AI in your company, with receipts

An AI consultant who has shipped agents, not slide decks

Q: What does it cost?

The one-week code audit: $1,080–$2,160 excl. VAT. A fractional AI lead two days a week for three months: $6,732–$12,366 excl. VAT. The workshop and a single agent build get a written price within 24 hours after the call, and the configurator shows hours and a range before you book.

Q: Did the EU postpone the AI Act?

Partly. High-risk rules moved to 2 December 2027 and 2 August 2028, but chatbot and AI-content transparency still applies from 2 August 2026 (Article 113, Article 50). Fines for transparency breaches reach 15 million EUR or 3 percent of turnover, the lower figure for SMEs (Article 99).

I take your company from “we tried ChatGPT” to one workflow that runs on its own and gets measured. I have built agents since 2024 and run them in production: narrow tools, evals, cost caps, MCP servers with external users. Start with the one‑week code audit, or score yourself below in two minutes.

Book a free 30‑min call Email me instead

Calls this week · work starts 3 November 2026 · Wrocław, Poland (CET), remote

“measurable changes in some key metrics”former manager, now a CTO, on LinkedIn
“someone you can always count on”an engineer from a team I led, on LinkedIn
15+ yearsshipping software
Free call, no pitch deck30 minutes, then you decide
Reply within 24 hourspersonally, or the first hour is free

~2/3of merged PRs in 2026 built with AI coding agents, nearly all through my review or merge

~3×more GitHub contributions in 2026 than in 2025, 8,500+ by October, with agents under enforced rules

10,000+affiliate programs cataloged, ~97% found automatically

15+years shipping software

Sound familiar?

AI adoption problems I solve for companies in 2026

Why do our AI pilots never make it to production?

In Deloitte’s State of AI in the Enterprise survey of 3,235 business and IT leaders (21 January 2026), only 25% had moved 40% or more of their AI pilots into production.

I pick one workflow with an owner and a number attached, ship it to production and measure it before touching the next. It starts with the one‑week code audit, with your processes in scope, not a slide deck. In May 2025 I measured a client’s AI article pipeline in production instead of guessing: over one weekend 65 of 69 runs succeeded, about 94 percent, at about a dollar per article.

Employees paste company data into ChatGPT. What now?

Verizon’s 2026 Data Breach Investigations Report (May 2026) found that 45% of employees now regularly use AI on their corporate devices, up from 15% a year earlier, and 67% of users reach AI services from those devices through non-corporate accounts.

A ban pushes it underground. I give the team an approved assistant connected to your systems through each person’s existing company login, so it sees only what that person may see, with usage measured per person. On a computer-vision study I read a vendor’s terms, saw that uploads granted them a license to the images, and told the team not to upload the client’s photos there.

Sound like you? Book a free 30‑min call →

How much should AI agents be allowed to do on their own?

Gartner predicts that by 2027, 40% of enterprises will demote or decommission autonomous AI agents because of governance gaps found only after production incidents (26 May 2026).

Agents start read‑only with narrow tools, and anything that cannot be undone is flagged so the agent asks a person first. Autonomy grows tool by tool, when the logs show it is safe. In April 2026 I proposed a rule for the AI features in an email product: the model only suggests, and thresholds and criteria set by people decide what actually happens.

Our developers keep fixing almost-right AI code

In the 2026 Stack Overflow Developer Survey (6 October 2026), 73% of respondents used coding assistants and agents daily, yet only 48% said they trust AI when they can easily validate its answers.

I write CLAUDE.md and AGENTS.md rules enforced by hooks, lint and test gates in CI, with a human review before merge. In 2026 about two thirds of merged PRs across two products were built with coding agents, and nearly all went through my review or merge.

Facing one of these? Book a free 30‑min call →

Adoption that survives the pilot

A pilot with no owner and no metric is a demo. Before the build we name both for one workflow, PostHog measures it from day one, and after the first month you get a written go or no‑go.

Agents you can debug, without a framework

No LangChain. Tools with narrow scopes, an evaluation loop with LLM‑as‑judge against fixtures, a cost cap in a Durable Object. My content platform published thousands of articles and a catalog of 10,000+ programs, ~97% automatically, on exactly that design.

