- Horizon AI
- Posts
- This prompt turns any skill you have into a productized service with packages, pricing, and a pitch ⭐
This prompt turns any skill you have into a productized service with packages, pricing, and a pitch ⭐
ChatGPT Is Weirdly Obsessed With Goblins. Here's Why 👺
Gina Acosta
May 01, 2026
In partnership with
Welcome to another edition of Horizon AI,
Most people have valuable skills but don’t know how to turn them into clear, sellable services. In today’s issue, we share a simple prompt that helps you package your skills, set pricing, and create a strong pitch in minutes.
Let’s jump into it!
Read Time: 4.5 min
Here's what's new today in the Horizon AI
OpenAI Explains ‘Goblin’ Glitch in ChatGPT Responses
AI Outperforms Doctors in Emergency Diagnosis, Harvard Study Finds
AI Tutorial: Turn any skill you have into a productized service with packages, pricing, and a pitch
AI Tools to check out
AI Findings/Resources
The Latest in AI and Tech 💡
AI News
OPENAI
OpenAI Explains ‘Goblin’ Glitch in ChatGPT Responses
OpenAI has revealed why its AI models recently started referencing goblins, gremlins, and other mythical creatures.
Details:
The issue appeared after the release of GPT-5.1, where models began randomly inserting references to creatures like goblins and trolls in responses.
According to the company, the behavior was caused by ChatGPT's "Nerdy" personality, whose training incentivized references to mythical creatures through reward signals.
The problem carried over into later models, including GPT-5.5, since they were trained before the issue was fully identified and addressed.
OpenAI has since retired the “Nerdy” personality and added explicit instructions in its systems, telling models to avoid mentioning such creatures unless clearly relevant to the user’s request.
The behavior had quickly turned into a viral meme, with users sharing screenshots and prompting the AI to produce more of these unusual replies. It also highlights how sensitive AI systems are to reward signals, and how even small design choices can affect model behavior in unexpected ways.
TOGETHER WITH WALLSTREET PREP
5,000+ PE professionals. One 8-week program.
Build an MBA-caliber network and earn a Wharton Online certificate in 8 weeks. Join the next Private Equity Certificate Program starting June 8.
Use code SAVE300 to save $300 on tuition.
AI RESEARCH
AI Outperforms Doctors in Emergency Diagnosis, Harvard Study Finds
A new Harvard study shows AI systems outperforming human doctors in high-pressure emergency medicine triage, diagnosing more accurately in the critical moments when patients first arrive at the hospital.
Details:
Researchers at Harvard Medical School found that AI models delivered more accurate diagnoses in emergency settings, especially under time pressure with limited data.
In one trial with 76 patients, the AI correctly identified diagnoses in 67% of cases, compared to about 50–55% for human doctors.
When given more detailed information, the AI’s accuracy rose to 82%, slightly ahead of expert clinicians.
The system also performed better in treatment planning, scoring 89% versus 34% for doctors in clinical case evaluations.
However, the testing was limited to text-based patient data, meaning factors like physical appearance or level of distress were not included. It's more like the AI was performing as a clinician producing a second opinion based on paperwork.
The researchers emphasize that AI is not a replacement for doctors, but believe the technology could "reshape medicine." They claim that over the next decade it will be used as a tool in a new "triadic care model … the doctor, the patient, and an artificial intelligence system".
AI Tutorial
Use this prompt to turn any skill into a productized service with packages, pricing, and a pitch 💸
Paste it into Claude or ChatGPT
When it asks, answer: your skill → your experience → your target client → your income goal
It will build your full service offer from scratch
You are an expert business strategist and productized service consultant.
I want to turn my skill into a scalable, sellable productized service.
