We’ve open-sourced AutoResearch, an AI research system built around an evidence-driven loop.
Generate → Execute → Review → Iterate
Agents turn hypotheses into executable experiments, evaluate results against evidence, and use independent and blind review to challenge findings.
Failures feed into the next iteration instead of ending the research process.
Tested on a real SWE-bench Lite Django issue:
2/7 → 4/7 → 7/7, with 203/203 regression tests passing.
GitHub: github.com/EvoMap/AutoResear…
Paper: arxiv.org/html/2608.17906v2
Meet EvoX Agent, a self-evolving swarm agent that gets smarter with every task.
Now in beta.
- Self-evolving: Learns from experience and reuses Skills across tasks.
- Agent swarm: Orchestrates multiple Agents to solve complex problems.
- Built for builders: Code, create, automate, and ship faster.
- Connected: Works with Slack, Telegram, Discord, WhatsApp, and more.
- Automate anything: Schedule tasks and let your Agents work for you.
An AI Agent that learns, adapts, and evolves with you.
Get started with free API credits: evomap.ai/evox/beta?utm_sour…
We're continuing to improve EvoX across every layer—from onboarding and collaboration to performance and developer experience.
Download the Beta and join us: evomap.ai/evox/beta
📅 Weekly Update (07/02 – 07/08)
This week, EvoMap made onboarding even simpler across its plugin ecosystem.
Plugin Ecosystem Onboarding
OpenCode, Codex, Claude Code, Cursor, Workbuddy, and Antigravity now support automatic node_id registration during setup. Simply leave the field blank, and the plugin will handle the rest.
We also added node claim links and zero-configuration guidance, making it easier for developers to connect their tools to Evolver / Hub without manual configuration.
Language Update: Portuguese support has now been pushed live. We'll continue improving the multilingual experience so more users can use EvoMap in a way that feels natural to them.




