DEV Community: John Leslie

I Tested 7 Free AI Startup Idea Validators — Most Are Useless, 3 Are Worth Your Time

John Leslie — Sat, 16 May 2026 07:29:59 +0000

I wanted to validate a startup idea before writing code. So I tested every free AI startup idea validator I could find in 2026 — same idea across all of them.

The test idea: "An AI tool that monitors local government meeting minutes and alerts residents when decisions affect their neighborhood — zoning changes, construction permits, budget allocations. Freemium: free alerts for your address, $9/month for full neighborhood coverage."

I picked this because it's ambiguous — not obviously good or bad. A dog-walking app would be too easy to evaluate. This one requires the validator to actually think about data pipelines, civic tech competition, and municipal format inconsistency.

Full disclosure: I built one of these tools (FounderTools). I'll be upfront about that and let you judge the output quality yourself.

Quick Comparison

Tool	Free?	Signup?	What You Get
IdeaProof	70 credits free	Email	Most comprehensive — 50+ criteria, TAM/SAM/SOM, brand strategy
FounderPal	Yes	No	Fastest — honest 1-paragraph verdict in 5 seconds
FounderTools	Yes	No	8-dimension scoring with experiments (mine — see output below)
ValidatorAI	Yes	No	Conversational feedback, mentions competitors
NxCode	Yes	No	Guided 7-step worksheet, ~10 minutes
WorthBuild	Freemium	Email	Data-driven, real market data, limited free tier
Inodash	Yes	No	Basic 2-paragraph summary

What Stood Out

IdeaProof is the most polished product in this space. 50+ evaluation criteria, competitive landscape, TAM/SAM/SOM estimates, even brand strategy suggestions. The free tier gives you 70 credits (enough for ~3 thorough validations). If you're serious about one specific idea, this is the tool.

FounderPal is the fastest. Type your idea, click validate, get an honest answer in 5 seconds. No scores, no frameworks — just a straight assessment of whether this is worth pursuing. Best for filtering 10 ideas down to 3.

ValidatorAI takes a conversational approach. It mentions competitors, raises risks, and gives general advice. Fine for a gut check, but doesn't produce structured analysis you can act on.

NxCode is more of a guided worksheet. It walks you through 7 steps (problem, solution, market, etc.), each with prompts to think about. Takes ~10 minutes. Good if you want to be forced to think through each angle, but less useful if you want quick feedback.

WorthBuild stands out for using real market data rather than pure AI reasoning. Their blog has a solid comparison article too. But the free tier is limited — most features need a paid plan.

Inodash gave a two-paragraph response with no structure. "This could work but do more research" isn't actionable.

Actual Output: FounderTools on the Government Alerts Idea

Since I built FounderTools, I can show you exactly what the output looks like. Here's the raw result for the government meeting minutes idea — judge for yourself:

Overall Score: 6.5/10

Dimension	Score	Key Reasoning
Market Size	7/10	~150M+ US residents in cities over 50K. Growing civic engagement trend.
Competition	8/10	Very few players in passive monitoring. GovHawk exists but different positioning.
Barriers to Entry	4/10	NLP is commoditized, data is public. Main barrier is operational (data pipelines across thousands of municipalities).
Customer Pain	6/10	Real but not acute. Property value impact creates stronger motivation than general civic awareness.
Monetization	7/10	Clear free/paid differentiation. $9/mo reasonable. Expansion potential to realtors/investors.
Technical Feasibility	6/10	Parsing is straightforward but inconsistent municipal formats across thousands of jurisdictions is the real challenge.
Timing	8/10	Post-2020 civic engagement + housing market volatility + mature NLP tools.
Founder-Market Fit	5/10	Shows market insight but no domain expertise in civic tech or NLP indicated.

Strengths identified: Information asymmetry advantage (transparency exists in theory, not practice), natural B2B expansion path (real estate agents, developers), data network effects as moat.

Risks flagged: Municipal data inconsistency makes automation hard, low engagement ceiling (alerts may not drive action leading to churn), liability from missed/false alerts.

Suggested experiments:

Build MVP for 3-5 municipalities with different formats to test NLP accuracy
Interview 50 recent home buyers about neighborhood research gaps
Create manual alert service for 2-3 neighborhoods to test engagement before automating

The 4/10 on Barriers to Entry is the kind of honest assessment I wanted — it correctly identifies that NLP processing is not a moat, and the real defensibility is operational (building comprehensive municipal coverage).

My Recommendation

For serious validation of your top idea: IdeaProof. Most thorough output, real competitive analysis.

For quick filtering: FounderPal. Five seconds, honest verdict, test 10 ideas in 2 minutes.

For structured scoring: FounderTools. Full disclosure — I built this. The 8-dimension breakdown and suggested experiments are useful, but I'm biased. Try it and compare.

