New Negotiated Autonomy — our trust protocol for high-stakes agents

4Gentic — a thesis-driven AI company

The first wave of AI was about conversations. The next wave is about commitments.

4Gentic is a thesis-driven AI company building long-running autonomous AI systems that assess, coordinate, act and dynamically adapt to produce specific outcomes.

Start a conversation

01 / Thesis

Important work does not happen in a single prompt.

Most AI systems today are built for short bursts of interaction: answer a question, summarize a document, complete a task. But many of the most valuable workflows in the real world do not behave like that. They unfold over weeks or months. They require continuity, judgment, accountability, and the ability to adapt as new signals arrive.

That is where conventional AI starts to break down.

When the work is longitudinal, high-stakes, and outcome-bound, the problem is no longer generating a clever response. The problem is staying aligned to the objective over time. Autonomously.

Session-based — context lost between each interaction Outcome-owned — state preserved, governed, and resumed
checkpoint human review resume

02 / What we build

AI systems built for outcomes, not sessions.

4Gentic develops governed, long-running AI systems designed to stay engaged for the full duration of a goal. Instead of treating intelligence as a sequence of disconnected prompts, we treat it as an ongoing commitment: tracking progress, coordinating work, preserving context, and adapting behavior as conditions and/or rules change.

These systems are designed for environments where memory matters, policy matters, and human oversight still has an essential role.

  • 01

    Persistent orchestration

    Systems that stay aligned to the goal over time rather than resetting with every interaction.

  • 02

    Structured memory

    Context that is refined, preserved, and resumed without collapsing into noise or runaway cost.

  • 03

    Dynamic governance

    Rules and policies that can be updated as the system runs, without rebuilding the system from scratch.

  • 04

    Human-aware execution

    Autonomy where it is useful, human judgment where it matters. Most importantly, an AI that can say “I don't know”.

    Negotiated Autonomy →

03 / How it works

A different operating model for AI.

4Gentic's systems are built around a simple idea: separate the agents that think from the agents that do.

System architecture showing orchestration, execution, and governance layers
Plate 01 — Operating model orchestration · execution · governance
  1. 01

    Orchestrators

    Persistent agents that hold the objective over time, monitor progress, and decide what should happen next.

  2. 02

    Executors

    Specialized agents spun up for specific tasks, then terminated when the work is done.

  3. 03

    State and context

    A structured memory layer that allows the system to checkpoint, resume, and maintain continuity across long-running workflows.

  4. 04

    Rules and policy

    A governance layer that makes behavior steerable, auditable, and better suited to high-stakes environments.

The result is an AI system designed not just to respond, but to stay accountable to an outcome.

04 / Ventures

Early proof points of a broader model.

4Gentic's ventures are not disconnected products. They are early expressions of the same underlying thesis: that AI becomes more valuable when it can persist, reassess, and act in service of a real-world outcome.

Venture / 01 · Real estate

Kinob

Kinob is our autonomous leasing platform — built around the full leasing lifecycle, from inquiry to placement, with AI handling the long-running workflow and human agents stepping in where judgment or licensed execution is required.

kinob.ca
Venture / 02

Clinical trials

A longitudinal participant support and retention model for clinical trials, where AI stays engaged between visits, detects drift, surfaces barriers, reassesses adherence over time, and creates better handoffs to coordinators before dropout happens.

Partners with usecases and workflows in diverse industries such as education and mining are actively leveraging our platform.

05 / The 4Gentic advantage

Building a company around a shared autonomy core.

Shared core 4Gentic
Orchestration long-running flows
Governance rules and audit
Execution domain actions
Memory persistent state

Each 4Gentic venture is a domain-specific expression of a broader operating model. The underlying architecture compounds: orchestration, memory, governance, and execution patterns can be carried forward, refined, and adapted across categories.

That means 4Gentic is not simply building one application at a time. It is developing a repeatable foundation for fully autonomous AI workflows in markets where continuity, accountability, and intelligent coordination matter.

This is how a thesis becomes a company.

06 / Research — trust protocol

Autonomy is not a switch. It's a negotiation.

Every team deploying agents in high-stakes domains ends up choosing between agents that act dangerously, agents that escalate everything, and brittle hard-coded rules. All three treat the boundary between autonomy and oversight as something you configure once. It isn't — it's a relationship.

Negotiated Autonomy is our answer: a protocol where the agent proposes its own Trust Contract, the human adjusts it, and operating evidence renegotiates it over time — tested across 4Gentic's ventures and published for others to pressure-test.

The agent proposes. The human adjusts. The system learns. The boundary lives.

07 / Featured open source project

Writing AI-assisted work is cheap. Reviewing it is the work.

Anyone using AI tools at work has seen their role quietly reverse. We are all moving from doing to managing — from drafting to reviewing, from producing to deciding what is trustworthy. New roles need new tools.

Obelus is a reviewer's surface for AI-assisted papers. Source editor, git-style diff review over project folders, Claude Code built in. Available as a desktop app, or directly in the browser as a Claude Code plugin that runs offline after first load. Paper bytes never leave the device. Open source, MIT-licensed.

08 / Company

Founders-led. Thesis-driven. Built for what comes next.

4Gentic is led by experienced founders with a track record of driving adoption of new technologies and a clear point of view on where AI must go next: beyond one-off interactions and toward systems that can sustain responsibility over time.

The company's work sits at the intersection of product thinking, systems architecture, and real-world commercial applications.

If you're thinking seriously about what comes after conversational AI, we should talk.

We're interested in exploring opportunities with exceptional builders, aligned investors, and partners operating in categories where continuity, judgment, and governance matter.