FAQThe questions people ask
The questions people ask
before the first call.
24 answers with no “it depends on the task” standing in for an answer. If your question isn't here, write to us and we'll add it.
Work and process
Why are you faster than a traditional studio?
Three reasons, and none of them is “we work nights”. Two or three senior engineers sit on the project with no layer of managers in between, so a decision is made the same day. The standard blocks — authentication, payments, roles, an admin panel, CRM integrations — sit in our library and are never written twice. And a demo every 2–3 days: any gap between the build and your expectations turns up in week two, while reworking is still cheap.
How long does a typical project take?
A landing page or a simple bot: 5–10 working days. A web app or a Mini App with payments: 3–6 weeks. An AI agent: 3–8 weeks, half of it spent on the knowledge base. The first demo lands on day 3–5 either way.
What if the result isn't what we wanted?
It rarely gets that far: you see a demo every 2–3 days, and any gap turns up in week two while it is cheap to fix. If the prototype makes it clear that we are a poor fit for each other, you pay for the work actually done and take everything that has been built.
Who on your side needs to be involved?
One person with the authority to decide, and 2–3 hours of their time a week: look at the demo, send corrections, hand over access. What derails projects most often is waiting for sign-off. A committee of five turns a two-day change into a two-week one.
Do you work under a contract?
Yes. A contract with an annexe covering scope, timelines, the acceptance procedure and the rights to the result. We work with sole traders and companies and issue the closing paperwork. We sign an NDA before the brief if the task is sensitive.
Can the requirements change along the way?
They can, and almost always should: the first demo shows up things nobody raised at the brief. A change of scope is recorded in writing and the timeline is recalculated the same day. You learn what a change costs before we start it.
Cost and payment
How do you calculate the cost?
By the hour, broken down by task. After the brief you get an estimate showing how many hours go into the front-end, how many into the CRM integration, how many into tests. Before the brief we name no figure: naming one means guessing the scope, and guessing inside someone else's budget is a bad habit.
What is the payment schedule?
Usually three instalments: 30% at the start, 40% at prototype acceptance, 30% after launch. On projects longer than two months we switch to monthly payment for work done. The estimate and the brief are free; working one out is never charged for in advance.
Why are you cheaper than a traditional studio?
We pay for no large office and no layer of account managers relaying the developer's words back to you. The standard parts — authentication, payments, roles, an admin panel, CRM integrations — sit in a library and are not rewritten on every project. The estimate covers only the part of the work that is specific to your task.
Can the estimate grow during the project?
Only if the scope grows, and only with your written agreement. There is no “refining the estimate” after the fact: if a change goes beyond the agreed scope, we show how many hours it adds and you decide whether to do it now, postpone it or drop it. Mistakes in our own estimate we finish at our own cost.
What happens to the price if the project has to stop?
You pay for the work actually done and take everything that has been built: code, designs, documentation, access. No penalties for termination and no holding the result back as leverage; that condition is written into the contract.
What affects the price most?
The number of integrations: each external system adds 15–40 hours for its limits and its failures. After that come load requirements, regulatory requirements and urgency. Urgency is the only factor we call a surcharge outright: 20–40% for parallel work and overtime.
Technology and AI
How is AI red teaming different from an ordinary penetration test?
A penetration test looks for mistakes in code: injections, authorisation bypasses, leaky permissions. AI red teaming looks for mistakes in the model's behaviour, where the vulnerability looks like text that talks the agent into doing something forbidden. A patch will not close it; only architecture will, through tool permissions, data isolation and checks on the answers. These are different skill sets, and we do both.
Will our data end up in foreign models?
Only if we agree on that. By default we use API mode with no training on your data. If the data is sensitive, we deploy an open model inside your own environment: lower quality, but nothing leaves the perimeter. Anonymising data before it is sent covers most cases, because models almost never need real names.
Why Go and Next.js when we already run another stack?
If you have a working team on another stack, we write in yours: leaving you a system nobody can maintain costs more than any saving on development. Go and Next.js are our default choice for new projects, for predictable performance, one language on the client and the server, and simple deployment.
Will the agent make up answers?
It answers only from your documents and has to cite a source; without a source the answer never reaches the customer. Before launch we run a set of 100–300 labelled questions and report the share of correct answers as a percentage.
Security and rights
Who owns the code after the project is handed over?
You do, entirely. The sources go into your repository together with exclusive rights, and this is written into the contract. You will never buy a licence for your own product, and there are no closed modules that only run on our side.
Do you work under a contract and do you sign an NDA?
Yes to both. We sign an NDA before the brief if the task is sensitive: nobody is obliged to discuss details on a promise. We ask for access to your systems at the minimum level and one system at a time, instead of “give us admin rights to everything”.
How do you store our credentials?
In a password manager separated by project, available only to the engineers on that project. After handover we ask you to revoke the keys issued to us, and we remind you ourselves if you forget.
What about 152-FZ (Russia's personal-data law)?
On any project with a form that takes a phone number or an email, we cover the basics: separate consents for processing and for marketing, a policy written around your processes, and a cookie banner that genuinely blocks the counters until it is answered. A full audit with a data-flow map is a separate service.
Support
What is included in support after launch?
A month of free fixes is always included. After that it is optional: a subscription starts at ₽30,000 a month — monitoring with alerts and a response within the working day. The tier with 20 development hours, dependency updates and a two-hour incident response starts at ₽80,000. Unused hours roll over, and you can cancel at any point.
What if we don't want a subscription?
That is normal and often the right call. You get the code, the documentation and the deployment instructions, and from there your own team or another contractor runs the project. We stay available for one-off tasks at an hourly rate.
What happens if something breaks at night?
On a subscription, monitoring sends the alert to us and to you at the same time, with a response within 2 hours, or within an hour under Enterprise SLA terms. Without a subscription the alerts go to you: we set them up at launch so the system tells you about a problem before a customer does.
Will you train our team?
Yes. The “Business” package includes two hours going through the code and architecture, recorded; Enterprise includes a full course on the system. A decent contractor works so that you can manage without them; people come back because it is easier that way.
Didn't find an answer?
Ask us directly.
We reply within two hours during working hours. Technical questions are answered by the engineer who will do the work.