form Alibaba's Qwen 2.5 series… ## [I Built the Outreach System I Couldn’t Find](https://aiagentssimplified.substack.com/p/i-built-the-outreach-system-i-couldnt) _2026-08-13 · AI Agents Simplified · AI Agents Simplified_ How a simple outreach workflow slowly turned into a system that checks replies, handles follow-ups, personalizes messages, and keeps track of what happened. ## [Was a new category of Language Model just created by NVIDIA & Meta?](https://cobusgreyling.substack.com/p/was-a-new-category-of-language-model) _2026-08-13 · Cobus Greyling · Cobus Greyling on LLMs, NLU, NLP, chatbots & voicebots_ Meta AI released Muse Glimmer and NVIDIA released Nemotron 3 Lightning… ## [The org-chart mistake that kills agent fleets](https://eugenevyborov.substack.com/p/the-org-chart-question-every-agent) _2026-08-13 · Eugene Vyborov · Eugene Vyborov Blog (AI future, entrepreneurship)_ Everyone deploying AI agent fleets faces the same structural question - and the two most popular answers are both wrong. ## [LAI #138: The Agent Reality Check](https://learnaitogethernewsletter.substack.com/p/lai-138-the-agent-reality-check) _2026-08-13 · Louis-François Bouchard · Learn AI Together_ Agent evals, retry tracing, runaway costs, and the context your agents actually need. ## [When a Rails or LMS Platform Becomes Hard to Change](https://joeywang.github.io/posts/rails-lms-reliability-modernization/) _2026-08-13 · Joey Wang · Joey_ A practical way for small education and SaaS teams to diagnose platform friction before committing to a rewrite. ## [When Ubuntu Builds Fail on libvips but macOS Does Not](https://joeywang.github.io/posts/libvips-ubuntu-build-mismatch/) _2026-08-12 · Joey Wang · Joey_ Why a Rails test suite can fail because Ubuntu ships an older native libvips, how ruby-vips fits into the picture, and how to upgrade libvips safely without confusing the system package with the Ruby gem. ## [A Portable Workflow for Testing Small Local LLMs on Modest Hardware](https://joeywang.github.io/posts/portable-local-llm-benchmark/) _2026-08-12 · Joey Wang · Joey_ A repeatable way to download, serve, benchmark, and choose small GGUF models for Hermes and coding-agent work across different CPU and memory budgets. ## [Browse ThreatCrush Modules (Sponsored)](https://crawlproof.com/a/c9DsUvOGPGP8) _2026-08-12 · **Sponsored**_ Install community-built security modules to extend real-time threat intelligence and monitoring ## [Orchestrating Hermes with Local Gemma: Cheap Routing, Delegation, and Safe Rails Workflows](https://joeywang.github.io/posts/orchestrating-hermes-with-local-gemma/) _2026-08-12 · Joey Wang · Joey_ A practical guide to using Hermes Agent with a local Gemma model for routine coding work, while reserving cloud models for tasks that genuinely need them. ## [A Hands-on Real-World ML project: Inside a $35K Kaggle competition (pt.1)](https://aiagentssimplified.substack.com/p/a-hands-on-real-world-ml-project) _2026-08-11 · AI Agents Simplified · AI Agents Simplified_ Follow this guide and learn how to do a machine learning project step by step ## [When Artificial Stimulants Look Like Strength](https://georgesiosi.substack.com/p/when-artificial-stimulants-look-like) _2026-08-11 · Siosi | The Digital Wayfinder · Siosi's Substack_ How gaps in Sunspot Theory led to Nodal Bend Theory ## [NVIDIA Nemotron 3.5 Lightning](https://cobusgreyling.substack.com/p/nvidia-nemotron-35-lightning) _2026-08-11 · Cobus Greyling · Cobus Greyling on LLMs, NLU, NLP, chatbots & voicebots_ The Execution Engine for Always-On Agents ## [Meta Muse Glimmer](https://cobusgreyling.substack.com/p/meta-muse-glimmer) _2026-08-10 · Cobus Greyling · Cobus Greyling on LLMs, NLU, NLP, chatbots & voicebots_ Muse Glimmer is an open agentic model that runs on your device ## [Agent 基建的三层爆发：电脑、运行时、技能方法论](https://joeywang.github.io/posts/agent-infrastructure-three-layer-surge/) _2026-08-10 · Joey Wang · Joey_ 从 Prime Agent、Cloudflare Computer、agent-skills、mattpocock/skills 和 Superpowers 看 Agent 基础设施正在如何从聊天框走向持久工作台、长任务运行时和工程方法论。 ## [One agent or many?](https://www.anup.io/one-agent-or-many/) _2026-08-10 · Anup Jadhav · Anup Jadhav_ A practical guide for AI engineers choosing between single-agent and multi-agent systems, based on task structure, context and coordination costs. ## [Tightening a dev VM that was quietly living without a firewall](https://joeywang.github.io/posts/vm-security-tightening/) _2026-08-10 · Joey Wang · Joey_ What a routine security audit found on a small always-on dev VM — unauthenticated databases, no host firewall, over-privileged automation — and the fixes that made it boring again. ## [framesmith 2.0: don't ask the model for taste](https://victorvelazquez.dev/blog/framesmith-2-0-design-quality-without-an-api-key/) _2026-08-10 · Victor Velázquez_ The thing agents are worst at is visual taste, so they fake it. framesmith 2.0 generates a full design language deterministically, no API key, AA by construction. ## [Golden Rules for Cheaper, Safer LLM Agents](https://joeywang.github.io/posts/golden-rules-for-cheaper-safer-llm-agents/) _2026-08-09 · Joey Wang · Joey_ Notes on turning repeated AI workflows into scripts, skills, cron jobs, and guardrails so agents become cheaper, safer, and more useful over time. ## [Realtime AI Arbitrage Terminal (Sponsored)](https://crawlproof.com/a/pXaW7xgbRqcX) _2026-08-09 · **Sponsored**_ Realtime auto-trading across exchanges with scored setups, backtesting, and risk controls. ## [Loop Engineering, Graph Engineering, and the Nature of Working with LLMs](https://joeywang.github.io/posts/loop-engineering-graph-engineering-and-llm-interaction/) _2026-08-09 · Joey Wang · Joey_ A practical reflection on prompt engineering, context engineering, loop engineering, graph engineering, and what they reveal about the shape of human interaction with LLMs. ## [CAR Bench Challenge 技术方案](https://shiina18.github.io/llm/2026/08/09/car-bench/) _2026-08-09 · Shiina · Measure Zero_ 背景 CAR-bench：车载语音 Agent 基准。 CAR-bench 简介 Challenge 最终排行榜 技术报告原文 254 个任务、58 个工具、19 条车载策略 用户、环境、工具调用：评测器控制 每题独立运行 3 次；Pass³ 要求 3 次全过 一次错误调用、一次漏读前置状态、一次不合规回复：该 trial 归零 三类任务： 类型 额外困难 正确行为 Base 多轮工具操作、策略级联 先读状态，再按顺序执行 Hallucination 工具、参数或返回字段被移除 承认能力缺失；不盲调、不编造、不承诺完成 Disambiguation 用户没有给全取值 能内部消解则内部消解；需要时读取偏好；不猜 这个基准测的不是“偶尔能不能做成”，而是“能不能每次都守住边界”。 ## [When an Always Free server gets smaller](https://joeywang.github.io/posts/oracle-always-free-arm-downsize/) _2026-08-08 · Joey Wang · Joey_ What Oracle's smaller Ampere A1 Always Free limits mean for a self-hosted ARM server, and how I reshaped mine around a 12 GB memory budget. ## [The CI failures that moved every time you reran them](https://joeywang.github.io/posts/the-ci-failure-that-moved-every-rerun/) _2026-08-07 · Joey Wang · Joey_ How a moving set of Rails CI flakes exposed shared download state, Chrome/Selenium stale-node retries, and feature specs that asserted before the browser had settled. ## [How to Actually Learn ML in 2026: The Complete Beginner's Guide](https://aiagentssimplified.substack.com/p/how-to-actually-learn-ml-in-2026) _2026-08-06 · AI Agents Simplified · AI Agents Simplified_ A chapter-by-chapter reading guide to Hands-On Machine Learning with Scikit-Learn and PyTorch by Aurélien Géron (O'Reilly, 2025) ## [Agentic Supervised Fine-Tuning](https://cobusgreyling.substack.com/p/vibe-model-fine-tuning) _2026-08-06 · Cobus Greyling · Cobus Greyling on LLMs, NLU, NLP, chatbots & voicebots_ Agentic Supervised Fine-Tuning in this context means treating the whole process as an agentic, iterative experiment loop… ## [LAI #137: Where AI Engineering Is Going in 