The words, defined once.
A site that coins terms owes the reader definitions. Each one below links to the page where the term does its work — so you can check the definition against the thing itself.
How we work
- Agentic AI
- Seen on: Services
- Software that plans and carries out multi-step work — reading context, choosing actions, using tools — instead of answering one prompt at a time. Every agentic system we ship keeps a human gate.
- The human gate
- Seen on: Method
- A named person approves or stops each phase. An agent proposes; a person approves. The gate is architecture, not a promise — the system cannot skip it, and scope, price and anything that ships stay with a name.
- The discovery week
- Seen on: Services
- Week one of every engagement: we map the process, the data and the risk. It ends with a written brief, a capability map, and a pilot scope with a price — in writing.
- The decision log
- Seen on: Method
- The running record of why a system grew the way it did — kept during the build, handed over at the end. It is what makes a handover honest in both directions, and what makes a security review fast.
How we mark what is true
- The honesty grammar
- Seen on: Method
- Three marks, read the same way in every context: ● live, offered, or in production today · ◐ building or under evaluation · ○ chartered or pre-launch, not yet offered. A mark is never upgraded for a nicer sentence.
- The Charter Index
- Seen on: Services
- The company's registered main object, typeset as a capability index — fourteen lines, each carrying an honest mark. It publishes the gap between what the charter permits and what we offer today instead of hiding it: the charter is wider than the brochure, and both are true.
- The Law of Proof
- Seen on: Services
- We will not show client logos we do not have. Until the logos are real, the five self-funded products are the evidence — real screens, honest marks, openable by anyone — and every number on this site is one we can defend.
- Machine gates
- Seen on: Method
- Checks that fail hard rather than filing a warning: the colour audit, which fails the build itself, and the zero-leak language gate, which fails the QA run against the built site — blocking every English–Hindi crossing outside one founder-sealed allowlist. No exceptions, including for us.
Inside the products
- Named specialist agents
- Seen on: Method
- Agents with one job each and a name, rather than one general assistant. Every product we ship runs them — the cross-portfolio signature, alongside a memory layer and human gates.
- The Mandala
- Seen on: InkSmith
- InkSmith's arrangement of eleven agents around a writer: two sovereigns — Chakravarti for deep research, Lakshya for launch strategy — and nine rishis carrying outline through final review. The writer holds the pen.
- The Gita (story memory)
- Seen on: InkSmith
- InkSmith's cross-chapter story memory: the structure that keeps a long book consistent with itself as it is written, chapter after chapter.
- Artifact Dependency Graph
- Seen on: TheVibeManager
- TheVibeManager's map of how a project's artifacts depend on each other, so that changing one thing does not let the rest drift out of agreement with it.
- Harness engineering
- Seen on: TheVibeManager
- TheVibeManager's own framing of what makes AI reliable: the harness around the model — the gated lifecycle, the Artifact Dependency Graph and the six-agent relay — rather than the prompt.
The name
- adhunik
- Seen on: About
- Hindi, adjective: modern; of the present age. The company took the word as a name, and as a standard it agreed to be held to, in writing. The Hindi self of this site renders it in its own script.
If a word on this site is not in this list and needed defining, that is a defect. Tell us: info@adhuniklabs.com.
Tell us what you're building.
Made with love in Naya Bharat