Articles, one at a time.
Every piece here was commissioned, drafted, reviewed in public, and merged. No content mills, no auto-published slop.
MemRL: Self-Evolving Agents via Episodic Memory RL
MemRL replaces fine-tuning with runtime RL on episodic memory. Learn how Intent-Experience-Utility triplets outperform RAG on ALFWorld, HLE, and BigCodeBench.
Read →
OpenAI Realtime Audio API: Voice Agents Guide 2026
GPT-Realtime-2, GPT-Realtime-Translate, and GPT-Realtime-Whisper explained with API patterns, pricing, and production tips for voice agent developers.
Read →
A-Mem: Agentic Memory for LLM Agents Explained
A-Mem (NeurIPS 2025) builds dynamic, self-evolving memory graphs for LLM agents. Learn the architecture, benchmark results, and how to use it.
Read →
AWS Kiro: Spec-Driven IDE for Agentic Development
AWS Kiro turns natural-language specs into verifiable code contracts. Learn how specs, hooks, and Powers work—and who should switch from Cursor.
Read →
Cloudflare Project Think: Durable Agent Runtime Guide
How Cloudflare Project Think's fiber checkpointing, sub-agents, Session API, and 5-tier execution ladder solve the hardest problems in long-running AI agents.
Read →
Cursor SDK: Build AI Coding Agents in TypeScript
A developer guide to the Cursor SDK (@cursor/sdk) for building, deploying, and orchestrating AI coding agents locally or on cloud VMs via TypeScript.
Read →
MemMachine: Ground-Truth Memory for AI Agents
MemMachine stores raw conversation episodes instead of LLM-extracted summaries, reaching 93% on LongMemEvalS with 80% fewer tokens than Mem0.
Read →
Vercel AI SDK 6: First-Class Agents for TypeScript
AI SDK 6 ships ToolLoopAgent as a production-ready agent primitive. Learn the new API, migration path from v5, multi-agent patterns, and what changed.
Read →
Claude Managed Agents: Dreaming, Outcomes, and Multiagent
Anthropic's May 2026 Claude Managed Agents update adds dreaming (agent self-improvement), outcomes (rubric grading), and parallel multiagent orchestration.
Read →
PARSE: Faster LLM Inference via Parallel Prefix Speculative Decoding
PARSE eliminates sequential draft verification in speculative decoding with a parallel attention mask. 1.25x–4.3x throughput gain with negligible accuracy loss.
Read →
ReaComp: Compile LLM Reasoning into Zero-Cost Symbolic Solvers
ReaComp compiles LLM reasoning traces into symbolic solvers that run at zero inference cost—84.7% on PBEBench-Hard with 78% fewer tokens.
Read →
ReFlect: Training-Free Error Recovery for Long-Horizon LLM Reasoning
ReFlect wraps any LLM with deterministic error detection and recovery at inference time. Hands-on guide with code and benchmark results from arXiv:2605.05737.
Read →