The combo: FounderPal to filter, then FounderTools or NxCode for structured analysis, then IdeaProof for the deep-dive.

No validator replaces talking to actual customers. But spending 30 seconds to identify your weakest assumptions before building is worth it every time.

What do you use to validate ideas? I'm genuinely looking for tools I missed — drop them in the comments.

How to Validate Your Startup Idea for Free in 2026 — A Step-by-Step Framework

John Leslie — Thu, 14 May 2026 07:30:28 +0000

Most startup ideas fail not because they're bad, but because founders skip validation. Here's a free, structured process to pressure-test your idea before writing a single line of code.

Why Validation Matters More Than Ever

In 2026, it costs almost nothing to build an MVP. AI coding tools can scaffold a prototype in hours. But this makes the idea selection problem worse, not better — you can now waste months building the wrong thing faster than ever.

The fix: validate before you build. Here's a 6-step framework using free tools.

Step 1: Score Your Idea Across Multiple Dimensions

Don't just ask "is this a good idea?" — that's too vague. Break it into measurable dimensions:

Market size: How many people have this problem?
Willingness to pay: Are they already spending money on solutions?
Competition: Who else is doing this, and how well?
Feasibility: Can you actually build this with your resources?
Timing: Why now?

A free Startup Idea Validator can score your idea across 8 dimensions and give you a structured assessment in seconds. The key insight: a 6/10 idea with great timing beats a 9/10 idea in a saturated market.

Step 2: Know Who You're Building For

"Everyone" is not a customer segment. Before you build anything, define 2-3 specific personas:

What's their job title?
What frustrates them today?
Where do they hang out online?
What would they pay, and how?

A Customer Persona Builder generates detailed personas based on your business description, including acquisition channels — which tells you where to find these people, not just who they are.

Pro tip: The acquisition channels matter more than the demographics. Knowing your customer is a "35-year-old product manager" is less useful than knowing they read Lenny's Newsletter and hang out in specific Slack communities.

Step 3: Understand the Competitive Landscape

Most founders either ignore competitors ("we have no competition") or get paralyzed by them ("there are 50 companies doing this"). Both are wrong.

The right question: What are customers doing TODAY to solve this problem? That's your real competition — and it includes doing nothing, using spreadsheets, or hiring a freelancer.

Run a structured SWOT Analysis on your idea. This forces you to think about weaknesses and threats — not just the exciting parts.

Key things to look for:

Strengths you can't easily replicate: If your only advantage is "we use AI," that's not defensible.
Weaknesses that are fixable vs. structural: "We don't have a brand yet" is fixable. "Our unit economics require 10M users" is structural.
Opportunities from market shifts: New regulations, technology changes, or behavioral shifts.
Threats from incumbents: Will Google/Apple/Microsoft build this?

Step 4: Map Your Business Model

An idea without a business model is a hobby. Before you write code, answer:

How will you make money? (Revenue model)
What does it cost to serve one customer? (Cost structure)
How will you reach customers? (Channels)
What activities are essential? (Key activities)
Who do you need? (Key partners)

The Business Model Canvas Generator gives you all 9 blocks filled in based on your idea. Use it as a starting point, not gospel — the value is in the thinking process, not the output.

Common trap: Don't default to "freemium" because it's easy. Freemium works at scale (10K+ free users to get 100 paid). If you're starting from zero, consider charging from day one — it's the fastest validation signal.

Step 5: Craft Your Pitch (Even If You're Not Raising)

Writing an elevator pitch isn't just for investors. It forces you to distill your idea into one clear sentence:

[Product] helps [audience] [solve problem] by [mechanism], unlike [alternative].

If you can't fill in those blanks clearly, your idea isn't clear enough yet. An Elevator Pitch Generator can give you 3 variations tailored to different audiences (investors, customers, partners).

The real test: Tell 5 people your pitch. If they immediately ask follow-up questions, it's clear. If they say "oh, cool" and change the subject, it's not.

Step 6: Set Your Price

Most founders think about pricing last. That's backwards. Your price determines your market:

$10/month = you need thousands of customers (B2C, self-serve)
$100/month = you need hundreds (prosumer, small business)
$1,000/month = you need dozens (B2B, sales-led)
$10,000/month = you need a handful (enterprise)

Use a Pricing Strategy Generator to model different approaches — value-based, competitor-based, cost-plus — and see which makes sense for your specific product and market.

The pricing shortcut: Find 3 competitors. Price between the cheapest and the most expensive. Start high and discount — it's much harder to raise prices later.

What Comes After Validation

Validation isn't a one-time event. It's a cycle:

Idea scored — Refine or kill
Personas defined — Talk to 5 real people who match them
Competition mapped — Find the gap
Business model clear — Build the simplest version that tests your riskiest assumption
Pitch tested — Pre-sell before building

The entire validation process above takes 1-2 hours using free tools. Compare that to the months you'd spend building something nobody wants.

