Evidence-backed opportunity intelligence
See the change before the consensus.
Important shifts leave traces before they become headlines. Antegraph is being built to discover emerging developments in original sources, connect evidence across markets and entities, and turn them into opportunities worth investigating.
From first signal to testable hypothesis, with the evidence for it, the evidence against it and a clear next step.
From signal to opportunity
From isolated signals to emerging opportunities.
A filing, a procurement notice or a change in capacity can seem insignificant on its own. Antegraph connects these developments to reveal what may be changing, who could benefit and which assumptions need to be tested.
Synthetic examples. Every source, event, relationship and outcome is invented to show the method.Signal
A regional health purchaser’s tender now lets up to 30% of laparoscopic instrument volume be supplied as certified reprocessed devices. Its previous tender did not.
Quoted · tender notice · lot 4
Connections
Two weeks later the national device regulator shortens the certification path for reprocessors. A sterilization services firm files to expand capacity in the same region.
Quoted · 3 independent origins
Opportunity hypothesis
If more purchasers copy the clause, certified reprocessing capacity, not demand, becomes the constraint within 12 to 24 months.
Mechanism · regulatory change → capacity constraint
What to investigate
Check whether other purchasing groups adopt the clause, count certified capacity in the region and see what single-use device makers already expect.
Kill if · no other purchaser adopts it by Q3
Anatomy of a hypothesis
Every case answers the same six questionsObserved change Observed
Procurement now admits certified reprocessed devices in a major tender lot.
Causal mechanism Inference
Cheaper certification moves volume to reprocessed devices, and certified capacity becomes scarce.
Who could be affected Inference
Gain: certified reprocessors, sterilization equipment and consumables. Pressure: distributors of high-volume single-use instruments.
Horizon
12 to 24 months, tied to the next tender cycles.
Evidence against Disputed
A hospital association warns that audit and liability costs may cap adoption.
Missing before acting
A second purchaser adopting the clause, regional capacity data and what incumbents already expect.
Signal
A data-center operator’s tender for a new campus requires direct-to-chip liquid cooling in every hall, with dripless couplings certified to a new connector standard.
Quoted · tender notice · section 5
Connections
A standards body publishes that connector standard the same month. A specialist coupling maker reports order lead times rising from 8 to 26 weeks.
Quoted · 3 independent origins
Opportunity hypothesis
If more operators specify certified couplings, connector qualification, not chips or power, becomes a gating item for new AI halls over the next 12 to 18 months.
Mechanism · new standard → component constraint
What to investigate
Count certified coupling suppliers, check whether other tenders cite the standard and compare with cooling integrators’ delivery guidance.
Kill if · lead times fall below 12 weeks
Anatomy of a hypothesis
Every case answers the same six questionsObserved change Observed
A new campus tender makes certified liquid-cooling couplings mandatory.
Causal mechanism Inference
A new connector standard narrows the qualified supplier base just as liquid-cooled capacity grows.
Who could be affected Inference
Gain: certified coupling and manifold makers. Pressure: integrators exposed to late components, and air-cooling-only suppliers.
Horizon
12 to 18 months, across the next campus builds.
Evidence against Disputed
A large operator says it qualifies couplings in-house and sees no shortage.
Missing before acting
The number of certified suppliers, lead times from a second source and whether integrators already expect it.
The platform
An intelligence workspace, not another feed.
Explore signals, follow events, inspect relationships and investigate hypotheses in one place. Antegraph keeps the underlying evidence connected, so your team can ask better questions instead of reading more summaries.
This week
synthetic dataCertified reprocessing capacity may become the constraint
3 origins for · 1 against · new supporting filing
Certified liquid-cooling couplings may gate new AI halls
3 origins for · 1 against · lead times rising
A contract manufacturer shortens accelerator lead times
1 origin · needs an independent second source
What changed
Reprocessed devices admitted in a regional tender
“Up to 30% of the volume in lot 4 may be supplied as certified reprocessed devices, subject to the accreditation set out in annex C.”
| Document | Origin | Relation |
|---|---|---|
| Tender · lot 4 | Purchaser | Evidence |
| Tender · lot 7 | Same purchaser | Same event |
| Regulator guidance | Regulator | Related event |
Certified capacity becomes the constraint
Open- Confirm the clause text in both lots
- Check the regulator guidance and its effective date
- Count certified capacity in the region
- Find a second purchaser using the clause
- Compare with what distributors expect
Counter-evidence. Hospital association: audit and liability costs may limit how much volume moves to reprocessed devices.
Hypothesis ledger
synthetic data| Hypothesis | Horizon | Kill if | Status |
|---|---|---|---|
| Certified capacity becomes the constraint | 12–24 mo | No second purchaser by Q3 | Open |
| Coupling qualification gates new AI halls | 12–18 mo | Lead times fall below 12 weeks | Open |
| Accelerator lead times keep falling | 3–6 mo | Lead times rise next quarter | Falsified |
| Dredging delays raise inland freight rates | 6 mo | Schedule restored by March | Expired |
Misses stay on the ledger next to the hits.
