All lab notes
What ciphers is building
ciphers is a Mumbai software and AI lab researching consequential Indian workflows and turning the strongest product theses into focused experiments, beginning with Startaflo.
ciphers is a Mumbai software and AI lab researching consequential Indian workflows and turning the strongest product theses into focused experiments, beginning with Startaflo.
Start a conversation
Tell ciphers™ about a consequential workflow, a collaboration, or a question worth thinking through. We read every serious note.
business@ciphers.techciphers™ is an independent software and AI lab in Mumbai, operated by Ciphers Tech Labs LLP. We study a small number of consequential problems, build our own product experiments, and let evidence decide which ones deserve to become enduring companies or products.
“Lab” is not a decorative word here. It describes the operating model. We are not a general software agency waiting for any brief, and we are not a gallery of weekend demos. We choose a problem, research the work around it, form a product thesis, build carefully, and remain willing to narrow, change, or stop when the evidence is weak.
Our current experiment is Startaflo™, a proposed founder operating system for the administrative life of an Indian company. It is at public-waitlist stage and is not publicly released.
Important work in India often runs across a difficult combination of formal rules and informal coordination. A process may involve a government portal, a professional adviser, an internal operator, a spreadsheet, a set of PDFs, several messages, and someone who remembers what happened last time.
Each individual tool can appear adequate while the end-to-end job remains fragile. Information exists, but context is scattered. Status exists, but only after somebody asks. A final document exists, but nobody can say which folder contains the accepted version. The most capable person in the room becomes the integration layer.
We look for workflows where that fragmentation has a real cost: lost attention, avoidable delay, uncertain responsibility, repeated interpretation, weak institutional memory, or decisions made without the full record.
That does not mean every untidy process needs a new product. A problem earns our attention when three things are true:
Better software is not simply a cleaner interface over the same fragmented work. It should understand the object at the centre of the workflow, preserve relevant context, make state and responsibility legible, and produce a record people can trust later.
For some products, that means a system of record: what is true, what changed, and what evidence supports it. It may also mean a system of work: what needs to happen, who owns it, what is blocking it, and what completion means. AI can add a system of assistance where interpretation, retrieval, or repetitive preparation creates avoidable effort.
The order matters. A probabilistic layer cannot rescue a product that has not modelled its rules, permissions, states, and sources of truth. We use deterministic software where the result must not drift. We use AI where uncertainty can be bounded, the output can be checked, and the user remains in control.
Building for India is not a localisation exercise performed after the main product is complete. Regulation, trust, language, service-provider relationships, device constraints, payment habits, organisational structure, and the way responsibility moves between people can change the product itself.
Taking that context seriously does not justify weaker craft. The standard for clarity, reliability, accessibility, performance, and security should be global. The context is Indian; the quality bar is not regional.
Running a company creates administrative work long after incorporation. Founders have to understand what applies, coordinate filings and registrations, gather inputs, follow up with professionals, and retain the acknowledgement, certificate, resolution, letter, or contract that proves what happened.
Startaflo’s thesis is that these activities should belong to one coherent company record. A founder should be able to understand the relevant event, see the inputs and ownership, follow execution, and retrieve the result without reconstructing the process from messages and folders.
That is the direction, not a claim that the product already exists in public. The scope, workflows, and delivery model may change as we learn from founders, operators, and appropriately qualified professionals.
We will not invent customers, traction, pricing, launch dates, access, or shipped functionality. We will not describe a possibility as a feature. We will not use AI as a substitute for professional judgement in legal, tax, accounting, company-secretarial, financial, safety, or other high-impact work.
Public pages stay conservative on purpose. Ambition is useful; ambition presented as availability is not.
If this way of building resonates, explore how the lab works, read the Startaflo product thesis, or start a conversation.