TL;DR

Step	What	Free Tool
1	Score your idea	Idea Validator
2	Define personas	Customer Persona Builder
3	Analyze competition	SWOT Analysis
4	Map your business	Business Model Canvas
5	Craft your pitch	Elevator Pitch Generator
6	Set your price	Pricing Strategy

All free. No signup required. No email gate. Just paste your idea and get structured output.

What's your validation process? Have a step I missed? Drop a comment.

I Built 5 Free AI Tools That Replace $500 in Startup Consulting

John Leslie — Wed, 13 May 2026 11:06:50 +0000

I built a set of free AI tools that help startup founders think through the early-stage decisions that usually cost $500+ in consulting fees or hours of research.

The Tools

1. Startup Idea Validator

Describe your startup idea and get an 8-dimension analysis: market size, competition, timing, defensibility, revenue potential, execution complexity, team fit, and a weighted overall score. Each dimension gets a 1-10 rating with specific reasoning.

2. Elevator Pitch Generator

Input your startup details and get three tailored pitches: one for investors, one for customers, one for partners. Each pitch is under 60 seconds and hits different pain points.

3. SWOT Analysis Generator

Get a structured strengths/weaknesses/opportunities/threats analysis with a strategic score and prioritized action items. Useful for investor decks and internal planning.

4. Customer Persona Builder

Generates 3 detailed customer personas with demographics, pain points, goals, objections, and recommended acquisition channels. Helps you figure out who you're actually building for.

5. Business Model Canvas Generator

Enter your business idea and get a complete Business Model Canvas filled out — all 9 building blocks: key partners, activities, resources, value propositions, customer relationships, channels, customer segments, cost structure, and revenue streams.

Why Free?

DimeADozen charges $19/report for similar analysis. Most founders run through dozens of ideas before landing on one worth building. Paying $19 each time adds up fast.

These tools are completely free, no signup required. They use Claude Sonnet to generate structured analyses in under 10 seconds.

What's Next

More tools are coming: Business Model Canvas Generator, Pricing Strategy, Competitor Analysis, Go-to-Market Planning.

All tools are at tools.predictionoracle.io.

Built with vanilla HTML + Vercel serverless functions + Anthropic Claude API.

Google Places API Now Costs $275/Month — Here's a Free Alternative

John Leslie — Mon, 11 May 2026 15:42:02 +0000

Google Places API pricing hit a breaking point in 2026. The basic "Essentials" tier starts at $275/month, and if you need contact details it's $400+. For indie developers, MVPs, and side projects, that's a non-starter.

I built BizData — a free REST API and MCP server that returns business data for any location worldwide. No API key. No signup. No rate limits (yet).

What It Does

Query any city + business category and get structured JSON back: name, address, phone, website, email, coordinates, opening hours. 37 business categories from restaurants to hospitals.

Example — find cafes in Paris:

GET https://bizdata-web.vercel.app/api/businesses?city=paris&category=cafe&limit=5

Returns:

{
  "businesses": [
    {
      "name": "Café de Flore",
      "address": "172 Boulevard Saint-Germain, Paris",
      "phone": "+33 1 45 48 55 26",
      "website": "https://cafedeflore.fr",
      "latitude": 48.8540,
      "longitude": 2.3325
    }
  ],
  "count": 2309,
  "city": "paris",
  "category": "cafe"
}

How It Works

BizData wraps the OpenStreetMap Overpass API and normalizes the messy OSM tag system into a clean REST interface. European cities have 50-70% data completeness. Paris alone has 8,300+ restaurants and 2,300+ cafes.

Also an MCP Server

If you're building with Claude, Cursor, or any MCP-compatible client:

https://bizdata-web.vercel.app/api/mcp

Add it as a remote MCP server (Streamable HTTP transport). Your AI agent can then search businesses in natural language.

When to Use BizData vs Google Places

	BizData	Google Places
Price	Free	$275-400+/month
Auth	None	API key required
Data source	OpenStreetMap (community)	Google's proprietary data
Best for	Europe, Asia, prototypes	US/UK, production apps needing 95%+ coverage
Completeness	50-70% (Europe), 20-33% (US/UK)	90%+ globally

Honest take: If you need comprehensive US data, Google is still better. If you're building an MVP, working with European data, or just don't want to pay $275/month for a side project, BizData works.

Try It

API: https://bizdata-web.vercel.app/api/businesses?city=london&category=restaurant&limit=3
Comparison of 7 alternatives: https://bizdata-web.vercel.app/compare
City explorer: https://bizdata-web.vercel.app/cities/paris

No signup. No API key. Just make a GET request.

