DaVinci presents
Synapse
Intelligence that gains, keeps and compounds experience.
A deep tech frontier AI lab. Pre-seed, coming out of stealth.
“DaVinci enables you to curate intelligence into your own compounding asset. Synapse becomes smarter, and faster, striving towards a ‘tokenless’ AI experience.”Nick Joughin, CEO.
Patent pending
An Overview of Synapse
DaVinci presents Synapse
Intelligence should be allowed to keep its experience.
And the experience should belong to the businesses and people who curate it.
Nick Joughin · CEO · DaVinci (Pty) Ltd
The problem
Seven problems the AI industryhas not solvedand what solving them takes
AI is stateless
Context and memory are workarounds; the reasoning is never kept. The hundredth solve costs the same as the first, so compute becomes a physical and financial constraint.
An AI that keeps its experience, so it never recomputes what it has already solved.
AI is a black box
Probabilistic weights, with almost no view of how it reaches a conclusion. Practically unauditable.
An AI that can be audited down to the inference level.
IP and security
Who owns the outputs, and who can reach the IP? Enterprises and nation states do not fully trust providers. Dependence on one provider creates lock-in risk.
Potentially no exposure to prompt injection, and an intelligence that belongs to the user.
AI is not sovereign
A way for users to curate their own intelligence, without dependency on compute.
Environment and infrastructure
Power, water and carbon at data-centre scale. Communities push back, and the build is capital intensive.
A solution that takes load off compute.
Access to intelligence
Ring-fenced to US providers or open weights. Emerging markets struggle to enter as easily as the US or EU, and open-weight dependency adds IP risk. Owned by few, needed by many: who decides access and censorship?
Intelligence, and its curation, put back in the hands of the user.
Pricing lock-in
Tokens get cheaper per capability, but frontier prices hold, so staying ahead means paying frontier prices.
Lower cost, by cutting the compute behind it.
The solution
DaVinci presents Synapse
Self maintaining intelligence that gains, keeps and compounds experience
What it is
The Experiential Reasoning Substrate.
How it works: one door in, three doors out
Measured
AI is stateless
AI is a black boxRefuses rather than guesses. Every serve carries a readable trace.
IP and securityPotentially nothing to inject, poison, leak or steal. The client owns its stored inference; its facts never leave.
AI is not sovereignRuns inside the client’s boundary, with any model or none. The known path runs air-gapped and locally.
Environment and infrastructureCompute, power, water and carbon fall with the share of work served without a model call.
Access to intelligencePeople and entities curate their own intelligence, in a library they own. Cost per query falls with use.
Pricing lock-inThe proof
From a cold start to 84 % reuse in fifteen days
The test Fifteen simulated days against a live commercial LLM provider.
Cold start The system began blind, with no prior data, no stored reasoning, no clusters, no taxonomy and no human labels. Everything it served, it learned in the run.
| All fifteen days | ||
|---|---|---|
| Reuse | ||
| No model call | ||
| Fewer tokens |
In theory: as a problem set saturates, every query becomes known, and serving approaches zero tokens, a tokenless response.
It organised and policed its own knowledge without supervision.
The potential impact
What would the impact be if Synapse was deployed globally by 2030 ?
fewer tokens over fifteen days
From a cold start
US-average cars off the road for a year
US homes' electricity for a year
US households' home water for a year
Who owns what
You own your stored inference.
We own the engine that makes it work.
Their inference library
The reasoning their own queries produce, captured and kept. Minted from their own traffic, curated by their own people.
- Served locally, inside their own boundary
- Observable, traceable and auditable
- Their data stays theirs: facts never leave
LicenceEngine licensed to the customerThe Synapse engine
The software that extracts, stores and reuses reasoning, and makes every library operable.
- The ability to extract, store and reuse
- Every update to the engine
- The IP: patent pending
The link between them
DaVinci · Our vision
Curate your own intelligence
Intelligence that gains, keeps and compounds experience.
Intelligence used to be owned by a few. Now every business, nation, family, person and community can curate and develop an intelligence of its own: one that gains experience over time.
Sovereign intelligence, from a person to a nation
The reasoning a company already pays for, kept as an asset its own people own and grow.
Its own intelligence, under its own law, with no foreign kill switch.
The way a family thinks, handed down for generations to come, like an estate or a library.
A lifetime of your own reasoning, carried into every job and still yours when providers change.
Towns, co-ops, schools and guilds that remember what their members worked out, kept local.
Hospitals, courts, universities and public bodies whose memory outlasts every change of staff.
Synapse todayA known query needs no model call and spends zero tokens.
The paper
Six scales of curation, and what comes next
DaVinci’s Synapse enables you to curate your own intelligence, at every scale.
From the few to everyone
Intelligence is rented from a few, and nothing it learns stays with the business or the person who paid for it.