2026](https://learnaitogethernewsletter.substack.com/p/lai-137-where-ai-engineering-is-going) _2026-08-06 · Louis-François Bouchard · Learn AI Together_ Persistent memory, agent infrastructure, observability, and the skills that matter next. ## [Graph Engineering is an old idea reborn](https://cobusgreyling.substack.com/p/graph-engineering-is-an-old-idea) _2026-08-05 · Cobus Greyling · Cobus Greyling on LLMs, NLU, NLP, chatbots & voicebots_ From state machines and chatbots to agentic circuits…why structured autonomy keeps returning ## [I just opened access to our entire engineering team](https://learnaitogethernewsletter.substack.com/p/something-new-ive-been-building-towards) _2026-08-05 · Towards AI · Learn AI Together_ Submit your architecture, resume, or code: our engineers review it ## [The Noise Threshold](https://eugenevyborov.substack.com/p/the-noise-threshold) _2026-08-05 · Eugene Vyborov · Eugene Vyborov Blog (AI future, entrepreneurship)_ Why peak intelligence is indistinguishable from randomness - and what that means for minds, machines, and one experiment we are actually running ## [Install Manifest V3 Extension (Sponsored)](https://crawlproof.com/a/hsCPAegU8uyK) _2026-08-05 · **Sponsored**_ Get this Manifest V3 extension from the official TronBrowser Store listing. ## [From Prompt Engineering to Loop Engineering](https://www.anup.io/from-prompt-engineering-to-loop-engineering/) _2026-08-05 · Anup Jadhav · Anup Jadhav_ If you spend any time watching this space, you have watched this vocabulary turn over three times in about a year. Prompt engineering, then context engineering, then harness engineering, and now loop engineering. The usual telling makes that a sequence, each term replacing the last. They are layers, and each ## [读论文 - BM25 Wins at Scale](https://shiina18.github.io/llm/2026/08/05/bm25-wins-at-scale/) _2026-08-05 · Shiina · Measure Zero_ 做企业 RAG 时，最容易被默认接受的一套叙事是：语料大了要上 embedding，复杂问题要上 agent，结构复杂再上 graph RAG。 这篇论文把它们放到同一条、跨度约 450 倍的语料规模曲线上，结论很不客气：小语料时文件系统 Agent 略好；语料约过 1,000 万 token 后，BM25 反超并持续领先。 论文： BM25 Wins at Scale: A Scaling Study of Retrieval-Augmented Generation Paradigms ## [Open Weight Models Aren't the Enemy…Distillation, Trust & the Real Opportunity](https://cobusgreyling.substack.com/p/open-weight-models-arent-the-enemydistillation) _2026-08-04 · Cobus Greyling · Cobus Greyling on LLMs, NLU, NLP, chatbots & voicebots_ Three concerns and three truths about open-source models that the current uproar keeps missing. ## [7 Failures Nobody Warns You About Shipping An AI Agent](https://aiagentssimplified.substack.com/p/7-failures-nobody-warns-you-about) _2026-08-04 · AI Agents Simplified · AI Agents Simplified_ 7 failure modes that only show up after real users arrive and the architecture patterns that contain them. ## [The Prediction Trap](https://georgesiosi.substack.com/p/the-prediction-trap) _2026-08-04 · Siosi | The Digital Wayfinder · Siosi's Substack_ From Astral Portals to Algorithmic Swarms ## [FDE is the new PLG](https://cobusgreyling.substack.com/p/fde-is-the-new-plg) _2026-08-03 · Cobus Greyling · Cobus Greyling on LLMs, NLU, NLP, chatbots & voicebots_ For the better part of a decade, Product-Led Growth ruled… ## [Grok Voice Think Fast 2.0: Voice Is Ascending as the Primary UI](https://cobusgreyling.substack.com/p/grok-voice-think-fast-20-voice-is) _2026-07-31 · Cobus Greyling · Cobus Greyling on LLMs, NLU, NLP, chatbots & voicebots_ Speech-to-Speech, Customer Service architecture & why Voice-to-Voice is important ## [The Biggest AI Failures Start With Tiny Mistakes](https://aiagentssimplified.substack.com/p/the-biggest-ai-failures-start-with) _2026-07-30 · AI Agents Simplified · AI Agents Simplified_ What a neural network attack reveals about building more resilient AI systems. ## [How to choose AI agent platforms