Summers are stressful.
It shouldn’t be that way.
But the lack of a consistent summer camp plan takes a wrecking ball to my schedule and routines.
After all, the pace of work slows down.
Plus, the process of finding a camp is hard as heck. It’s a fragmented market, with different constraints, costs and peculiarities.
Not to mention your child’s unique preferences. (“No camps with too many boys.”)
Thankfully, ChatGPT made this process effortless.
If you’re sleeping on Deep Research, this post is for you.
Please research all summer camp programs within a 15-minute drive from zip code XXXX (Manhattan Beach, CA) for two girls ages 8 and 11. Focus on non-sleepaway day camps that emphasize creativity and general activities over competitive sports. Camps must end by July 23, 2025.
Requirements:
- Day camps only (not overnight)
- Suitable for creative, non-sports-focused children
(...)
- Week-long or multi-week sessions welcome
- Casual outdoor activities like swimming, hiking, games are fine as part of program
For each camp, provide:
1. Program name and organization
2. Direct link to camp information/registration
3. Schedule details:
- Session dates (highlighting any that end by July 23)
- Daily hours (full-day vs half-day)
- Session length options
(...)
8. Rating (1-5) based on fit for creative, non-sporty kids:
- 5 = Perfect fit: Heavily creative/artistic focus with variety
- 4 = Great fit: Strong creative elements plus engaging general activities
- 3 = Good fit: Balanced program with solid creative options
- 2 = Okay fit: Some creative elements but less emphasis
- 1 = Poor fit: Limited creative opportunities
Sort results from highest to lowest rating.
Include at least 15 options if available, covering various creative focuses (art, theater, music, maker/STEM creativity, nature/environmental, general enrichment).
Wait, I thought this was only for management consultants?
You’ve probably been ignoring Deep Research because you think it’s limited to folks whose job is to do research. We’re talking McKinsey associates, fund managers and private equity investors.
But here’s another way to think about Deep Research — any time you do a Google Search, you’re doing research.
And let me tell you, this Deep Research report on Summer Camps was an absolute banger.
It found 10 camps and spanned 15 pages.
So what is Deep Research?
Deep Research is a specialized feature that’s part of all the major LLM models.
It’s basically a very smart web crawler and researcher. Upon receiving its instructions (via a prompt) it goes off to the web, social media sites and YouTube and attempts to answer your research request — in a digestible package.
The process can take up to 25 minutes and the final result can approach 70 pages.
And a Deep Research report comes with citations, tables, statistics and a list of all the sources.
wrote a fantastic guide on Deep Research and I highly recommend you go read it.The guide broke down the pricing across all the various model families:
Since I have the $200 version of ChatGPT (which I still struggle to maximize) I get 250 of these queries a month. (Yet I only do <3 per week.)
How can you use Deep Research?
Like any new technology, identifying the potential use cases often requires you to stretch your imagination. Here’s a starting list:
Professional:
Conduct in-depth product, market, or competitor research
Compile a client briefing prior to a meeting
Create a list of pre-FDA biotech companies targeting Alzheimer’s disease
Do a deep dive on a new industry
Identifying guests to interview on your podcast
Personal Life
Plan vacations: research destinations, compare hotels, and build travel itineraries
Compare products (e.g., cars, electronics) with detailed pros and cons
Research personal projects, such as event planning or home renovations
Go deep on a new intellectual curiosity (Stoicism, art history or F-1 racing)
And below, here are some actual prompts.
Example 1: Identifying permitting, zoning and safety regulations to build a pool
Here’s the prompt via @realestatetrent:
Hey so we want to install a new pool in our yard in (omitted) area. Am I going to have any issues with the City? What are my limitations? Please research all pool applications within the last 15 years within that City, any pushback an applicant got from the building or planning department, what the pushback was, how it was resolved, and how long it took.
Example 2: A detailed timeline of the financial crisis from a hedge fund’s perspective
I actually lived through this and the output became this podcast episode:
Example 3: Identifying nominee managers to protect the privacy of an investor
This is some mega-billionaire type sh*t, but it was shared with me:
Can you come up with 100 individuals that are willing to serve as nominee managers in an LLC so that the company can have additional privacy? Please exclude companies from the list
Sign up for our new course
If you’re looking to master AI for your career, check out our course below. Deep Research is a component of the course and it’s aimed at professional knowledge workers who are looking to build their AI advantage. 👇👇👇
ChatGPT Mastery for Professionals Course
A self-paced course for exclusively for our the Founding Members of Future-Proof Your Career with AIThis course is available exclusively for our Founding Members. Upgrade Below.
Example 4: Researching biotech names applying novel approaches to drug discovery
From a friend at a Biotech VC:
I am researching AI biotech companies, namely those who are developing novel therapeutics by applying AI and computational approaches to drug design. Specifically, I want to develop a list of companies who are focused in Hit identification using these techniques. Companies like Insilico Medicine, VantAI, Iambic, and Isomorphic are all top of mind. In practice, these companies have their own specialized models or use models from companies like Isomorphic to ingest target data, then design a new molecule (small or large biologic) that can bind that target. Ideally, their approaches will either make for a better molecule (I.e. more efficacy or less toxicity) and/or they will offset the costs of downstream lab work such as in vitro lead optimization and preclinical in vivo validation.
Example 5: Planning a blow-out 50th birthday
Yup, I’ve got some bougie friends (who don’t like to play golf):
I need you to research and compile 20 unique, high-end birthday trip ideas for a group of 10 men celebrating a 50th birthday. The trip will be 5 days (including a weekend and weekdays) in May 2026.