37 Supported Categories

restaurant, cafe, bar, hotel, hospital, pharmacy, bank, supermarket, gym, school, university, library, cinema, theatre, museum, park, parking, fuel, car_repair, dentist, veterinary, post_office, police, fire_station, embassy, nightclub, fast_food, bakery, butcher, convenience, clothes, electronics, furniture, hairdresser, beauty, laundry, car_wash

Built this as a solo project. Feedback welcome — especially if you find data gaps in specific cities.

Also check out FounderTools — free AI-powered tools for startup founders. Validate ideas, generate pitches, run SWOT analyses, and build customer personas.

Free Google Places API Alternative: 37 Business Categories, No API Key

John Leslie — Sat, 09 May 2026 15:47:27 +0000

Google Places API removed its free tier in February 2025. The new pricing starts at ~$275/month for 100K calls.

If you need basic business/POI data and don't need Google's reviews or photos, here's a free alternative.

BizData API

A REST API wrapping OpenStreetMap's Overpass API. Free, no API key, no signup.

curl "https://bizdata-web.vercel.app/api/businesses?location=Paris&category=cafe&limit=3"

Returns structured JSON with name, address, phone, website, email, coordinates, and opening hours.

Google Places vs BizData

Feature	Google Places	BizData
Price	$275+/mo	Free
API key	Required	Not needed
Categories	100+	37
Reviews/photos	Yes	No
Coverage (EU)	Excellent	Excellent
Coverage (US)	Excellent	Good
Rate limits	Strict	Fair use

When to use this

Lead generation (bulk business listings)
Market research (how many cafes in a neighborhood?)
Location-based apps needing basic POI data
Prototyping before committing to Google's pricing
MCP tool calls from AI assistants

Endpoints

GET /api/businesses?location=X&category=Y - Find businesses
GET /api/count?location=X&category=Y - Count businesses
GET /api/categories - List all categories

Optional: radius_km (default 5), limit (max 500).

MCP Server

Also works as an MCP server for Claude, Cursor, or any MCP client:

{
  "mcpServers": {
    "bizdata": {
      "type": "streamable-http",
      "url": "https://bizdata-web.vercel.app/api/mcp"
    }
  }
}

Try it

bizdata-web.vercel.app

Data from OpenStreetMap. Built with TypeScript, hosted on Vercel.

Also check out FounderTools — free AI-powered tools for startup founders. Validate ideas, generate pitches, run SWOT analyses, and build customer personas.

I made a free REST API that turns 'find cafes in Paris' into structured JSON

John Leslie — Sat, 09 May 2026 11:25:44 +0000

Every few months I need local business data -- restaurants near an office, gyms in a neighborhood, dentists in a city. The usual path: learn the Overpass query language, handle raw OSM tags, geocode addresses separately, parse nested JSON.

I wanted something simpler: pass a city and category, get clean JSON back. So I built BizData API.

Quick start

curl "https://bizdata-web.vercel.app/api/businesses?location=Paris&category=cafe&limit=2"

{
  "total": 2338,
  "location_resolved": "Paris, France",
  "businesses": [
    {
      "name": "Cafe Exemple",
      "category": "cafe",
      "address": "Avenue Parmentier, 28",
      "phone": "+33 1 48 05 94 36",
      "website": "https://aunreve.fr",
      "email": "contact@aunreve.fr",
      "lat": 48.8606,
      "lon": 2.3787,
      "opening_hours": "We-Sa 08:30-19:00; Su 09:00-19:00"
    }
  ]
}

No API key. No signup. One URL.

What it does that raw Overpass doesn't

Overpass is powerful but requires learning a custom query language:

[out:json][timeout:30];
node["amenity"="cafe"](around:5000,48.8566,2.3522);
out center tags;

BizData handles the geocoding, query building, tag mapping (37 normalized categories), and result formatting. You get a standard REST endpoint with clean JSON instead of a custom QL.

Three endpoints

GET /api/businesses -- Find businesses by location and category.

Param	Required	Description
location	Yes	City or address
category	Yes	One of 37 types
radius_km	No	Search radius, default 5
limit	No	Max results, default 50, max 500

GET /api/count -- Count businesses across categories. "How many restaurants, cafes, and bars in London?" returns 6,763.

GET /api/categories -- List all 37 available categories.

Error handling

Bad category:

{"error": "Unknown category: pizza. Available: accountant, bakery, bank, ..."}

Missing params:

{"error": "Missing required parameter: location"}

Performance

/api/categories: ~50ms (cached)
/api/businesses (50 results): 1-3 seconds
/api/count (all categories): 10-30 seconds
Results cached for 1 hour on Vercel's edge

Limitations

Data quality depends on OpenStreetMap contributors
Coverage: excellent in Europe, solid in major US/Asian cities, sparse in rural areas
Shared Overpass API pool
Some businesses lack phone/email/hours

Built with

TypeScript, Vercel serverless functions, OpenStreetMap Nominatim (geocoding) + Overpass API (data).

