Developer reverse-engineers Codex ghost text to simulate his own thinking for autonomous agents
urbantechgoods · reddit · 2026-11-12
A Reddit developer shares how he turned Codex's predictive ghost text into something more powerful: instead of just completing what he'd type, it suggests the direction he wants to steer the conversation.
- He instructed his agent to reverse-engineer how the predictive text works, and to keep learning how he thinks from past and future interactions
- The goal is letting the agent iterate on its work without his input, using his simulated preferences as an internal reviewer
- The skill was just written and is still being fine-tuned; code and details are on his blog, with feedback welcome
More from coding & agent
- One prompt, Blender MCP: Qwen4-Max builds a ramen stall at ~1/4 of DeepSeek's token cost — teortaxesTex · 2026-11-12
- 1.7B model shown driving browser and writing HTML autonomously, claimed to hit 4o-level benchmarks — JosephJacks_ · 2026-11-12
- Hackathon Project Uses Hermes Agent to Turn Prompts into 3D-Printable CAD Designs — alexcovo_eth · 2026-11-12
- Baidu's ERNIE 5.5 Flash writes its own code to build editable CAD skyscraper model — heyshrutimishra · 2026-11-12
- Gamma's UGC pages become parasite SEO target, risking domain-wide Google penalties — gefei55 · 2026-11-12
- A neat prompt to make your agent expose its hidden assumptions about you — keyanzhang · 2026-11-12