Daniil takes the first call personally and gives technical answers from the executive vision.

We are distributed and remote-first. If you have a Computer Science background and care about building things people actually read and use, talk to us.

Engineers in every time zone

Freedom for people and for clients: no lock-in, open decisions, the right to walk away.

Freedom

Computer Science first. We understand the systems before we change them.

Expertise

A tool, not a goal. AI where it measurably helps, and nowhere else.

Technology

Every step tracked with data. Every step produces a document you can read.

Transparency

Your data stays yours. Security is designed in, not bolted on.

Security & privacy

HMR: Human Must Read. We make things people want to read, not noise.

Respect for people

Direct answers

Where AI should work. And where it shouldn't

its.xyz, a personal intro, a partner website, any channel

The protocol

What you get, in writing

Notes from the team — engineering, design, and product.

Notes from production

Every post here is written or edited by a person who did the work. No generated filler.

Human Must Read

Blog

AI built into how your software is specified, coded, reviewed, tested and released.

AI SDLC Pipelines

The right answer from your own data, and your business rules applied the same way every time.

Search & Rule Engines

The +1 step: one real process, automated end to end.

Process Automation

Legacy systems moved to a modern stack, without stopping the business.

Legacy to Modern Solution

Continuous, automated security checks of your code, dependencies and infrastructure, with findings a human can act on.

Security Automated Audit

Robots that walk through your product like real users, in production, around the clock, and alert you before your customers notice.

True Smoke Tests & Production Liveness Robots

Turnkey solutions. Shipped, not slideware

You own all of it

Duplicated, forgotten or idle resources that still cost money.

Redundant services

Instances, databases and clusters sized for a peak that never comes.

Over-provisioning

Managed services where simpler, cheaper ones would do the job.

Architecture shortcuts

A readable map of who and what spends your cloud budget.

Cost visibility

Thousands of bill lines, logs and config files. AI reads all of it; our engineers decide what is safe to change.

AI reads bills, metrics and infrastructure code

Engineers map what each service really does

We implement the changes you approve, safely

Every scope change comes back to you with a re-estimate. Nothing is decided for you, nothing silently.

Keep you deciding

Clear rhythm, clear owners, people and AI agents in one flow.

Organise

Complex scope broken into steps a team can actually deliver.

Structure

Faster delivery is the start. The goal is a more detailed, deeper product.

Deepen the product

AI takes the status updates and reminders. The lead takes the rhythm and the structure. You keep every call that matters, with the numbers to make it.

Scope changes re-estimated on the fly, never silently

Retros that change how the next sprint works

Authentication, authorisation, rate limits and input validation on every endpoint.

API hygiene

Access rights, sessions and how people actually use your tools day to day.

User interaction hygiene

Leaked keys and vulnerable libraries found before they are exploited.

Secrets & dependencies

Continuous automated checks, with findings a human can act on.

An unchecked endpoint, a leaked token, a permission nobody revoked. Hygiene closes them before someone else finds them.

We review your APIs, access model and code

The critical gaps closed first, with your team

Ongoing monitoring as part of <support>support</support>

How work and data really move through your systems.

Process & data mapping

How systems should talk to each other, and where they shouldn't.

Integration architecture

What else breaks, moves or speeds up when one part changes.

Change impact analysis

Specifications developers, managers and AI agents can all work from.

Readable specs

Every change is modelled, its impact traced across processes and data, and written down so that people can read and agree on it.

Processes, data flows and integrations as they are

Specs ready for your team or <build>ours to build</build>

AI drafts test cases from specs and code. Engineers review every one.

AI-assisted test design

Robots walk critical user journeys and contracts on every build.

UI & API test robots

Not "the page loads", but "a user can actually complete the task".

True smoke tests

24/7 checks in production that alert you before your customers notice.

Production liveness robots

Your QA engineers then spend their time on what robots can't do: thinking about what could go wrong.

We map your critical user flows and current coverage

Liveness robots keep checking production, day and night

{title} — its.xyz, engineering agency since 2015.

Who decides, who builds, who checks. Clear roles for AI work.

AI operating model

Your engineers and operators work next to ours on real tasks.

Hands-on enablement

Rules for what AI may produce and what a human must read.

HMR standards

We help your people prepare for Anthropic and Google Gemini certifications.

Certification path

AI drafts, people decide. Clear checkpoints where a human reviews, approves or corrects.

Human-in-the-loop flows

We build an AI-powered software delivery lifecycle for your organisation: from specs and code to review, testing and release.