What it replaces
Replace the feed.
Keep the evidence.
Newsletters, alerts and analyst notes tell you what someone thinks happened. Antegraph keeps the original document, the exact quote and the path from source to hypothesis, so you can check the reasoning instead of trusting it.
Replaces reading
- Morning newsletters
- Keyword alerts
- Forty open tabs
- Summaries you cannot trace
Replaces guessing
- Conviction nobody can explain
- Single-source headlines
- Claims that were never re-checked
- Theses with no exit condition
Gives you instead
- A workspace you can question
- Evidence you can open
- Hypotheses you can track to an outcome
The intelligence engine
An evidence engine for emerging opportunities.
Discover. Verify. Connect. Prioritize. Investigate. Each stage preserves the distinction between what the sources establish and what the system infers.
| Source | Status |
|---|---|
| Regional tender portal | Acquired |
| Device regulator notices | Acquired |
| Company filings | No fresh entry |
| Trade association news | Failed · reported |
Each fetch keeps its raw bytes, a content hash and the time it was received.
Finds what changed.
Antegraph reads primary sources and keeps a receipt for every request: what was asked for, what came back and when. A source that fails is reported. It is never quietly dropped or replaced.
“Up to 30% of the volume in lot 4 may be supplied as certified reprocessed devices, subject to the accreditation set out in annex C.”
Keeps only what it can quote.
An observation is accepted only when it matches an exact passage in the original. Documents that report the same thing resolve into one event, so repetition is never mistaken for confirmation.
| Event | Origin | Relation |
|---|---|---|
| Reprocessed devices admitted | Purchaser | Supports |
| Certification shortened | Regulator | Supports |
| Capacity expansion | Reprocessor | Supports |
| Adoption may be limited | Hospital assoc. | Disputes |
Three websites repeating one press release count as one origin.
Connects the evidence.
Events are linked to the companies, technologies and hypotheses they concern. Support is counted by independent origin, not by the number of documents or websites.
Decides what deserves your time.
Hypotheses compete for a fixed attention budget, ranked by their impact if true and by the quality of the evidence behind them. Volume never buys a place.
- Mechanism
- Regulatory change → capacity constraint
- Affected
- Gain: certified reprocessors. Pressure: single-use distributors.
- Horizon
- 12 to 24 months
- Still needed
- A second purchaser adopting the clause
- Certified capacity in the region
- What incumbents already expect
- Outcomes
- ValidatedFalsifiedExpiredInsufficient evidence
- Action
- Research, not execute
Says what could follow, and how to check.
The top hypotheses become research cases: the mechanism, who could be affected, what is still missing and what would prove them wrong. They start your analysts’ work. They are not trade signals.
Evidence graph
Discover connections others miss.
Antegraph surfaces candidate relationships across events, companies and industries, then shows what each one rests on. Every link says whether it was observed in a source, inferred, disputed or only suggested, so your team investigates the connection instead of trusting it.
How Antegraph finds connections
Event-centric evidence graph
Explicit relationships between events, companies, technologies, documents, contradictions and hypotheses. Every relationship keeps its provenance and the time it became known, so a past view never leaks later information.
Semantic & graph embeddings Graph embeddings experimental
Passage embeddings find related developments even when sources use different words. Graph embeddings may suggest related neighborhoods worth reviewing. Neither is used to decide that two documents describe the same event.
Hybrid retrieval
Keyword and vector search, entity and time filters, and graph traversal, combined with a ranking you can inspect. Results always come back with their original citations.
Similarity suggests where to investigate. It is never treated as independent confirmation or as proof of causality.
Open by design
Your intelligence. Where decisions happen.
Work in the Antegraph app, access your entitled research through an API, or let your agents ask evidence-grounded questions through MCP. Send selected updates into the tools your team already uses.
The main workspace for research, review and hypothesis tracking.
Scoped read access to observations, events, relationships and hypotheses.
Let authorized agents query evidence with your permissions.
Notify your systems when a hypothesis changes materially.
Selected updates in the channels your team reads.
A digest for people who work from their inbox.
Selected updates for self-hosted teams.
Your own destinations through the API and webhooks.
Which events weakened the reprocessing hypothesis this month?
// tool call explore_relationships({ "hypothesis": "hyp_reprocessing_capacity", "relation": "CONTRADICTS", "since": "2026-10-01" })
One source disputes it. A hospital association says audit and liability costs may limit adoption, and it is recorded as counter-evidence with the original quote.evt_hospital_statement
Read-only tools, scoped to your workspace and plan.