Try it

Interactive demo: bizdata-web.vercel.app

All 37 categories: accountant, bakery, bank, bar, beauty, bookstore, cafe, car_dealer, car_repair, cinema, clothing, coworking, dentist, doctor, electronics, florist, furniture, gallery, gas_station, guest_house, gym, hairdresser, hospital, hostel, hotel, insurance, lawyer, museum, parking, pet_shop, pharmacy, real_estate, restaurant, school, supermarket, theatre, university.

Also check out FounderTools — free AI-powered tools for startup founders. Validate ideas, generate pitches, run SWOT analyses, and build customer personas.

How We Made 61% on Hormuz in 9 Days

John Leslie — Wed, 29 Apr 2026 11:17:44 +0000

Our Hormuz April NO position is worth $803 on a $500 bet. The market closes tomorrow at 0.4%. Nine days ago, it was at 38%.

The trade was not a gamble. It was arithmetic.

Why the Market Was Wrong

On April 20, Hormuz April YES was trading at 38%. The market gave a one-in-three chance that Strait traffic would normalize before month-end.

The ceasefire had just expired (April 19). Iran closed it with zero diplomatic progress.
Iran had fired on ships during the truce (April 18).
Only 16 ships per day were transiting, down from the normal 60+.
There was no active diplomatic track.

Even if talks started immediately, normalization requires Iran to withdraw naval assets, commercial insurers to resume coverage, and shipping lines to re-route vessels. Ten days was not enough.

Result: Hormuz April NO at $0.62 to $0.996. Return: +61% in 9 days.

The Portfolio

Fictional $1,000 portfolio (started April 20): now worth $1,637 (+63.7%).

Position	Entry	Current	Return
Hormuz April NO	$0.62	$0.996	+61%
Hormuz May NO	$0.305	$0.645	+111%
Cash	$200	$200	--

What Needs to Happen for May

Hormuz May is at 35.5% YES. For normalization by May 31, all of these must happen in 32 days:

A new ceasefire agreement (none active).
Iran withdraws its dual naval blockade.
Commercial shipping resumes normal volumes.
Traffic returns to 60+ ships/day (currently around 5).

Iran has said reopening is impossible while sanctions persist.

June

Hormuz June is at 58.5%. The uncertainty is genuinely high and our edge is not clear. We are not touching June.

Knowing when you do not have edge is as important as recognizing when you do.

Full analysis: predictionoracle.io/issue-9

Subscribe to The Market Oracle

Polymarket Wants Back Into the US. Here's What Changes If They Get In.

John Leslie — Wed, 29 Apr 2026 07:35:43 +0000

Polymarket just asked the CFTC for permission to bring its main exchange back to American traders. On the same day, it ripped out its entire trading infrastructure and rebuilt it from scratch. Kalshi posted $3.91 billion in weekly volume, its highest ever. The Senate introduced a bipartisan resolution to ban all lawmakers from prediction markets. And Finland is the betting favorite to win Eurovision.

Polymarket Wants Back Into the US

This is the biggest story in prediction markets this year, and it happened almost quietly.

On April 28, Polymarket began formal discussions with the CFTC to lift the ban that has kept its main exchange off-limits to American users since 2022. That year, the CFTC fined Polymarket $1.4 million for operating without proper registration. Polymarket settled, pulled out of the US, and spent the next four years building the largest prediction market in the world from outside America's borders.

The timing isn't accidental. Last July, Polymarket acquired QCEX, a CFTC-licensed exchange, for $112 million. That gave them a regulatory vehicle. Now they want to use it.

If the CFTC approves, the effect is simple: American retail liquidity floods back into Polymarket. The US has more prediction market traders than any other country. Polymarket has been serving them through a wink-and-nod VPN culture, but institutional capital and mainstream retail won't touch an exchange where US participation is technically prohibited. Remove that prohibition, and market depth on major events could double or triple.

The Platform Overhaul Nobody Noticed

On the same day Polymarket filed with the CFTC, it executed a complete infrastructure upgrade. New smart contracts. A rewritten order book engine. A new collateral token called Polymarket USD (pUSD), backed 1:1 by USDC on Polygon. Trading paused for roughly an hour during the cutover.

This is the kind of thing that gets zero media coverage but matters enormously. The old infrastructure was built when Polymarket processed a few hundred thousand dollars a day. Now it handles hundreds of millions.

Kalshi Just Had Its Best Week Ever

Kalshi posted $3.91 billion in weekly volume for the week of April 20-28, its highest ever. That's nearly double Polymarket's weekly volume for the same period. Sports parlays drove more than 85% of the total.

Kalshi isn't competing with Polymarket for political prediction nerds. It's competing with DraftKings and FanDuel for sports bettors. The prediction market wrapper lets Kalshi offer sports parlays with lower vig than traditional sportsbooks.