Key Parameters:
Budget: $5,000 per person ($50,000 total) excluding airfare
Locations: US, Canada, Caribbean, or Mexico (accessible from both East and West Coast)
Weather: Must be warm in May (avoid cold destinations)
No golf focus, but other amenities welcome (beach, fitness, etc.)
Mix of active and relaxed activities preferred
Nightlife not a priority
What Makes These "Unique":
Focus on experiences where the high budget removes typical barriers - think private access, exclusive experiences, chartered services, or activities that most people can't access due to cost.
For each of the 20 ideas, please provide:
Destination & Overview (2-3 sentences on why this is special)
Exclusive/Unique Activities (3-4 specific activities with:)
[Truncated]
Example 6: Potential buyers of a strawberry farm in the Central Valley
A reader runs a residential real estate brokerage firm, but got a strawberry farm listing. They used Deep Research to send physical mailers to prospective buyers (and are looking to do more agriculture deals):
I need help researching potential buyers for a $5 million active strawberry farm in California's Central Valley. Please create a comprehensive buyer prospect list for a direct mail campaign targeting qualified agricultural land investors within a 100-mile radius of the Central Valley.
Research Scope:
Transaction timeframe: Last 5 years (2019-2024)
Geographic focus: 100-mile radius from Central Valley, CA
Transaction types: Agricultural land purchases including:
Strawberry farms
Other berry/fruit operations
General agricultural land
Properties in the $3-7 million range
Required Information for Each Prospect:
Contact Details:
Buyer name (individual or entity)
Mailing address
[TRUNCATED]
Example 7: Profiling a new CEO at a Fortune 500 company
Brett Caughran is a former hedge fund investor who’s on the forefront of using AI in the stock selection process. He’s chronicling his Deep Research experiments on Twitter and here’s how he analyzed Starbucks’ hiring of Brian Nichols (and the full 3 page prompt):
How do I write a good Deep Research prompt
Once again, I’ll reference
’s Deep Research guide. He recommends that you include:A simple and clear goal
Context on your situation (i.e. your “why”)
Structure
Content (which sections to include)
Style (e.g. table, bullets, prose)
This approach works extremely well — but is also very time consuming.
A Deep Research Prompt Shortcut
Recently, I’ve switched to having another model interview me — and then create the prompt.
The process works as follows:
Do a verbal dump of my idea (that loosely covers the categories above)
Use Claude Opus-4 or ChatGPT o3
Ask the model to ask you follow questions
Have it generate your actual Deep Research prompt
Here’s the prompt that gave me the prompt for the friend’s 50th:
Help me come up with a deep research prompt that will be used for ChatGPT with O3. I'm helping a friend come up with birthday party ideas for his 50th, it's a guy's trip. Money is not an issue. Assume that 10 people can each spend $5,000 excluding airfare. The goal would be to do something for five days, spanning a couple of weekdays and a weekend. It would be in May, 2026. Most people would be coming from the East Coast or West Coast, so US, Canada, Caribbean, and Mexico seem to be good locations. Having access to good food would be important. Nice lodgings either a big house or hotel rooms near one another. Amenities like working out, beach, (no golf) are a plus but not necessary. We want the unique experience to be unique. So if there are some atypical and outdoor activities, that would be great. Come up with 20 ideas, links to the companies or individuals coordinating them. Description of the activities, amenities, lodging. Very happy to use a concierge service or travel agent to book.
Before you give me the prompt, ask me any clarifying questions.
Some Advanced Deep Research Tricks
In addition to using a separate model to create your Deep Research prompt there are a few other stealthy ways to make your output even better.
In Getting the most from Deep Research,
describes setting up an entire Claude Project to improve his prompt. This included creating 4 very specific context files:STYLE_GUIDANCE
SOURCE_HIERARCHY
OUTPUT_FORMATS
CONTEXT_BRIEF
I found the Source_Hierarchy file to be pretty powerful. Here’s how Alex prioritizes Business and Economic sources:
Next, there’s Deep Research Connectors (from OpenAI). This was quietly released as a beta on June 4th, 2025:
This feature allows Deep Research (and even regular chats) access files from your internal knowledge base by connecting to Google Drive, Microsoft Sharepoint, Dropbox and Github.
I’ve yet to test this out in my own use cases, but am working with some consulting clients on trying it out on a “ring-fenced” data set.
You can set up connectors by going to Settings → Connectors and then triggering them from the Sources drop-down.
What about those junior analysts?
As a 21 year old investment banking analyst, a big part of my job was “Internet Research.”
My boss would come to me (usually at 8pm) and say:
Can you pull up all possible companies (both public and private) that might buy this small payments company we’re representing?
I’d scour PR Newswire, Dunn and Bradstreet, public company filings and trade publications — meticulously placing them in a spreadsheet.
Now — with a short prompt, you get the report. For $200/month, you get 250 of these reports.
If Deep Research lives up to the hype — and I believe it does — what happens to the future junior researchers of the world?
Are you looking for live tutorials and case studies?
I started a Build with Me series which consist of 10 minute video tutorials showcasing various tools and use cases. I’ll release 1 video a week and they’re free to our paid ($10/month) tier.
Agree with this. Even the less powerful DR is severely underrated. I send it off on research tasks all day every day; love the status module on iPhone.
This nails the everyday value of Deep Research. What clicked for me is the architectural part: if you can orchestrate multi-engine pipelines with progressive learning, deduplication, and scoring, it stops being just a “longer Google” and becomes a reusable discovery system.
We built one starting in K–12, but it now pulls 2,000+ vetted resources per run across any domain — just by swapping engines, not rewriting code. Breakdown here:
https://trilogyai.substack.com/p/ai-discovery
Would love to hear if you’ve tested source prioritization or confidence scoring to handle affiliate noise and low-signal results. It gets real fast.