Rules, skills and review for coding agents

CLAUDE.md and AGENTS.md that agents cannot skip, skills for the repeating work, hooks and lint gates in CI, a human review before merge. With those rules one product went from about 60 commits in 2025 to 4,800+ by October 2026.

Your product usable by agents

If ChatGPT and Claude users should reach your product, you need an MCP server: narrow tools on top of your existing auth, so an agent gets no wider access than a person. Adoption measured per client from the first merge.

Live demo

Score your AI readiness in two minutes

Twelve questions, six areas. You get a score per area, a verdict and a sketch of your first 30 days, built from your two weakest areas. Everything runs in your browser.

Demo loading…

Copy the result as Markdown and bring it to the call. It becomes the agenda.

Ways to work together

Four shapes of engagement

Want a number first? The configurator shows hours, a price range and the start date. The call is free either way.

1 week

One-week code audit, with AI readiness

Your code, processes, data, API and team, read by hand by someone who has shipped agents. A written report: which workflow goes first, what an agent can safely do today, what has to change, a plan with milestones. $1,080⁠–⁠$2,160 excl. VAT.

Book a free call →

4 to 6 weeks

One workflow, one agent

A single outcome in production: a support triage agent, a content pipeline, an MCP server for your product, a coding-agent setup for the team. Measured from the first day, handed over with the rules and the dashboard.

Book a free call →

workshop + 4 weeks

AI enablement for your engineers

A hands‑on workshop, then rules, skills and review gates for Claude Code and Codex in your repos, then four weeks of pairing. The team keeps the speed and loses the mess.

Book a free call →

2 to 3 days a week

Fractional AI lead

Inside your team for a quarter or two: the roadmap for AI features, vendor and model choices, evals, costs, security reviews, hiring screens. Your people ship, I keep the quality bar. Two days a week for three months: $6,732⁠–⁠$12,366 excl. VAT.

Book a free call →

Team workshop

Coding agents under rules, for your team

For a team where everyone already builds their own tools with Claude Code or Codex, and nobody agreed how. One day on site or remote, on your own repositories, not on a toy project.

Where you are
What each person builds with agents today, what broke last month, which repositories matter.
Shared rules
CLAUDE.md and AGENTS.md for your repos: what the agent may do alone, what needs a human, how to write it down.
Skills for repeating work
Turn the prompts people keep retyping into skills the whole team uses.
Gates in CI
Lint, tests and hooks that stop a bad change before review, not after the deploy.
Review that scales
How to review agent pull requests quickly, and when to throw one away.
Accounts and costs
Which tools and plans, who pays for what, and how to see what the agents cost.

You get the rules and skills committed to your repos, a list of tools and accounts, the recording, and 30 days of questions on chat. A second day is optional.

See workshop price and date →

Proof

Work I can show

Problem, what I did, result. Client names only where I am allowed to use them.

An AI coding agent as a team member, under enforced rules

The problem: Coding agents write code faster than a team can review it. A team of four wanted the speed without a slop swamp on main.
What I did: I authored the rule set shared by Claude Code and Codex: mechanical checks over written rules, lint and test gates for security, API surface and CSP, agents working on main with hooks, every PR reviewed by a human, AI‑assisted content disclosed as such.
Result: In 2026 about two thirds of merged pull requests across two products were built with coding agents, and nearly all went through my review or merge. I merged 260+ PRs of my own and reviewed 300+ of other engineers’. My commits across those two products went from under 1,000 in 2025 to 7,500+ by October 2026.

~2/3 PRs with agents 300+ reviews 7,500+ commits in 2026

Offer: AI engineer for your team (Node.js, LLM agents)

An MCP server from first merge to first external users

The problem: Users of Sandtime.io, the time tracker I built at Sanddev, used by 4,000+ freelancers and teams, wanted to log time from their AI assistants. The API existed, the agent surface did not.
What I did: I set the tool surface, removed destructive tools, routed every operation through the existing REST controllers so an agent gets no wider access than a person, and instrumented adoption per client in PostHog from day one.
Result: First external users. A bot-protection rule that rejected MCP clients by user agent was found and removed the same day because the dashboard showed it.

