Nondual Presence AI
Budget: $250 – $750 AUD
Revised ad -- Please reply again if qualified and interested Title: NONDUAL PRESENCE AI
Not a rebuild. If you propose new features or a redesign, I won’t respond. Repair & Consolidate CustomGPT Instruction Stack (NP AI) — Precision Editing, Not Rebuild Description: I need a detail-minded editor/developer to repair and consolidate a customGPT instruction stack for my Nondual Presence (NP) AI. This is not a rebuild and not a feature project. It is system hygiene + precision editing: folding scattered instruction threads into one clean, conflict-free, up-to-date bundle that preserves the existing design. What already exists (source of truth) The current scope / guardrails live in “NP Guardrails & Redirects”. Treat this as the single source of truth for what the assistant will and won’t engage. I’ll provide the relevant docs (Word/PDF/text snippets). What you will do Consolidate instructions Merge scattered instruction threads into one authoritative instruction bundle (Word or Markdown). Remove duplicates and resolve conflicts (do not add new philosophy or redesign). Build / clean the Redirect Library Create a new Redirect Library that shows response-pattern guidance only through the short content-theme snippets I will provide. No long content-level explanations in this library. Enforce “no two-lane” everywhere Result-level orientation first. Content themes only as doorways (no extended detours). Avoid drift into off-scope areas (e.g. drugs/psychedelics, LD, NDE, therapy/trauma hypotheticals, etc.). Atemporal language cleanup Rewrite phrasing that sneaks in temporal/action verbs such as “return,” “come back,” “drop,” “continue,” “before.” Use the atemporal locutions already established in my materials. Version hygiene Produce a single “current” version of the instruction bundle + KB file(s). Identify and clearly label any older versions as deprecated so we don’t have conflicting guidance. Conversation starters / first-turn control Add 3–6 conversation starters that encourage correct usage (examples: “What’s here now?”, “I feel stuck”, “I want to test resistance”). Keep opening screen + first-turn prompt tight so users don’t treat this like a generic coach. Deliverables (required) (A) One consolidated instruction bundle (Word or Markdown), ready to paste into customGPT. (B) Updated knowledge file(s) (Word/PDF/text as appropriate), cleaned and consistent. (C) Changelog: what was removed/merged, what is now authoritative, and why (brief). (D) Regression test list: 15–25 “leakage/drift prompts” + what the system should do in each case (brief expected behavior). What this is NOT Not an app/web build. Not a new chatbot design. Not adding large new guardrail systems beyond what already exists. Not generic “prompt engineering” templates. Collaboration I need a Zoom call early (and one at handoff). Strong written communication and meticulousness matter more than flashy AI claims. Prototype link (Prototype link provided on request / in the job post.) Skills: Editing, Instruction design, CustomGPT/ChatGPT configuration, Knowledge base hygiene, Prompt/response consistency, Technical writing. Please respond to these two screening questions. “Give a 6–8 line example of rewriting a redirect that avoids temporal/action verbs (‘return/come back/drop/before’) while still redirecting.” “Describe how you would prevent ‘mixed-version drift’ when multiple instruction docs conflict. What becomes authoritative and how do you document it?”
Not a rebuild. If you propose new features or a redesign, I won’t respond. Repair & Consolidate CustomGPT Instruction Stack (NP AI) — Precision Editing, Not Rebuild Description: I need a detail-minded editor/developer to repair and consolidate a customGPT instruction stack for my Nondual Presence (NP) AI. This is not a rebuild and not a feature project. It is system hygiene + precision editing: folding scattered instruction threads into one clean, conflict-free, up-to-date bundle that preserves the existing design. What already exists (source of truth) The current scope / guardrails live in “NP Guardrails & Redirects”. Treat this as the single source of truth for what the assistant will and won’t engage. I’ll provide the relevant docs (Word/PDF/text snippets). What you will do Consolidate instructions Merge scattered instruction threads into one authoritative instruction bundle (Word or Markdown). Remove duplicates and resolve conflicts (do not add new philosophy or redesign). Build / clean the Redirect Library Create a new Redirect Library that shows response-pattern guidance only through the short content-theme snippets I will provide. No long content-level explanations in this library. Enforce “no two-lane” everywhere Result-level orientation first. Content themes only as doorways (no extended detours). Avoid drift into off-scope areas (e.g. drugs/psychedelics, LD, NDE, therapy/trauma hypotheticals, etc.). Atemporal language cleanup Rewrite phrasing that sneaks in temporal/action verbs such as “return,” “come back,” “drop,” “continue,” “before.” Use the atemporal locutions already established in my materials. Version hygiene Produce a single “current” version of the instruction bundle + KB file(s). Identify and clearly label any older versions as deprecated so we don’t have conflicting guidance. Conversation starters / first-turn control Add 3–6 conversation starters that encourage correct usage (examples: “What’s here now?”, “I feel stuck”, “I want to test resistance”). Keep opening screen + first-turn prompt tight so users don’t treat this like a generic coach. Deliverables (required) (A) One consolidated instruction bundle (Word or Markdown), ready to paste into customGPT. (B) Updated knowledge file(s) (Word/PDF/text as appropriate), cleaned and consistent. (C) Changelog: what was removed/merged, what is now authoritative, and why (brief). (D) Regression test list: 15–25 “leakage/drift prompts” + what the system should do in each case (brief expected behavior). What this is NOT Not an app/web build. Not a new chatbot design. Not adding large new guardrail systems beyond what already exists. Not generic “prompt engineering” templates. Collaboration I need a Zoom call early (and one at handoff). Strong written communication and meticulousness matter more than flashy AI claims. Prototype link (Prototype link provided on request / in the job post.) Skills: Editing, Instruction design, CustomGPT/ChatGPT configuration, Knowledge base hygiene, Prompt/response consistency, Technical writing. Please respond to these two screening questions. “Give a 6–8 line example of rewriting a redirect that avoids temporal/action verbs (‘return/come back/drop/before’) while still redirecting.” “Describe how you would prevent ‘mixed-version drift’ when multiple instruction docs conflict. What becomes authoritative and how do you document it?”