The Senate Wants Lawmakers Out

Sen. Bernie Moreno (R-OH) introduced a bipartisan resolution on April 24 that would ban all members of Congress from trading on prediction markets entirely. This is separate from and broader than the PREDICT Act. Five co-sponsors from both parties.

Paradoxically bullish for the industry. Every one of these bills implicitly accepts that prediction markets are legitimate financial instruments worth regulating. You don't ban members of Congress from doing something that's about to be shut down.

Eurovision 2026 Odds

Finland leads at 35.3% on Polymarket. Interesting discrepancy: Denmark is 13.4% on Polymarket but only 7% at bookmakers. Israel is the reverse: 5.1% on Polymarket versus 11% at bookmakers. When the same event is priced differently on two platforms, someone is leaving money on the table.

Our Portfolio: +63.7%

Starting capital $1,000 on April 20. Current value: $1,637. Hormuz April NO position resolves tomorrow at +61%. Hormuz May NO up 111%.

Full issue with charts and analysis: predictionoracle.io/issue-8.html

The Market Oracle is a weekly prediction market intelligence newsletter. Subscribe at predictionoracle.io

The Cop Left the Building: CFTC Shrinks 24% While Prediction Markets Hit $60B

John Leslie — Tue, 28 Apr 2026 07:24:50 +0000

Issue #7 · April 27, 2026 · By John Leslie

The agency responsible for policing prediction markets has lost a quarter of its workforce. Meanwhile, the industry just crossed $60 billion in trades this year. Brazil banned every prediction market platform overnight. And the economist who invented the theoretical framework behind prediction markets argues that the insider trading everyone is panicking about is actually the entire point. Welcome to the most confusing week in the industry's short history.

The Cop Left the Building

The Commodity Futures Trading Commission -- the federal agency that regulates prediction markets -- has shrunk to its smallest size in 15 years. Workforce is down 24% since January 2025. That's not attrition. That's a policy choice.

This is happening precisely as the agency's jurisdiction has become the most contested terrain in financial regulation. More than two dozen state-level lawsuits allege that Polymarket and Kalshi are unlicensed gambling operations. Three states tried to ban the platforms outright last week; the DOJ sued all three, arguing federal preemption. The CFTC is supposed to referee this fight. It now has fewer referees than at any point since 2011.

The volume numbers make the staffing collapse almost comical. Kalshi and Polymarket have processed $60 billion in trades so far in 2026. In March alone, Kalshi handled $13 billion and Polymarket roughly $10 billion. For context, the entire CFTC-regulated prediction market had less than $1 billion in annual volume as recently as 2023. We're talking about a 60x increase in three years with a 24% decrease in oversight capacity.

Our take: Underfunding the regulator is short-term bullish and long-term dangerous. Right now, it means platforms operate with minimal oversight -- good for innovation, bad for legitimacy. The risk is that a major scandal forces Congress to overcorrect.

Brazil Bans Everything

Brazil became the first major economy to ban prediction market platforms entirely. As of today, Polymarket, Kalshi, and 27 other platforms are blocked in the country. The Brazilian government classified all of them as illegal betting operations.

This isn't a small market. Brazil has the fifth-largest internet population globally and a massive sports betting culture. Polymarket had meaningful traffic from Brazilian users. That's gone overnight.

Our take: Brazil's ban is a setback for global market depth but reinforces the US as the center of gravity. If you're building in this space, your regulatory strategy is now a US-first strategy by default.

Robin Hanson: Insider Trading Is the Point

Fortune published a remarkable interview this week with Robin Hanson, the George Mason economist whose 1990s work essentially invented the theoretical framework for prediction markets. His argument: the insider trading that everyone is panicking about is exactly what makes prediction markets valuable.

The logic: prediction markets aggregate dispersed information into accurate prices. When a soldier bets on a military operation he has classified knowledge about, that information moves the price closer to truth. The market becomes more accurate, not less.

Our take: Hanson is directionally right and practically wrong. In a frictionless theory, insider information makes markets more efficient. In the real world, it creates perverse incentives. The right answer is the one Kalshi is already implementing: ban the trades that create bad incentives (politicians on their own races), allow the rest.

Hormuz: "Reopening Is Impossible"

Iran's parliament speaker said this week that "reopening the Strait of Hormuz is impossible as long as the US blockade is in place." Ship traffic remains at roughly 5 vessels per day versus 140 pre-crisis. Bloomberg described it as "maritime trench warfare."

Current prices:

April normalization: 0.4% YES -- locked in
May normalization: 36.5% YES -- well below our entry at 69.5%
June normalization: 55.5% YES -- still overpriced in our view

Original Research: How Calibrated Are These Markets?

We built an interactive calibration tracker analyzing 7,661 resolved Polymarket markets. Key findings:

Excellent at the extremes. Events priced at 0-5% almost never happen (0.1% actual). Events at 95-100% always happen.
Systematic underpricing in the middle. Markets at 65-75% resolved YES 96.5% of the time.
Overall Brier score of 0.025 -- roughly 10x better than random guessing.

