Current projects

Projects, demonstrations and research

Practice by the founder of Time into Form: products, an operating system, a manual demo and AI research. Each has its own results, evidence and readiness level.

Current portfolio

Multi-agent event processing

AgentDesk

When messages arrive from different sources, the system helps identify what matters and prepare the next step. Decisions and sending remain with the owner.

Founder’s role
Creator, architect, and developer
What this demonstrates
An agent registry, persisted state, five portfolio adapters, shared cycle health, and an independent Watchdog connect domain review with verifiable system state.
Verified status
Private product · production 0.14.1 verified 19 September 2026
A DoMiSoul tube amplifier in a private listening space

Owned productUltra Hi-End audio

DoMiSoul

Presents tube amplifiers, private listening sessions, and personal audio-system setup.

Founder’s role
Founder and engineer
What this demonstrates
Connects the commercial presentation of a real product with an operating digital system.
Verified status
The site runs on a managed VPS with nginx and PHP 8.3-FPM and is published through immutable releases.
01

97.4% — why is it not enough?

One critical error in 39 cases violated the acceptance condition.

One critical error in 39 cases violated the acceptance condition.
02

A harder task needs more resources

1.48 → 15.54 seconds. Both difficulty and reasoning mode change.

1.48 → 15.54 seconds. Both difficulty and reasoning mode change.
03

No configuration passed unconditionally

At the tested difficulty level, an unconditional positive conclusion is premature.

At the tested difficulty level, an unconditional positive conclusion is premature.

Three Stage 1 results. Synthetic tasks and local tests; conclusions apply to the stated configurations and conditions. Figures are in Russian; English headings summarize each finding.

Independent applied AI research laboratory

LydStone Lab

A convincing AI answer is not necessarily reliable. The laboratory tests errors, limitations and resource use on defined tasks, so deployment decisions can be grounded in measurements.

Founder’s role
Programme initiator and author of the method and study
What this demonstrates
Seven task groups studied, from clarifying requests to handling conflicting sources and constrained text. Results and evidence boundaries are recorded.
Verified status
The first stage of the laboratory development programme is complete: a study using synthetic tasks in a local environment.

What the cases allow you to assess

Examples show product framing, implementation and operations. They support assessment of the approach without guaranteeing results for a new task.

  1. Product framing

    I connect the problem, audience, constraints, and a useful outcome before choosing technology.

  2. Development and architecture

    I bring the solution into operation with explicit boundaries, verification, and a recovery path.

  3. Positioning and promotion

    I translate product complexity into a clear offer, useful content, and a path to the next meaningful action.

  4. Operations and evolution

    I observe the system, repair failures from evidence, and evolve the product without presenting plans as current facts.

Other areas, labelled by their actual status

The evidence platform, a product in development, and future directions remain separate from the four current projects.

Meta layer · outside the primary four

Vaneev and PROOF

The founder’s personal website and the independent evidence facet of Time into Form. PROOF connects tasks, results, sources and limitations; publication lifecycles remain separate.

Open: Vaneev and PROOF

Manual demo available · product in development

Kernel

A system for collecting, checking and managing engineering data, from measurements and documents to resolved findings, approved decisions and an issued report.

Open the Kernel manual demo: Kernel

Future directions · not implemented

Meridian and Facility

International technical workflows and safe agentic AI in operations remain future directions.

Open: Meridian and Facility

A question about a result?

Ask about its evidence or discuss the conditions that would need checking for a similar task.

Ask about a case