AI SDLC implementation

Your team, upgraded

Walk the transformation path yourselves

Free one-week audit: where AI fits and where it doesn't

We build the first win together with your team

You run the next steps on your own. We're a call away

Half of a good transformation is knowing what to leave alone. Your team learns to tell a real +1 step in automation from a stream of generated noise.

Real enterprise AI integrations, taken all the way to production, taught us where transformations stall: unclear ownership, unreadable output, and AI pushed into places it doesn't belong. We design around all three from day one.

We've walked this path before

The page you asked for does not exist or has moved.

Page not found

Have a process that eats hours?

What needs to be done, stack, project stage, timeline…

Please describe your task in at least 2–3 sentences

Could not send

Name

Request received

Describe your process

1h call with your executives: goals, constraints

1h call with your operations team: how work really flows

Report: where AI fits, where it doesn't, and what "done" will look like

want us to build it? → we shape the ask, resource plan and work breakdown, and kick off prefer to go on your own? → the report is yours to keep

Free AI audit. One week.

Data from many sources, normalized, with metrics and reports kept accurate in real time.

BI system for 20M+ users

A trading bot for the oil industry that wins deals with different tactics across exchange instruments.

HPT Quantum Bot

Routing on a microservice architecture: fewer delays, lower courier costs.

Smart delivery routing

CRM and operations for wake parks: bookings, payments, day-to-day.

Plotva

Shipped and running

Signed before the audit, not after. Your template or ours.

NDA before the first call

We work in your cloud, your VPN, your access rules. Models can run on-prem where the data must not leave.

Your infrastructure if you prefer

Nothing from your process goes into a model, a demo or a case study without written consent.

No training on your data

Your process stays yours

Pilot to production in 1–2 months

Map the process and the data. Agree the acceptance criteria.

Pilot with real users, measured against the criteria

We know what done looks like before we start

HMR: Human Must Read

Most requests are precise and reasonable. Some of them automate the wrong step. A week of audit tells which one you have, and what the working version looks like, before anyone writes code.

We understand the systems AI runs inside.

Engineers first

Specialists available across time zones, so work keeps moving.

Every time zone

We certify with Anthropic and Google Gemini, not just read about them.

Certified

Computer Science people

Find where AI should work and bring value.

Build the automated solution, from spec to production in 1–2 months.

Support, maintain and update when needed. We stay with you for as long as you need.

Audit. Change. Build. Support.

Account layer, bench rates, a team you meet after the contract. Months to the first working slice.

One tool, one demo, no production history. Your process has to fit the product.

The people on the first call write the code. Production systems since 2015. The product fits your process.

The only question that matters now is how we will know the result is the one you need. We turn it into acceptance criteria, then break the work down to the level a model or a script can execute reliably. That decomposition is the skill; the execution is hours of an agent.

"Done" is written down before the first commit.

Senior engineers set the architecture and the criteria. A trained team and AI tooling execute inside a process we have run since 2015. You get senior judgement on every decision at a cost a large vendor cannot match, because there is no bench and no account layer between you and the people writing the code.

A senior core, a lean team, a tested delivery process.

Not every process needs a model. We tell you where AI pays back, and where a script, a form or a person does the job better.

We'll show you where not to use AI.

What we actually do differently

One process. One outcome.

A senior lead, engineers and our AI tooling, working as one unit on one outcome.

A squad, not a seat

Acceptance criteria written down with the process owner before the first commit.

"Done" agreed in week one

We check the ask against the process, not just implement it.

Product thinking

From discovery and architecture to release and monitoring.

Full lifecycle

We adapt to your tools, access policies and compliance.

Your stack, your rules

NDA before the first call, your infrastructure if you prefer, nothing from your process in a model or a case study without written consent.

Confidential by default

No handoffs through three layers of documents. We talk to the people who do the work, write down what "done" means with them, and build to that. They keep running the process; we build the tool around it.

Week one: NDA, access, and what "done" means, written down with the process owner

We map the process and the data with the people who run it

Working slices every week, measured against the criteria, production in 1–2 months

Your team takes it over, or we keep <support>supporting it</support>

Monitoring, incident response, fixes. Specialists in every time zone.

Keep it running

Quality checks on AI output, data drift and model updates.

Keep it accurate

The next +1 step: new features and automations, on the same foundation.

Keep it improving

Documentation stays current. HMR applies after launch too.

Keep it readable

With you for as long as you need

We agree on response times and scope that fit you