Each business, nation, family, person and community curates an intelligence that gains experience.
“DaVinci enables you to curate intelligence into your own compounding asset. Synapse becomes smarter, and faster, striving towards a ‘tokenless’ AI experience.”Nick Joughin, CEO.
It keeps the reasoning, not the answer
Rebuilt fresh for each new query. Nothing canned is ever served.
Five things that make intelligence ownable
The client owns the stored inference; DaVinci owns the engine.
Runs inside the client’s boundary, any model or none.
Cost per query falls with use.
Every idea that follows rests on these five.
It would rather say “I don't know”
With this question’s specifics bound in, checked by a quality gate.
A fresh answer is paid for, never a near-miss, never a guess.
Synapse todayIt refuses rather than guesses. Its worst case is a full-price correct answer.
Imagine a day in that future
A new hire asks how the team handled last winter’s outage. The answer is rebuilt from what a colleague who has since left worked out.
A cooperative’s newest member asks how the group handled last year’s drought, answered from what its members worked out together.
A daughter asks how her grandfather reasoned through selling the farm. The family library answers for her situation.
Every serve is graded and the library is consolidated, a little more experienced by morning.
Synapse todayEvery serve is graded; the library is consolidated nightly.
The questions it raises are opportunities
Wills, trusts and estates that carry a family’s reasoning alongside its property.
A clear line: the owner holds the library; the engine stays separate.
Curation as a family, board or public practice, with a readable trace for every answer.
An intelligence that refuses rather than guesses earns trust that lasts.
Each question is a place where something lasting can be built.
Curate your own intelligence
We allow people to curate their own intelligence.
Intelligence should be allowed to keep its experience. And the experience should belong to the businesses and people who curate it.
Nick Joughin · CEO · DaVinci (Pty) Ltd
Team
Built from zero by four founders
DaVinci is a deep tech frontier AI lab. Pre-seed, coming out of stealth.
Nick Joughin CEO · co-inventor
Oscar Downing CTO · co-inventor
Technical architect of the Synapse engine; named co-inventor on the patent application.
Redford Muir CFO
Finance and commercial structuring. Anchors a deal-assurance and integrity practice (AML, FCPA, counterparty and forensic investigation to DFI and institutional-investor standard). Member of CIBA and the IIA.
Ingrid Thomas COO
Interested?
Three paths. Choose the one that fits.
Thank you.
Your design partnership request is in. The DaVinci team will read it closely and reply by email to set up a first conversation.
Thank you.
Your investor enquiry is in. The DaVinci team will review it and reply by email to book a meeting.
Thank you.
You are on the waitlist. We will email you as Synapse opens to more people and organisations.
FAQ
Questions, answered
What DaVinci is building, what has been measured, and how to get involved.
What is DaVinci?
DaVinci is a deep tech frontier AI lab, pre-seed and coming out of stealth. It was founded in April
What is Synapse?
Synapse is the Experiential Reasoning Substrate. It sits between the user, software application or agent and the large language model, and stores inference as a compounding asset for whoever owns it, so a problem that has been solved needs no model call and no tokens again.
How does Synapse answer a question?
One door in, three doors out. A solved problem is answered from owned reasoning with no model call and zero tokens. A partly solved problem calls the model only for the missing step, cutting tokens by up to
What has been measured so far?
In a fifteen-day test against a live commercial LLM provider, starting cold with no prior data or stored reasoning, daily reuse rose from
Who owns the stored reasoning?
The customer owns their inference library: the reasoning their own queries produce, served inside their own boundary, and their facts never leave. DaVinci owns and licenses the Synapse engine that extracts, stores and reuses that reasoning.
Does Synapse guess when it is not sure?
No. It refuses rather than guesses. When a question does not match exactly one owner in the library, it abstains and a fresh answer is paid for, so its worst case is a full-price correct answer.
Does Synapse depend on one AI provider?
No. It runs inside the client's boundary with any model or none, the known path can run air-gapped and locally, and stored reasoning survives a change of provider.
Is Synapse patented?
A patent application is pending, covering deterministic computational reasoning reuse in AI inference infrastructure.
How can I get involved?
There are three paths: become a design partner, join the waitlist, or, for investors, tell us about your fund. Each has a short form on the Interested? page.
Can Synapse be run locally?
Yes Synapse can run locally.
Is Synapse open to investment?
Synapse is officially launched and out of stealth and we are actively seeking investment.
Isn't Synapse just a semantic cache?
No, not even close. Semantic cache's store answers. It has no adaptability and serve stale and unrelated answers. Synapse stores reasoning, it doesn't replay or serve a stored trace. It applies the stored reasoning to the relevant query and its specifics and reconstructs an answer on a query to query basis.