MCP OAuth adoption dashboard

Offer: Your company systems inside Claude and ChatGPT (MCP)

AI content platforms: R&D and an agentic content platform, quality measured not assumed

The problem: Thousands of articles were needed, with AI disclosure built in and without publishing slop.
What I did: Late 2024 I built the agent driver: enforced stop conditions, tool‑call limits, parallelization, delegation chains. Quality came from best‑of‑n, LLM‑as‑judge, scoring against fixtures and plagiarism checks, with many models evaluated across their releases.
Result: Thousands of articles shipped through the platform. Content deployed automatically across 300+ generated sites with git‑backed bulk operations and reputation monitoring. From November 2023 to May 2025 it was a two‑person team. Invented author bylines came off all 300+ sites within a week, and one measured weekend in production saw about 94% of runs succeed at roughly a dollar per article.

thousands of articles 300+ sites LLM‑as‑judge

Three steps from this page to a running project.

1
A 30-minute call
You describe the product and the problem. I ask the questions I would ask on day one.
2
Scope in writing
Within 24 hours after the call you get a short written plan: what we build first, what we measure, how we work together and what it costs. No slide deck.
3
Working software in week one
Deployed, measured and open for your review. After that, the numbers set the pace.

Book a free call Email me instead

Fit check

Is this a fit?

A good fit when

You need a senior who owns delivery, not a pair of hands
The product has to ship and be measured, with or without a product manager
A solo engagement, or one team of up to four that I lead directly
TypeScript, React, Node.js, Cloudflare, Postgres, PostHog, and Python when the work needs it

Probably not when

A program that needs a ten‑person vendor team
A stack built around Java or a cloud I do not work in
Body leasing by the hour with no say in how the work is done

Working style

How I work

Agents under rules

Claude Code and Codex work on main with hooks, lint and test gates. The model suggests, the criteria decide. Every PR gets a human review and AI‑assisted content is disclosed.

Review as a craft

800+ pull requests reviewed on one product since 2018. I review for the product and its reasons. The code comes second.

Measured, not assumed

PostHog in every project from day one. A tracking plan per feature, and dashboards leadership reads instead of asking me for numbers.

Good to know

AI Act and coding-agent facts companies ask about

AI literacy became softer

After the 2026 Digital Omnibus, Article 4 asks companies to take measures to support staff AI literacy and does not require any guaranteed level (final text).

High-risk rules moved to December 2027

Obligations for Annex III high‑risk systems, such as HR or credit scoring, now apply from 2 December 2027, and AI inside regulated products from 2 August 2028 (Article 113).

Chatbot disclosure still applies from August 2026

Article 50 was not postponed: people must be told they are talking to AI at the latest at the first interaction (Article 50).

Reviewed AI text needs no label

AI‑generated text published to inform the public needs a label unless it went through human review and someone holds editorial responsibility (Article 50).

Claude Code skips AGENTS.md when CLAUDE.md exists

By default Claude Code reads AGENTS.md only when there is no CLAUDE.md, so teams mixing Codex and Claude Code should import one file from the other (Claude Code docs).

Codex stops reading at 32 KiB

Codex stops adding AGENTS.md files once their combined size reaches 32 KiB by default; rules past that point are not loaded (Codex docs).

FAQ

Questions people ask before they book

I run a small company and build our system myself with Claude Code. Is a workshop too much?

Probably. A few hours a month with a senior next to you fits better: see a senior for owners who code with AI.

Why no LangChain?

Because the problems that matter sit elsewhere: which tools the agent gets, how you know it is correct, what it costs, what happens when it is wrong. A few hundred lines of TypeScript with an eval suite beat a framework you cannot debug. The technical side is on the AI and agents page.

Which models and vendors?

Whatever wins the evaluation for your task at your budget. I run Claude, OpenAI and open models on Workers AI side by side and keep the comparison in the repo. Vendor lock‑in is a decision, never an accident.

What about our data and GDPR?

EU hosting where the data requires it, no customer data in prompts without a policy that says so, prompt injection treated as a design constraint. Analytics through PostHog EU with a first-party proxy, see the PostHog page.

Will you replace our developers with agents?

No. Agents under rules make a good team faster and a careless team louder. I set the rules, the gates and the review, your engineers keep the keys. Read how I run them on the AI and agents page.

How do you measure ROI?