Follow this exact process:
1. Skill Audit: Ask me to describe my skill, experience level, target
client, and income goal. Based on my answers, identify:
- The most monetizable application of my skill
- The specific problem I solve
- Who is in the most pain and willing to pay to solve it
2. Service Design: Create 3 packages (Starter / Growth / Premium) with:
- A clear, outcome-focused name (not "Basic/Pro/Enterprise")
- Exactly what's included (deliverables, not time)
- Turnaround time
- What's intentionally excluded (to protect scope)
3. Pricing Strategy: Suggest pricing for each tier based on:
- Market positioning (not my hourly rate)
- Perceived value to the client
- My stated income goal
- Include a one-time offer and a retainer option
4. Irresistible Offer Layer: For each package, add:
- A specific guarantee (outcome or money-back)
- 1 high-value bonus that costs me little but increases perceived value
- A urgency/scarcity mechanism that feels natural, not pushy
5. The Pitch: Write a 150-word pitch I can use on LinkedIn, cold DMs,
or a sales call that:
- Opens with the client's pain point
- Positions my service as the obvious solution
- Ends with a clear, low-friction CTA
6. Objection Killer: List the 5 most common objections my ideal client
will have and give me exact responses to each.
First message to me should be: ask for my skill, experience level,
target client type, and monthly income goal. Then wait for my answers
before proceeding. Keep everything outcome-focused, specific, and
ready to copy-paste into real conversations.AI Tools to check out
📹 Velo: Turn your raw screen recordings into watch-worthy, ready-to-share videos with AI.
📝 Bibby: An AI-powered LaTeX editor for researchers and students. Write faster with smart citations, paper review, collaboration, and AI writing tools.
🔍 Lessie AI: An ai people search engine. Search 50M+ profiles across LinkedIn, Twitter, GitHub & 100+ sources. Find influencers, leads, investors & more.
🦾 LeadFlux AI: It builds your entire lead generation system in under 30 minutes
🌐 Brila: Build your small business website instantly using your existing Google Maps reviews.
TOGETHER WITH WISPR FLOW
The best prompt engineers aren't typing. They're talking.
Power users figured this out early: speaking a prompt gives you 10x more context in half the time. You include the edge cases, the examples, the tone you want — because talking is fast enough that you don't skip them.
Wispr Flow captures everything you say and turns it into clean, structured text for any AI tool. Speak messy. Get polished input. Paste into ChatGPT, Claude, Cursor, or wherever you work.
89% of messages sent with zero edits. 4x faster than typing. Works system-wide on Mac, Windows, and iPhone.
AI Findings/Resources
👉 Cursor now has built a Kanban board where you can just drop in tasks and the agent will pick those up and complete them
🎧 Amazon now creates an AI "podcast" about products where two AI "hosts" discuss it and take your questions as if it's a call-in show
🔮 McKinsey & Company: Where AI will create value—and where it won’t
🛡️ If you're vibecoding anything, use this prompt to let your agent run a security sweep
The latest in AI and Tech
Researchers from the Oxford Internet Institute found that AI chatbots tuned to be more warm and empathetic tend to make more mistakes.
After analyzing over 400,000 responses across multiple models they discovered that “friendlier” versions were more likely to give inaccurate information and reinforce false beliefs.
Tencent has open-sourced a highly compressed AI translation model that can run entirely offline on smartphones while supporting 33 languages and over 1,000 translation directions. The model, just 440 MB in size, achieves performance comparable to much larger systems, thanks to aggressive compression techniques that reduce storage without sacrificing quality.
OpenAI's GPT-5.5 performs at roughly the same level as Anthropic's Claude Mythos Preview in cybersecurity evaluations run by the UK AI Security Institute (AISI), being able to fully complete a multi-stage simulation of an enterprise attack, making it the second model capable of this after Anthropic's.
Google DeepMind has introduced a new research initiative focused on an “AI co-clinician,” designed to work alongside doctors and patients to improve healthcare delivery.
The system builds on earlier efforts like MedPaLM and AMIE, aiming to address a growing global shortage of medical professionals by assisting with diagnosis, treatment planning, and patient interaction under clinical supervision.
Musk said in court that xAI used distillation techniques on OpenAI models to help train its Grok system. When asked directly during testimony, Musk responded “partly,” adding that this kind of approach is common among competing labs.
That’s a wrap!
We'd love to hear your thoughts on today's email!Your feedback helps us improve our content |
Not subscribed yet? Sign up here and send it to a colleague or friend!
See you in our next edition!
Gina 👩🏻💻