LawMan
Indian law, answered with the source attached.
LawMan is an AI system built for one thing - Indian law. It grounds every legal claim in a real, retrieved source. When it cannot verify a claim, it says so instead of inventing one.
It is a research and drafting instrument, not counsel. LawMan researches and drafts; a qualified person reviews and signs.
Fully specified, and not yet built. Everything on this page describes the intended product, not running software.
What is LawMan
LawMan is PunkRaven's reasoning layer, built for Indian law as its first body of authority. It retrieves the governing material before answering, attributes each claim to its source, verifies every reference against the source text, and abstains when the sources do not support an answer. LawMan is specified and not yet built.
It is specialised in Indian law - statutes, case law, court procedure, and the rules and circulars that change week to week. It is not a general-purpose chatbot pointed at legal questions.
It is a system designed around a single constraint: a legal answer is only worth having if you can trace it back to the authority it came from. Everything in LawMan follows from that - how it is trained, how it retrieves, how it checks itself, and what it does when it is unsure. It is the same discipline behind how PunkRaven builds every layer of the stack.
LawMan researches, explains, and drafts. It works in Indian languages, and it reads the documents legal work actually arrives in. It is built to run on infrastructure you control, so confidential material never has to leave your hands.
Why LawMan is required
General-purpose AI fails at Indian legal work in ways that are not cosmetic.
- It invents authority.
- Fabricated sections and judgments arrive confident and correctly formatted. In law, that is not a rough draft - it is exposure.
- It does not know current law.
- Legal facts move constantly. A general model's knowledge is frozen at the moment it was trained, and it cannot tell you which parts have gone stale.
- It does not know Indian law specifically.
- Indian statutory structure, court hierarchy, procedure, and drafting convention are not well represented in models built elsewhere for everyone.
- It cannot be trusted with your files.
- Confidential material cannot be handed to an external service by anyone carrying a duty of confidence.
- It does not know when to stop.
- A general model would rather produce a plausible answer than admit it has none. That instinct is exactly backwards for legal work.
LawMan exists because each of these has to be solved deliberately, at the level of how the system is built - not patched over with a disclaimer.
How it works
Grounded, never freehand.
LawMan does not answer from memory. It retrieves the governing material first, then answers from what it found, and attributes the claim to its source. An uncited legal claim is treated as a defect, not a stylistic choice.
Every reference is checked before you see it.
Citations are verified against the actual source text automatically. A reference that does not hold up does not reach the answer.
Skill in the model, facts in the sources.
LawMan is trained to reason, to use legal language correctly, and to cite properly. The facts themselves always come from live retrieved material. That separation is why its knowledge does not go stale as the law moves.
It works step by step, and checks its own work.
Complex questions get broken down. LawMan plans an approach, gathers what each step needs, and verifies before moving on - rather than producing one confident block of text in a single pass.
Trained to abstain.
LawMan is deliberately taught to say "I could not verify this" when its sources do not support an answer. Abstention is a designed behaviour, not a failure mode. A wrong legal answer is worse than none.
Fails closed.
Verification runs in layers. If any layer cannot confirm a claim, the system withholds it rather than shipping it unverified.
Built in Indian languages.
Questions asked in an Indian language are researched against the same material and answered in that language.
Reads real legal documents.
Scanned judgments, filings, and the print quality that Indian court output actually arrives in - not just clean digital text.
Runs where you decide.
LawMan is built to be deployed on your own infrastructure. Public law can live on shared systems; confidential material never has to.
A human still signs.
LawMan researches and drafts; a qualified person reviews and takes responsibility. That review step is part of how the system is designed to be used.
When to use LawMan, and when not to
When should you use LawMan?
- You need to find the governing provision or authority on a point, see where it comes from, and know whether that position still holds today or what it was at an earlier time.
- You are working through material in an Indian language, or across several of them.
- You need to work with documents that cannot be sent to an outside service.
- You are drafting and want a first pass that already carries its references, with the underlying material laid out so you can check it yourself.
When should you not use LawMan?
- You need legal advice. LawMan is a research and drafting instrument. It is not counsel and does not replace professional judgement.
- The output will go out unreviewed. Consequential work needs a qualified human sign-off.
- The question is outside Indian law. LawMan is specialised, and specialisation is the point.
- You want a confident answer more than a correct one. LawMan will tell you when it cannot verify something, and that is the behaviour it was built for.
Specified, not shipped.
An AI system for Indian law, grounded in the source and honest when it cannot verify.
LawMan is a specification, not a service you can call today. If you hold a body of authoritative material that has to be reasoned over this way, we would like to hear about it: what the corpus is, where it is allowed to live, and who signs off on the output.