The companies that win the next market are the ones that saw it forming first.
Every market that exists today started as pain. Someone complained on a forum. Someone built a workaround. Someone switched tools out of frustration. These signs pile up for months before a market becomes obvious — before the first TechCrunch article, before the first VC memo, before the first analyst report.
Most companies see a market only after it has formed. We exist to change that.
How It Works
Complaints in. Market signals out.
Junctiq watches public conversations around the clock — Reddit, HackerNews, startup job boards, and more. We look for real, recurring problems: not trending topics or keyword spikes, but the kind of pain people build workarounds for, switch tools over, and pay money to escape.
We group the signals that point to the same problem, then score each one on 10 factors — how urgent it is, how many people feel it, how long it has lasted, whether they're ready to switch, and whether they'll pay to fix it. When the evidence is strong enough, we publish the call to a permanent public record.
We don't guess what might happen. We spot what's already forming.
Why Trust Us
Anyone can claim they saw it coming. We prove it.
Most of the prediction business has no accountability. Analysts publish reports and never check back. VCs claim they spotted the pattern but never share their hit rate. Consultants hand over a deck and disappear.
We built proof into everything we call: a permanent public record locked with a cryptographic hash, accuracy graded with real math, and a public list of every market we thought was forming but wasn't.
The scorecard isn't our product. It's the reason you can trust our product.
What We Believe
Six principles. No exceptions.
Find the market, don't just describe it.
We read what real people say in public — Reddit threads, HackerNews complaints, GitHub issues, G2 reviews. We watch what they actually do, not what they tell a survey. The signs of a new market are already out there. Our job is to spot them before everyone else does.
A real call is one you can't take back.
A prediction you can quietly edit later isn't a prediction. Every call we make is locked with a cryptographic hash (SHA-256) and a timestamp on a public record. That means nobody can change it later — not even us.
Prove the accuracy. Don't just claim it.
We grade ourselves with Brier scores — a rigorous scoring method used in academic forecasting research. Not testimonials. Not case studies. Math.
Publish the failures too.
We keep a public record of every prediction we got wrong, with an honest breakdown of why. The failures aren't something to hide. They're the proof that our scorecard is real.
Every call should make the next one sharper.
Every prediction that resolves — right or wrong — feeds back into the engine. Each market that forms teaches it what the next one will look like. We get more accurate the longer we run.
You earn the right to advise by showing your scorecard.
If a prediction engine won't grade itself in public, it isn't an engine — it's marketing. We show everything, not because it's comfortable, but because it's the only thing that earns real trust.
The Enemy
Retroactive insight.
The analyst report that tells you what happened last quarter. The platform that tracks the present and calls it “foresight.” The VC “thesis” with no timestamp. The strategy deck with no proof attached.
The enemy isn't a company. It's the habit of describing markets after they form and calling it intelligence.
Our Commitment
Every call is published to a permanent public record.
Every call is locked with a cryptographic hash (SHA-256) the moment we publish it. It can't be changed.
Every call has a deadline. When it arrives, we face the outcome.
Every resolved call — right or wrong — goes on our public track record.
Every wrong call is published with an honest breakdown of why.
We publish our accuracy score the moment we have enough data. Not before. Not after.
We don't edit history. We learn from it.
Verify, don't trust.
Our entire history is public. Every call. Every outcome. Every failure. See the record for yourself.
The Junctiq Manifesto · Accountable Market Detection · 2026
Engine Architecture
Seven stages. Zero guessing.
Every brief follows the same fixed steps, every time. The engine doesn't make things up — it shows its work. Here is exactly how.
Ingest
142 Sources across Pain, Hiring, and Capital
The engine reads about 885 posts a week across 142 sources, including Reddit, HackerNews, job boards, and startup launches. Every post is cleaned up and de-duplicated, so nothing gets counted twice.
Extract
AI reads and sorts each signal
The engine reads each post and sorts it: is this a complaint, a workaround, a new hire, or money moving in? It also pulls out the root cause, who's affected, and what people are using to cope. That means you get structured evidence, not a pile of raw text.
Cluster
Same problem, grouped together
People describe the same problem in different words. The engine groups them. A Reddit complaint and a GitHub issue about the same broken workflow become one signal, so you see the real size of the problem.
Score
Scored on 10 factors
Each grouped problem is scored on 10 factors: how often it shows up, how many different sources mention it, whether people will pay to fix it, and more. That means you get a ranked picture of which problems are real, not a hunch.
Promote
Only the strongest signals advance
Only problems that score high enough and have at least 3 separate sources become opportunities. Everything else goes to a public archive with a note on why it didn't make the cut. That means you only see signals that cleared a real bar.
Monitor
Tracked week over week
The engine watches each opportunity over time. Evidence piling up faster means the window is opening. Evidence going quiet means it's closing. That means you know not just what's forming, but how fast.
Project
Published for you to read
The final opportunity is published in two layers: a free layer anyone can read, and a deeper paid layer with the full timeline, past failures, and the structural edge. That means you get a clear brief, not raw engine internals.
See it in action.
The engine runs every 6 hours. Browse the validated opportunities — each one is the output of this exact pipeline.
Browse Opportunities