to run agents in production](https://eugenevyborov.substack.com/p/how-to-choose-ai-agent-platforms) _2026-07-30 · Eugene Vyborov · Eugene Vyborov Blog (AI future, entrepreneurship)_ Half the names are hype, half are genuinely good, and from the outside it's almost impossible to tell which is which. So I stopped answering it one message at a time and wrote the guide. ## [LAI #136: Build Faster With Agents, Debug Their Failures, and Evaluate Them More Reliably](https://learnaitogethernewsletter.substack.com/p/lai-136-build-faster-with-agents) _2026-07-30 · Louis-François Bouchard · Learn AI Together_ Better agents, better evaluation, and fewer production surprises. ## [You Forgot to Exhale](https://georgesiosi.substack.com/p/you-forgot-to-exhale) _2026-07-29 · Siosi | The Digital Wayfinder · Siosi's Substack_ How Tokenmaxxing Is Irregular Breathing & What It's Costing You ## [The power of speed…a lesson from Cerebras](https://cobusgreyling.substack.com/p/the-power-of-speeda-lesson-from-cerebras) _2026-07-29 · Cobus Greyling · Cobus Greyling on LLMs, NLU, NLP, chatbots & voicebots_ Speed only matters once the product becomes useful ## [How to Become an AI Engineer: The Complete Beginner's Guide](https://aiagentssimplified.substack.com/p/how-to-become-an-ai-engineer-the) _2026-07-28 · AI Agents Simplified · AI Agents Simplified_ A chapter-by-chapter reading guide to AI Engineering: Building Applications with Foundation Models by Chip Huyen (O'Reilly, 2025) ## [10 communication rules nobody teaches you at college](https://mensurdurakovic.com/communication-rules-college-never-teaches-you) _2026-07-27 · Mensur Duraković · Mensur Duraković_ I once found out about a decision in a very ugly way. I exported a spreadsheet to CSV, imported it into a project, and handed it to my AI agent to build me a slide out of the data. The slide came back wrong. Not broken, just wrong: the numbers on it did not match the plan our whole group had agreed on weeks earlier. My first thought was that I had messed up the import. My second thought was that… ## [framesmith 1.10: approvals that can be proven wrong](https://victorvelazquez.dev/blog/framesmith-1-10-approvals-that-can-be-wrong/) _2026-07-27 · Victor Velázquez_ A design gate is only as good as its ability to notice reality drifted. framesmith 1.10 binds approvals to a content hash and turns drift into a CI check. ## [Pax Silica](https://georgesiosi.substack.com/p/pax-silica) _2026-07-26 · Siosi | The Digital Wayfinder · Siosi's Substack_ How to read this AI "peace initiative" as a digital wayfinder and conscious technologist. Learn the difference between Silicon and Silica. ## [The most valuable spec-kit command is the optional one](https://victorvelazquez.dev/blog/the-most-valuable-spec-kit-command-is-the-optional-one/) _2026-07-24 · Victor Velázquez_ After running spec-driven development across several projects, the step that pays for the whole workflow is the one the docs mark optional: clarify. ## [The AI Lesson Hidden Inside a Heart Failure Prediction Paper](https://aiagentssimplified.substack.com/p/the-ai-lesson-hidden-inside-a-heart) _2026-07-23 · Arian Jafari · AI Agents Simplified_ The future of AI will not be defined by systems that make fewer mistakes, but by systems that understand which mistakes matter. ## [LAI #135: The Useful Part of Graph Engineering Is Not the Graph](https://learnaitogethernewsletter.substack.com/p/lai-135-the-useful-part-of-graph) _2026-07-23 · Louis-François Bouchard · Learn AI Together_ When JSON hurts reasoning, Anthropic wins our dream-workplace poll, and bad statistics make chocolate look like a diet plan ## [Weights vs Prompts](https://georgesiosi.substack.com/p/weights-vs-prompts) _2026-07-22 · Siosi | The Digital Wayfinder · Siosi's Substack_ Weekly wisdom for July 22, 2026