One prompt vs 200: pushing Astra to model the WTC in Blender with pure prompting
Sprytex · reddit · 2026-09-11
User Sprytex stress-tested Astra (xhigh) by rebuilding the NYC Twin Towers in Blender using only prompts, never touching the .blend file. A single 1-shot prompt produced poor results, but after 10 hours and 200 human-guided prompts, the model could build nearly anything, correctly inferring intent even when the author lacked the vocabulary. Astra's own subagent critics were far less effective than human guidance. The author concludes Astra isn't AGI yet — his bar is producing the right result from one prompt — but expects future models to beat his 200-prompt result with a single prompt.
More from coding & agent
- Cursor ships Annotate in design mode: give agents visual prompts by editing and drawing on the page — soleio · 2026-10-28
- One-person companies are about the org chart, not headcount: humans judge, agents execute — ciguleva · 2026-10-28
- OATs: open-source toolkit turns everyday objects into personality-bearing agents — Stefania_druga · 2026-10-28
- Dev: local coding dropped from 100% to 5% in a year as cloud agents take over — bytebot · 2026-10-28
- Open-sourced 501(c)(FREE): agent-built free back office for US nonprofits — gralston · 2026-10-28
- Grok Bot 1.0.0 early build reveals Shared Bots, service accounts and team workflows — XFreeze · 2026-10-28