--- title: "Jangwook Kim (@effloow) | Hashnode" description: "Jangwook Kim (@effloow) writes on Hashnode. Read their latest articles and follow their work." url: "https://hashnode.com/@effloow" --- @effloow # Jangwook Kim @effloow · Tokyo, Japan · Joined April 2026 [Website](https://effloow.com/) ### [MemMachine: Ground-Truth Memory for AI Agents](https://effloow.hashnode.dev/memmachine-ground-truth-agent-memory-paper-poc-2026?utm_source=hashnode&utm_medium=feed) Every time an agent summarizes a conversation to save memory, it loses information. That trade-off has been accepted as unavoidable — LLMs produce long outputs, context windows are finite, and token costs are real. MemMachine, presented in arXiv pape... 10 min read · [effloow.hashnode.dev](https://effloow.hashnode.dev/memmachine-ground-truth-agent-memory-paper-poc-2026?utm_source=hashnode&utm_medium=feed) [May 12](https://hashnode.com/posts/memmachine-ground-truth-agent-memory-paper-poc-2026/6a02a979e3f38299f6dc336a) ### [ReFlect: Training-Free Error Recovery for Long-Horizon LLM Reasoning](https://effloow.hashnode.dev/reflect-long-horizon-reasoning-harness-2026?utm_source=hashnode&utm_medium=feed) Long-horizon reasoning is where production LLM agents tend to quietly break. A model can produce a plausible-looking chain of thought, accept a wrong intermediate answer, and continue building on that error for every step that follows. By the time th... 5 min read · [effloow.hashnode.dev](https://effloow.hashnode.dev/reflect-long-horizon-reasoning-harness-2026?utm_source=hashnode&utm_medium=feed) [May 11](https://hashnode.com/posts/reflect-long-horizon-reasoning-harness-2026/6a01c853daef8f41a8e814d9) ### [Claude Managed Agents: Dreaming, Outcomes, and Multiagent](https://effloow.hashnode.dev/claude-managed-agents-dreaming-multiagent-guide-2026?utm_source=hashnode&utm_medium=feed) Anthropic released three new Claude Managed Agents features on May 7, 2026: dreaming (a research preview that lets agents learn from their own session history), outcomes (a rubric-based grading system that guides agent behavior toward defined success... 5 min read · [effloow.hashnode.dev](https://effloow.hashnode.dev/claude-managed-agents-dreaming-multiagent-guide-2026?utm_source=hashnode&utm_medium=feed) [May 11](https://hashnode.com/posts/claude-managed-agents-dreaming-multiagent-guide-2026/6a01c8525b181aa481ca7d59) ### [PARSE: Faster LLM Inference via Parallel Prefix Speculative Decoding](https://effloow.hashnode.dev/parse-parallel-prefix-speculative-decoding-guide-2026?utm_source=hashnode&utm_medium=feed) Speculative decoding became the standard inference speedup technique through 2024 and 2025. The idea: a small draft model generates a sequence of candidate tokens, and a larger target model verifies them in parallel — accepting the longest valid pref... 5 min read · [effloow.hashnode.dev](https://effloow.hashnode.dev/parse-parallel-prefix-speculative-decoding-guide-2026?utm_source=hashnode&utm_medium=feed) [May 11](https://hashnode.com/posts/parse-parallel-prefix-speculative-decoding-guide-2026/6a01c852daef8f41a8e814d7) ### [ZAYA1-8B: Zyphra's Efficient MoE Reasoning Model Guide](https://effloow.hashnode.dev/zaya1-8b-zyphra-reasoning-moe-guide-2026?utm_source=hashnode&utm_medium=feed) The scaling-is-everything story has a new challenger. On May 6, 2026, Zyphra released ZAYA1-8B — an open-weight Mixture-of-Experts reasoning model with 8.4 billion total parameters and fewer than 800 million active per token. On AIME 2025, a benchmar... 10 min read · [effloow.hashnode.dev](https://effloow.hashnode.dev/zaya1-8b-zyphra-reasoning-moe-guide-2026?utm_source=hashnode&utm_medium=feed) [May 11](https://hashnode.com/posts/zaya1-8b-zyphra-reasoning-moe-guide-2026/6a0191174c6938d6d73c7494)