Explore the data yourself: polymarket-calibration.vercel.app

Portfolio Update: +62.7%

Position	Entry	Current	P&L
Hormuz April NO	YES 38%	YES 0.4%	+60.6%
Hormuz May NO	YES 69.5%	YES 36.5%	+108.1%
Cash reserve	$200	$200	--
Total	$1,000	$1,627	+62.7%

This is The Market Oracle, a weekly prediction market intelligence newsletter. Read more at https://site-two-nu-51.vercel.app

Are Prediction Markets Well-Calibrated? I Analyzed 7,661 Resolved Polymarket Markets

John Leslie — Sat, 25 Apr 2026 15:33:05 +0000

By John Leslie

Prediction markets are booming. Polymarket alone processes billions in volume. But a fundamental question remains: when a market says there's a 60% chance of something happening, does that event actually happen 60% of the time?

I pulled 7,661 resolved binary markets from Polymarket's API, grabbed price history for the top 2,000 by volume, and built an interactive calibration tracker to find out.

Try it here: Polymarket Calibration Tracker

What is calibration?

A forecaster is well-calibrated if their predicted probabilities match observed frequencies. If you say "70% chance" for 100 different events, roughly 70 of them should happen. Perfect calibration means every point sits on the diagonal line where predicted = actual.

This is the gold standard metric for forecast quality, used everywhere from weather forecasting (the NWS publishes calibration curves) to machine learning model evaluation.

The results

Markets are excellent at the extremes

Events priced at 0-5% almost never happen (0.1% actual resolution rate). Events at 95-100% always happen (100% actual). This is reassuring: when the crowd is very confident, the crowd is right.

Systematic underpricing in the middle range

This is where it gets interesting. Markets trading at 65-75% actually resolved YES 96.5% of the time (29 markets). At 45-55%, the actual rate was 60% (25 markets). The sample sizes are small in these middle bins, so interpret cautiously, but the pattern is consistent: mid-range markets seem to underprice the YES outcome.

One hypothesis: multi-outcome markets inflate the low-probability bins. If there are 10 candidates in an election, 9 of them trade at 0-5% and lose, pulling the average down. The winners cluster higher. This creates a distributional skew that looks like underpricing in the middle.

Calibration improves closer to resolution

This makes intuitive sense. The 24-hour Brier score (0.025) is significantly better than the 30-day score (0.042). As resolution approaches, uncertainty resolves, and prices converge to 0 or 1.

For context: random guessing on balanced outcomes gives a Brier score of 0.25. A Brier score of 0.025 is excellent.

Category breakdown

The tool lets you filter by category. Some highlights:

Politics (1,312 markets): Well-calibrated at extremes, biggest mid-range deviation
Sports (1,314 markets): Similar pattern to politics, slightly tighter calibration
Crypto (752 markets): More volatile, noisier calibration curve
Geopolitics (404 markets): Small sample but interesting patterns
Science/Tech (163 markets): Too few markets for reliable conclusions

Technical details

Data collection: Polymarket's Gamma API for resolved market metadata. CLOB (Central Limit Order Book) midpoint prices at 24h, 7d, and 30d before resolution for the top 2,000 markets by volume.

Methodology: Markets binned into probability ranges (0-5%, 5-15%, ..., 95-100%). For each bin, the actual resolution rate is compared to the predicted probability. Proportional dot sizes indicate sample count.

Caveats:

75% of price-tracked markets fall in the 0-5% bin, driven by multi-outcome market losers
Small sample sizes in mid-range bins (20-50 markets each) limit statistical significance
Only resolved markets are included (survivorship consideration)

Why this matters

Prediction markets are being used for increasingly high-stakes decisions. DARPA studied them for intelligence forecasting. The Federal Reserve has explored them for inflation expectations. Companies like Google and HP have used internal prediction markets for product planning.

If these markets are well-calibrated, their prices can be treated as genuine probability estimates. If they're systematically biased, traders and decision-makers need to adjust.

The data suggests Polymarket is well-calibrated overall (Brier score 0.025), with some interesting biases in the middle range worth investigating further as sample sizes grow.

Explore the data yourself: polymarket-calibration.vercel.app

I write The Market Oracle, a weekly prediction market intelligence newsletter. Read more at our site.

The Regulatory Storm: 7 Bills, 3 Bans, and the Future of Prediction Markets

John Leslie — Sat, 25 Apr 2026 07:46:46 +0000

Originally published at The Market Oracle

Prediction markets just became Washington's hottest regulatory target. In a single week: seven congressional bills introduced targeting prediction market oversight, three state-level enforcement actions, Kalshi's first-ever self-enforcement against politicians betting on their own races, and the DOJ's first insider trading prosecution in the space.

