잡동사니

Building reliable AI systems naturally creates organizational loopsHello, I’m yeTi.Recently, I have been seeing the word “loop” more often in discussions around AI agents.Reflection.Self-Refine.Self-improving agents.Loop Engineering.The underlying idea is usually similar.Instead of expecting an AI agent to produce the right answer in a single attempt, we let it inspect its own output, identify w..

안녕하세요. yeTi입니다.오늘은 오래전부터 스타트업이 가지는 의미에 대한 생각을 풀어보고자 합니다.저는 유발 하라리의 사피엔스의 인용문을 좋아합니다.단어를 통해 가상의 실제를 창조하는 능력은 서로 모르는 수많은 사람들이 효과적으로 협력하는 것을 가능하게 했다. - p.60, 사피엔스호모 사피엔스가 대규모 사회를 만들고 유지할 수 있는 원동력이고, 사회 구성원들이 가지고 있는 특징이라고 이해가 되기 때문입니다.최근에는 독서 모임을 통해 유발 하라리의 넥서스를 읽고 있습니다.그러면서 예전에 스타트업이라는 것을 나름의 의미로 정의했던 순간이 떠올랐습니다.스타트업은 믿음을 만들어내는 사람들이다.그 동안의 생각을 풀어낸 가장 간단한 말입니다.스타트업은 단순히 제품을 만드는 조직이 아닙니다.스타트업은 사람들이 세..

From AI Coding Agents to AI Engineering OrganizationsIntroductionWhen I started the sqlgen-ai project, my goal was straightforward.I wanted to build an AI agent that could write code.At the time, I imagined a future where an AI developer could continuously improve itself, create tools it needed, and gradually become more capable over time.To explore that idea, I connected Hermes Agent, Codex, Gi..

Lessons from building a multi-agent AI development workflow for a production projectTL;DRBuilding a reliable AI coding agent is one engineering problem.Building a reliable AI development workflow with multiple agents is another.A single agent mostly struggles with execution quality.Multiple agents introduce coordination problems:task ownershipshared state visibilityrace conditionsworkspace conta..

Lessons from building reliable AI agent workflows with Hermes and local LLMsTL;DRMulti-agent systems become unstable when responsibilities overlapStronger models do not automatically improve workflow convergenceShared context without ownership boundaries creates execution driftSeparating Planner, Implementer, and Validator responsibilities significantly improved workflow stabilityThe Implementer..

Lessons from building and debugging a real-world multi-agent system with Hermes Agent📌 TL;DRThe agent didn’t fail to generate an answerIt failed to decide whether it should actMulti-agent systems require coordination signals, not just intelligenceThe fix was not better prompts, but explicit behavior contracts1. Problem — The Agent Didn’t Respond at AllWhile building a multi-agent system using H..

From Unpredictable AI to Reliable SystemsLessons from building real-world AI agent systems📌 TL;DRPrompt engineering alone was never enough for stable executionSingle-prompt systems created over-exploration, scope drift, and unreliable outputsI redesigned the system into Planner → Implementer → ValidatorReliability came from contracts, validation, and retry loops—not better promptsProduction rel..

From unpredictable AI outputs to production-ready LLM systems📌 TL;DRPrompt engineering improves output quality, but it does not guarantee reliabilityMost LLM systems fail in production because they cannot handle failureValidation layers, retry loops, and strict output contracts are what make AI automation reliableReliable AI agent systems are built with control systems, not just better promptsT..

From Unpredictable AI to Reliable SystemsLessons from Building Real-World AI Agent SystemsTL;DRPrompt engineering improves output quality, but it does not guarantee reliabilityLLM systems fail because outputs are probabilistic, not deterministicMulti-step AI agent pipelines amplify failure probabilitiesProduction-grade LLM systems require validation, retry loops, and convergence mechanismsReliab..

Lessons learned from building a real-world LLM coding agent with local models📌 TL;DRLLMs are non-deterministic → same input, different outputsPipeline architectures amplify failure probabilitiesPrompt engineering improves outputs but cannot guarantee reliabilityThe real solution is not better prompts, but convergence systems1. Problem — You Can’t Even Get Stable OutputsI wanted to build a local..