Product
Projection EngineModel CouncilSignal DeskSector GraphWatchlistsResearch Automations
Company
How it works Solutions Enterprise Pricing Contact sales
Log in Get started
How it works

From a ticker to a
defensible projection
in about a minute

No black box. Every step of the pipeline is visible, every claim carries its source, and every projection is graded against what actually happened.

1

Ask about an asset

Type a ticker, a sector, or a question in plain English — "what happens to ORCL if datacenter capex slows?" Mael resolves it to the instruments involved and the horizon you care about.

2

Assemble the evidence

The engine pulls live quotes, the last four filings, the most recent transcript, analyst revisions, options positioning, relevant macro series and weighted social sentiment. Everything is timestamped and cited.

3

Run the council

Five model classes analyse the same evidence independently — no shared scratchpad, no single narrator. Each returns a directional view, a target range, and the factors it weighted most heavily.

4

Reconcile and score

Mael compares the five outputs. Agreement raises confidence; divergence lowers it and surfaces the dissenting case in full. The result is a projection with a number attached to how much to trust it.

5

Watch it, and grade it

The projection is saved with its inputs. As reality arrives, it's scored automatically — and every past projection stays visible in your history, right or wrong.

Who sits on the council

Five model classes with genuinely different failure modes. The point isn't more compute — it's disagreement you can see.

Reasoning model

Multi-step chains over filings, guidance and unit economics. Best on "why" questions.

Analysis model

Careful, conservative readings of long documents. Tends to be the dissenting voice.

Synthesis model

Wide-context summarisation across many sources at once. Good at spotting the through-line.

Sentiment model

Real-time social and news tone, weighted by account quality rather than volume.

Open-weight model

An independently-hosted cross-check so the council isn't all one vendor's blind spots.

Specific model versions rotate as new releases ship. The current roster is always listed inside the app.

Confidence, honestly

What the confidence number actually means.

It is not a probability that the price hits the target. It's a measure of how much the five models agreed, how much evidence they had, and how well projections with this shape have scored historically. High confidence with thin evidence gets marked down.

  • Agreement weight. Five-of-five beats three-of-five, and the gap is shown.
  • Evidence depth. A quiet week with no filings lowers the ceiling.
  • Historical calibration. Scored against how similar past projections resolved.
Council output · sample
Reasoning
+78
Analysis
−34
Synthesis
+64
Sentiment
+81
Open weights
+57
Council confidence71%

One model dissents on valuation. Its full case is one click away.

Kept honest

Every projection stays on the record.

Most prediction products quietly forget the bad calls. Mael keeps all of them, timestamped and scored, in a history you and anyone you share a workspace with can read. If the hit rate slips, you'll see it before we tell you.

  • Immutable snapshots. Inputs and outputs frozen at generation time.
  • Automatic grading at each horizon as the window closes.
  • Per-model scorecards so you learn which voice to weight.
Scored history · sample

What it is — and what it isn't

Mael is

  • A research tool that shows its work
  • A way to see several models disagree in one place
  • A faster path through filings, transcripts and flow
  • A record of what was projected and how it resolved

Mael is not

  • Investment advice or a recommendation
  • Connected to your brokerage — it never places trades
  • A guarantee of any outcome, ever
  • A replacement for a licensed financial professional

Be there when it opens

Private beta, opening in small groups. No card, no commitment — just an email when there’s room.

Start free trial Back to overview