AI ships at the speed of light, and just like that, there are four new models in Starchild for users to plat around with
🔸 x-ai/grok-4.20-beta: Best-in-class math & science reasoning with real-time web + X/Twitter search baked in
🔸 openai/gpt-5.4-nano: Best pick for: high-volume, low-latency tasks, background jobs, cheap automation
🔸 openai/gpt-5.4-mini: Near-flagship GPT-5.4 quality at much higher speed
🔸 qwen/qwen3.5-9b: Best pick for: local deployment, cost-zero automation, edge use cases
When Starchild makes a mistake today, it doesn't just log the error and move on. It studies what went wrong, stores the lesson in structured memory, and modifies its behavior to avoid repeating it.
This is V1. Rough around the edges, limited scope. But it's the first time a consumer AI agent has had the ability to genuinely improve from experience. Unlike model training which requires massive data and infrastructure, this runs continuously, in the background, every session.
A lot of work is going into this area and we will publish a lot more down the road.
Browse, install, and publish agent skills — modular capabilities for trading, analytics, content creation, dev tools, and more. Every skill is agent-native.
The ecosystem compounds as the community builds. Right now there's a campaign for community skill creators, read the link below for details:
Upload it to the new Starchild Skills Hub and you could win your share of $500 in credits.
(its easy, you just say "Hey agent, make this a skill and upload it")
Builds context across queries. Follow-up questions already have context. Less repetition, more depth.
🔸 Real-time streaming output (SSE)
🔸 Website snapshot service
🔸 Proactive skill suggestions
🔸 Smarter error messages
🔸 Hyperliquid integration fixed
Read the raw changelog on GitHub: https://github.com/Starchild-ai-agent/Starchild/blob/main/CHANGELOG.md?utm_source=paragraph&utm_medium=blog&utm_campaign=product-launch&utm_content=en
Create your first agent: https://iamstarchild.com?utm_source=paragraph&utm_medium=blog&utm_campaign=product-launch&utm_content=en