For investors · As of October 5, 2026
BuiltGoal
Intelligence for the work of whole organizations.
We are developing an intelligence. It is meant to plan, carry out and verify the work of corporations and public agencies. A person always approves what goes out.
Since September 2026
Built as a prototypeThe control of approvals, tested in-house
Today, October 5, 2026 No customer for the intelligence yet
In developmentGoal: AGI, a general intelligence for whole organizations
01 / The bet
We are betting on an intelligence that does not just do one thing, but understands work as a whole.
Researchers call an intelligence like this AGI. We are developing it for whole organizations. Nobody can say today when it will succeed.
02 / Why now
Since 2019, leading language models have been completing ever longer tasks.
The length of these tasks has doubled about every seven months.
Source: METR, March 19, 2025, confirmed with more tasks on January 29, 2026. Software and research tasks that models complete in half of all attempts; length measured in the working time of professionals.
Our thesis
Organizations only hand over work when they can verify what the intelligence is allowed to do. That is why we are building the control first, and the intelligence is meant to earn every new permission with measured cases.
03 / What is built
What is built so far is the control.
We are still developing the intelligence itself.
A yes counts exactly once, and a copy is rejected.
Built as a prototype · tested in-house
Every entry in the log carries the verification value of the entry before it.
Built as a prototype · tested in-house
04 / The gates
These are the three things we want to prove next.
- Gate 1OpenIn a conversation, an organization should name its own work area to us.
- Gate 2OpenA paid pilot should be measured before and after.
- Gate 3OpenAt least two customers should work with it in day-to-day operation.
05 / Conversation
Come and talk to us.
We will show you the prototype, the plan and everything that is still missing.
- Horatiu-Adrian Baloi: Owner, product and build.
- Kevin Podstawka: Sales.
Or write directly to [email protected]
06 / Letter
Letter from the owner
Halle (Saale), October 5, 2026
To investors and funders
An organization does not work in one program, but between many. Each program knows only one piece of the work, and people carry the pieces from program to program. Work gets stuck along the way, because nobody sees the whole picture at once.
We are developing an intelligence meant to see this whole picture and to plan, carry out and verify the work in it. We want to start with a single task. We only take the next step when measured cases show that the step before holds. That is how we want to grow through a workflow and a department to the whole organization.
The organization decides what the intelligence may do on its own. It may do more only when measured cases without errors support it and the customer approves it in writing. People always decide about people and about everything that goes out.
We have been building this since September 2026. What is built and tested in-house so far is a prototype of the control. We are still developing the intelligence itself.
Two things can fail. The intelligence may be harder to build than we think today. And organizations may hesitate to trust a new provider with their work. We test the second one first, at gate 1, and we will say openly what comes of it.
Horatiu-Adrian BaloiOwner
- September 2026
- start
- September 14, 2026
- control of approvals in code
- September 29, 2026
- all tests re-run by us
- October 5, 2026
- status of this page
07 / Measures
What we measure ourselves by
Five numbers are public here, each with today's value and the gate at which it counts.
| No. | Measure | How we count it | Gate | Today | As of |
|---|---|---|---|---|---|
| 01 | Paid pilot engagements | Payment received for a pilot engagement | Gate 2 | 0 | Oct 5, 2026 |
| 02 | Deployments at customers | A workflow runs under contract in an organization's systems | Gate 3 | 0 | Oct 5, 2026 |
| 03 | Error-free cases at customers | In the customer's log, approved without correction | Gate 3 | 0 | Oct 5, 2026 |
| 04 | Steps up with a customer's written approval | The customer's written approval per work area | after gate 3 | 0 | Oct 5, 2026 |
| 05 | Work areas where the same software runs without rework | Another work area without a new build | after gate 3 | 0 | Oct 5, 2026 |
Every number is at zero today. We publish them so you can hold us to them.
- Gate 1 is reached when an organization names its own work area to us in a conversation.
- Gate 2 is reached when a paid pilot runs in one work area that was measured beforehand. This includes the involvement of the works council or staff council and a written agreement on where the intelligence runs.
- Gate 3 is reached when at least two customers use the intelligence in day-to-day operation after a measured pilot. This includes a monthly report on cases, errors and processing time.
08 / Limits
What is missing, and where our limits are
What is missing
- We have no customer for the intelligence yet.
- We are still developing the intelligence that plans, carries out and verifies work.
- There is no connection to a customer's programs yet.
- There is no operation and no certification yet.
Our limits
- People always decide about people, and a person approves what goes out.
- In defence we only develop for maintenance, logistics, procurement and readiness, with no weapons, targeting or effects functions.
09 / Contact
Request a conversation
Write to us if you would like to talk with us.
An introduction to a head of procurement, finance or public administration also helps us.
A person reads your message and replies to you.
Or write directly to [email protected].
SnapSite · sole proprietorship · Halle (Saale), Germany