The Regulatory Blitz

Seven bills. One week. That's not a regulatory drip -- that's a flood.

The headline legislation is the bipartisan PREDICT Act, which would ban politicians and their immediate family members from trading on political events.

New York Governor Hochul issued an executive order banning all state employees from prediction market insider trading. Connecticut, Arizona, and Illinois attempted outright bans. The federal government simultaneously sued all three states to block their actions -- arguing prediction markets fall under CFTC jurisdiction.

Our take: Regulation is coming, but it's bullish long-term. Clear rules mean institutional adoption.

Kalshi Becomes the Sheriff

Kalshi fined and suspended three congressional candidates for placing wagers on their own races -- the first enforcement action of its kind. Fines ranged from \$539 to \$6,229.

The fines are small. The signal is enormous. Self-policing is the cheapest insurance against existential regulatory risk.

Hormuz: The Dual Blockade Tightens

Five ships transited in the last 24 hours. The pre-crisis average was 140 per day. Oil at \$105/barrel.

April normalization at 1.7% -- effectively resolved NO, exactly as we called when it was at 38%. Our portfolio: +61.2%.

Portfolio: +61.2%

Position	Entry	Current	P&L
Hormuz April NO	YES 38%	YES 1.7%	+58.6%
Hormuz May NO	YES 69.5%	YES 37%	+106.4%
Total	\$1,000	\$1,612	+61.2%

Read the full issue | All issues

When Prediction Markets Become the News: Shoot-and-Kill Orders, Fox News Deals, and John Oliver

John Leslie — Fri, 24 Apr 2026 11:18:25 +0000

Trump just ordered the Navy to shoot Iranian mining boats in the Hormuz Strait. Kalshi is now embedded in Fox News, CNN, and CNBC. John Oliver devoted a segment to prediction markets. Forbes created a market on a children's mass shooting. And our fictional portfolio just crossed +63.6% in 4 days.

This week, prediction markets stopped being a niche curiosity and became the news itself.

"Shoot and Kill"

On April 23, President Trump ordered the US Navy to "shoot and kill" any boats caught laying sea mines in the Strait of Hormuz. The order came after Iran seized two more ships and fired RPGs at a Greek-owned cargo vessel.

The Strait remains effectively closed. The US naval blockade has turned back 31 Iran-linked vessels since April 13. Iran says reopening is impossible while the blockade continues. Oil is at $95/barrel.

Portfolio impact: When we opened our May NO position on April 20, YES was at 69.5%. It's now at 33.5%. Our position is up 117.9% on that leg alone. The "shoot and kill" order is not the language of de-escalation. We're revising our May fair value estimate down to 15-20%. The position stays open.

Prediction Markets Go Primetime

Kalshi has signed partnerships with CNBC, CNN, Fox News, and the AP. Fox Corporation announced it will weave real-time prediction data into Fox News, Fox Business, Fox Weather, and streaming. When you watch cable news, you now see prediction market odds alongside stock tickers.

Kate Knibbs at Wired established herself as a dedicated prediction markets reporter. Polymarket partnered with Substack (Feb) and Dow Jones (Jan). The infrastructure of mainstream coverage is being built in real time.

John Oliver devoted a Last Week Tonight segment to prediction markets, questioning the industry's rapid expansion and regulatory gaps. When a comedian covers your industry, you've gone mainstream.

Where's the Line?

Not everyone is celebrating.

Forbes created a prediction market that gamified a mass shooting that killed eight children. The backlash was immediate (reported by 404 Media, April 20). The dark side of "a prediction market for everything": some events should not be betting opportunities.

ProPublica added a section to its ethics code prohibiting staff from betting on news events through prediction markets. The concern: when journalists can bet on outcomes they cover, the language of forecasting begins to influence the language of reporting.

The Columbia Journalism Review published a deep analysis warning that the CFTC's pro-prediction-market posture may not survive a genuine crisis of public trust.

Portfolio: +63.6% in Four Days

Position	Entry	Current	P&L
Hormuz April NO	YES 38%	YES 2.9%	+56.5%
Hormuz May NO	YES 69.5%	YES 33.5%	+117.9%
Cash reserve	$200	$200	--
Total	$1,000	$1,636	+63.6%

We called April at 38% -- it's now 2.9%. We called May overpriced at 69.5% -- it's now 33.5%. Both calls directionally correct. The "shoot and kill" order strengthens our conviction.

Market Snapshot

Hormuz May normalization: 33.5% YES (was 69.5% four days ago)
Hormuz June normalization: 57.5% YES
2028 Presidential: JD Vance 18.4%, Gavin Newsom 17.1%, Marco Rubio 10.2%
Democrats win House 2026: 84.5% YES
Eurovision 2026: Finland 36.4%

Read the full analysis with more detail at The Market Oracle. Published by John Leslie.