FounderHub journal

Notes from the work.

Research reports, engineering decisions and practical views from inside a working AI innovation company.

Raising Sprout: what deterministic reasoning looks like in practice

A progress report on a symbolic reasoner being developed as an observable educational experiment.

The work

Sprout is a deterministic, zero-GPU symbolic reasoner. It is being developed in public as an engineering and educational experiment, not presented as a finished general intelligence system. That distinction matters. The point is to make progress measurable and failure visible.

What the evidence says

As of 29 July 2026, the live system held 93,429 facts and had recorded zero hallucinations across 251 adversarial tests. Its current capability is estimated at an early-to-mid Grade 2 level. Those numbers are a bounded system observation, not a claim that Sprout is broadly intelligent.

Why FounderHub is doing it

Most AI development hides uncertainty behind fluent output. Sprout explores the opposite route: explicit knowledge, deterministic behaviour and testable progress. The research informs FounderHub's wider work on governed reasoning, evidence-led learning and dependable AI operations.

Grant intelligence needs verification, not another directory

Finding an opportunity is useful. Knowing whether it is real, relevant and worth the work is the actual product.

The problem

Funding websites tend to stop at aggregation. They collect titles, deadlines and links, then leave founders to determine whether an opportunity is current, whether the organisation is eligible and what evidence a credible application needs.

The FounderHub approach

Our opportunity engine tracks official UK and Manchester sources, verifies material fields and separates open opportunities from the archive. The founder workspace then connects discovery to preparation: profile information, saved opportunities, evidence and draft application answers.

The outcome

The objective is not to create the largest list. It is to reduce wasted applications and help organisations put serious effort behind opportunities they can credibly win. Funding intelligence belongs inside the wider innovation path from idea to product and growth.

Governance has to exist before autonomous systems enter production

An autonomous system is only useful when its actions can be bounded, evidenced and challenged.

Control is part of the product

Governance should not be added after an AI system begins acting on real infrastructure, financial workflows or customer data. Identity, authority, evidence and approval boundaries have to be present in the execution path itself.

Evidence before confidence

FounderHub's engineering principle is straightforward: evidence before opinion, deterministic before probabilistic, and governance before autonomy. A system should not receive operational responsibility because a demonstration looked convincing.

Commercial relevance

This is not abstract safety work. Organisations adopt automation when they can understand what happened, why it happened and who had authority. Governed systems reduce operational risk and make AI adoption easier to defend to customers, auditors and leadership.

Work with FounderHub

Research should lead somewhere.

[email protected]