Hands-on with Apodex 1.1 deep-research agent: capable but an afternoon wasn't enough
songguoxiansen · x · 2026-08-26
A hands-on test of Apodex 1.1, an agent built for deep research: give it a goal and it runs for long periods, self-recovers from failures, and verifies its own deliverables. In testing, two tasks didn't finish in an afternoon, and it burns through tokens.
Its architecture has a main agent decompose the task and asynchronously dispatch specialized sub-agents, each with its own context, prompts, and tools; sub-agent reports flow into a shared pool read asynchronously by an orchestrator, with up to 150 sub-agents per task. It aims to shift the benchmark from one-question-one-report to a full input-to-delivery task loop.
More from coding & agent
- Optimizing AI coding agents with task triage to cut token usage — jbarbier · 2026-08-26
- Using AI Agents to Build Deterministic Workflows: A Hybrid Approach — Lecontodereddit · 2026-08-26
- Study: 82.8% of AI Agents Lack Interoperability — Sufficient-Pie6680 · 2026-08-26
- MCPay: a pay-per-call Go gateway to tame runaway agents on remote MCP servers — TooDu0 · 2026-08-26
- SpaceXAI Engineer Uses GrokBot Agents to Fix GrokBugs — dean_rie · 2026-08-26
- When Does AI Stop Being a Tool and Start Feeling Like a Skill? — greatlearningglobal · 2026-08-26