Anthropic just published a great blueprint on building always-on agents. Here is what happens when you build those five primitives locally in agy-hq—and why background runs used just 0.41% of quota across 1.51B real tokens.
Sharing agy-hq with the community: a local-first workspace and zero-dependency Go helper that makes Google Antigravity work like a responsive personal assistant across coding, notes, inbox triage, and daily admin—routing quick turns to Flash and deep background tasks to models like Gemini 4 Carbon.
Google Antigravity is built for coding, but I wanted one workspace on my phone and laptop to juggle code, research, and daily admin at the same time. Here is why prompt rules were not enough, and how a small Go helper keeps tasks from colliding.
Two tuning stages on eight repeated development cases: raise the response cap from 8K to 32K, then compare medium with xhigh. More repairs finished, waiting increased, and confirmation remains unrun.
A controlled Qwen3.8-27B Q4_K_M and OpenCode comparison: MTP raised generation throughput by 56% and cut successful task time by 20–40%. Paired coding results explain why throughput alone is not enough.
Deploy Qwen3.8-27B Q4_K_M on an RTX 3090 24GB and connect it to OpenCode through llama.cpp. Measured 128K capacity, 20.9–36.4 tokens/s generation, and four coding tasks with successes, timeouts, and review findings.
Self-host OpenCode Web IDE with authentication and HTTP/WebSocket reverse proxy. Access one coding-agent workspace from phone and laptop, with model integration examples and LAN or VPS setup steps.
With your permission, we use Google Analytics and cookies to understand traffic sources, page engagement and download-link clicks. You can read everything without analytics. Privacy notice