GET /v1/hypotheses/hyp_reprocessing_capacity?include=evidence { "statement": "Certified capacity becomes the constraint", "status": "open", "horizon": "12-24 months", "evidence": [ { "event": "evt_devices_admitted", "relation": "SUPPORTS" }, { "event": "evt_hospital_statement", "relation": "CONTRADICTS" } ], "kill_condition": "No second purchaser adopts the clause by Q3" }
Every object returns with its citations and the time it became known.
POST https://your-app.example/hooks/antegraph { "type": "hypothesis.updated", "hypothesis": "hyp_reprocessing_capacity", "change": "contradicting_evidence_added", "evidence": "evt_hospital_statement", "idempotency_key": "a41f…", "attempt": 1 }
Subscription filters, retries and idempotency keys, so a change is acted on once.
Developer interfaces and third-party integrations are part of the product roadmap. Availability will vary by release and plan.
Traceability
Evidence you can challenge.
Origin, quote, independence, contradictions and unknowns stay visible on every hypothesis.
Synthetic dataGrounded
- Original stored as fetched
- Exact quote matched
- Independent origins counted
- Contradictions recorded
- Your decision
Contradictions are evidence
Built so that a denial weakens the claim it disputes and you see both sides.
Unknowns stay visible
- Certified capacity in the regionUnknown
- A second purchaser adopting itUnknown
- What distributors already expectUnknown
Each hypothesis lists what is missing before anyone should act on it.
Origin, not host
Two documents from one purchaser count once. Different websites are not different sources.
Plans
From early signals to forward-looking intelligence.
One evidence engine. Three ways to work with it. Coming soon.
€29/ month
PLANNED LAUNCH PRICING
Detect what matters.
For teams that need to know what changed and where it came from.
- Curated brief
- Prioritized signals
- Linked primary evidence
- Chronology, developments and uncertainty
€99/ month
PLANNED LAUNCH PRICING
Connect the dots.
For teams that need to see how developments connect and why they matter.
- Everything in Observe, plus
- Contextual research
- Event and entity links
- Explainable themes and timelines
- Evidence graph and contradiction analysis
€249/ month
PLANNED LAUNCH PRICING
See what’s next.
For teams that investigate what could happen next, and test it.
- Everything in Understand, plus
- Forward-looking hypotheses
- Scenario monitoring and leading indicators
- Falsification tests
- An inspectable outcomes ledger
Enterprise
FutureConnect intelligence to your reality.
Potential private-data integrations, custom workflows, organizational exposure mapping and enterprise controls. Custom pricing.
Published planned monthly pricing in EUR, subject to change before launch. No subscriptions or payments are available yet. Every plan runs on the same intelligence engine.
Method
Built to measure what matters.
We intend to record hypotheses before outcomes, preserve the original evidence and measure both successes and misses. Discovery quality, timing, false positives and whether a finding improves a decision matter more than an impressive accuracy number.
Write it down first
Each hypothesis is recorded before its outcome is known, with a horizon, a kill condition and its original evidence.
Compare with a baseline Calibrating
The evaluation protocol reserves held-out signals, so results can be compared with a baseline instead of the system’s own memory.
Report the misses
Results will be reported alongside misses, false positives and unresolved cases. No accuracy figure appears here until enough hypotheses have resolved.
The questions investment teams ask first.
“Isn’t this another AI news summary?”
A summary compresses what was said. Antegraph keeps what was said, in the original, connects it to other evidence and asks what could follow. Nothing becomes a hypothesis without a quote you can open.
“What is available today?”
An experimental engine that runs internally in test cycles on a small set of sources, with its output reviewed by hand. The recurring SaaS service on this page, including the workspace, API, MCP server and integrations, is planned and separate. Beta access is not yet available; the waitlist will open soon.
“Are these investment recommendations?”
No. Antegraph produces research hypotheses: what may be changing, who it could affect and how to test it. What to do about it stays with your team.
“How does semantic search avoid false connections?”
Similarity only suggests where to look. Two documents count as the same event, or as independent support, only after their origin and content have been checked. A similarity match is shown as a lead, never as evidence.
“How do you know it works?”
We don’t yet, and we say so. Hypotheses are recorded before outcomes and scored after, against a baseline. Results will be published once enough have resolved, with the misses.
“How much will it cost?”
Planned launch pricing: €29/month Observe, €99/month Understand, €249/month Anticipate. Enterprise pricing will be custom and available later. Prices may change before launch; there is no checkout or billing today.
Why we built Antegraph
Conviction should cost evidence.
Most decisions are made on summaries of summaries. By the time a change reaches a newsletter, the people who read the original have already acted.
Antegraph is built for the slower, surer habit: read the source, write the hypothesis down, and let reality score it.
Beta waitlist · Coming soon
Be early to what matters.
Antegraph is preparing a beta of its evidence-backed opportunity intelligence platform.
- Discover the planned Observe, Understand and Anticipate plans
- Get an invitation when beta access becomes available
- No payment or subscription commitment
The waitlist opens once secure registration and our privacy notice are ready. Beta access is not available yet.