How Triljon works
A walkthrough of how Triljon works: the signals companies are measured against, how a match score is formed, and how each investor’s engine learns from their decisions.
Triljon reads companies, describes them in a shared vocabulary of signals, and weighs that description against what an individual investor cares about. Every investor runs on their own engine, built in house and tuned by their own activity. This report goes through the parts in order: how companies are read, how a match score is formed, and how the engine learns over time.
The ideas
Each section builds on the one before it, so they read best in order. Together they describe how an investor and a company, two very different things, end up described in the same terms.
01
Signals, a shared language
Investors care about different things. So do companies. Triljon describes both in one vocabulary, a set of named qualities called signals. More than a thousand are measured at once, and the set is never fixed. Signals are added, retired, and recalculated as the engine learns what matters. Each one holds several underlying values, so a signal stands for more than its name alone. The full list, and the scoring behind it, is not public.
02
Meaning, not matching words
Triljon reads text by what it means, not by the words it happens to use. Every phrase becomes a position in a space of meaning, so phrasings that say the same thing sit close together even when they share none of the same words. Two ex-Spotify engineers and an experienced technical team register as one idea. This part of the system is built in house.
03
Tidbits, the facts worth keeping
Wherever there is information, a deck, a profile, a note, a conversation, Triljon reads it and lifts out tidbits: small, self-contained facts. A revenue figure. A detail about the team. A shift in the market. From there the engine reasons about these facts rather than the raw material, so the same fact carries the same weight whether it came from a document or a message.
04
Representation, how strongly a signal shows up
Each tidbit is measured against the signals, and the closer it sits to one, the more it counts toward it. Gather that across everything known about a company and you get its representation: how strongly the company carries each signal. A company can represent a signal heavily, faintly, or not at all.
05
Match score, where you line up
Each investor weights the signals differently: a few matter a great deal, most barely at all. Triljon lays a company’s representation against those weights. Strong representation on a signal you prioritize counts heavily, while representation on a signal you are indifferent to counts for little. The result is a single match score, with the contributing signals still attached.
06
It learns from what you do
A preference is hard to write down in full, so Triljon also learns it from behavior. Every action you take is tracked and weighed, and each one adjusts your engine slightly. Onboarding sets the starting point, and use refines it from there. Nothing resets it along the way.
07
Cases like the ones you liked
To estimate how you will respond to a new case, Triljon finds the cases you have already seen that sit closest to it, and looks at how you responded to those. That learned read is weighed together with the strategy you have described, and the combination decides where the case lands in your inbox.
For investors
The ideas above run underneath the product. This is how they surface, view by view.
The inbox
Founders pitch you through a personal link, and other investors can forward cases your way. Everything arrives in one inbox, already structured and ordered by match score.
Cases
Opening a case shows the company in a structured layout: the story, the team, the documents, and which of your signals it represents and why.
Scout
Scout is a directory of companies that have opted in to being found. Each one is shown with how it fits your signals, and you can reach out directly.
The pipeline
Shortlisted cases move into your pipeline, where notes, messages with founders, and meetings sit together. Passing a case removes it from the inbox.
For founders
On the founder side, the same engine reads your profile and your materials. This is what that looks like in practice.
Pitching
Each investor has a profile you can pitch through. You fill in your story, attach your materials, and submit. The pitch arrives on the investor side as a structured case.
Coverage
Triljon measures how well your profile covers the qualities investors look at, so you can see where information is missing before you send anything.
Coverage enrichment
When something is missing, Triljon asks a focused question and folds your answer into your profile. The questions target the gaps that matter most.
Scout and follow-up
If you opt in to Scout, investors can find you while they browse. Once you have pitched, you can see where each case stands and keep the conversation going when interest is mutual.
What Triljon does not do
Triljon does not make decisions. It organizes and explains: when a case sits near the top of your inbox, you can see which signals placed it there. Every shortlist, every pass, and every conversation is yours.
Each engine also runs privately, for one investor. Your dealflow, your preferences, and your conversations are not shared or pooled.