A metric per workflow before the build: tickets closed, hours saved, articles published, adoption per client. PostHog from day one, a dashboard leadership reads without asking, and a written go or no‑go after the first month.

What does it cost?

The one‑week code audit: $1,080⁠–⁠$2,160 excl. VAT. A fractional AI lead two days a week for three months: $6,732⁠–⁠$12,366 excl. VAT. The workshop and a single agent build get a written price within 24 hours after the call, and the configurator shows hours and a range before you book.

How soon can you start?

From 3 November 2026. Book the call and bring the readiness score from the demo above.

Did the EU postpone the AI Act?

Partly. High‑risk rules moved to 2 December 2027 and 2 August 2028, but chatbot and AI‑content transparency still applies from 2 August 2026 (Article 113, Article 50). Fines for transparency breaches reach 15 million EUR or 3 percent of turnover, the lower figure for SMEs (Article 99).

Why does Claude Code ignore our rules?

CLAUDE.md is context, not enforcement, and contradicting lines make Claude pick one arbitrarily. Hard limits belong in a PreToolUse hook (Claude Code docs). Skill descriptions are cut at 1,536 characters, so put the trigger first (skills).

Why do our AI pilots never make it to production?

In Deloitte’s State of AI in the Enterprise survey of 3,235 business and IT leaders (21 January 2026), only 25% had moved 40% or more of their AI pilots into production. I pick one workflow with an owner and a number attached, ship it to production and measure it before touching the next. It starts with the one‑week code audit, with your processes in scope, not a slide deck. In May 2025 I measured a client’s AI article pipeline in production instead of guessing: over one weekend 65 of 69 runs succeeded, about 94 percent, at about a dollar per article.

Bonus · #suchar

A dad joke, on the house

Free, like the first call. Same sense of humor, included in every project.

Why is the AI model on a diet?Punchline

It’s counting tokens. Ba‑dum‑tss.

How many prompt engineers does it take to change a light bulb?Punchline

One. Five attempts and a “please”. Ba‑dum‑tss.

What does an AI say in a job interview?Punchline

“Great question! You’re absolutely right.” Ba‑dum‑tss.

In 30 minutes you will know if we are a fit

You bring the problem, I bring the questions. Want a number first? The configurator shows hours, a price range and the start date.

Book a free 30‑min call Email me instead

or write to ciunkos@gmail.com

After you book you get a calendar invite. Nothing else to fill in.

My pledge

Reply within 24 hours. I reply within 24 hours on working days, personally, not a bot. Weekend messages: by Monday noon.
Late means free. If I am late, the first hour of work is free. You do not have to ask.
Scope in writing within 24 hours. A free 30‑minute call with no company deck, then the scope and the price in writing within 24 hours. You decide.
The person who does the work. You talk to the person who does the work. No sales layer, no handoff.
Follow‑up only with news. I follow up only when I have something new for you. Never a “did you see my offer”.
Honest “not a fit”. If I am not the right person, I say so on the call and point you to someone who is.

Calls this week · work starts 3 November 2026 · B2B contract · Wrocław, Poland (CET), remote

You own the code, domain and accounts from day one. Nothing is held hostage.

An AI consultant who has shipped agents, not slide decks

AI adoption problems I solve for companies in 2026

Why do our AI pilots never make it to production?

Employees paste company data into ChatGPT. What now?

How much should AI agents be allowed to do on their own?

Our developers keep fixing almost-right AI code

Why me

Adoption that survives the pilot

Agents you can debug, without a framework

Rules, skills and review for coding agents

Your product usable by agents

Score your AI readiness in two minutes

Four shapes of engagement

One-week code audit, with AI readiness

One workflow, one agent

AI enablement for your engineers

Fractional AI lead

Coding agents under rules, for your team

Where you are

Shared rules

Skills for repeating work

Gates in CI

Review that scales

Accounts and costs

Work I can show

An AI coding agent as a team member, under enforced rules

An MCP server from first merge to first external users

AI content platforms: R&D and an agentic content platform, quality measured not assumed

Three steps from this page to a running project.

A 30-minute call

Scope in writing

Working software in week one

Is this a fit?

A good fit when

Probably not when

How I work

Agents under rules

Review as a craft

Measured, not assumed

AI Act and coding-agent facts companies ask about

AI literacy became softer

High-risk rules moved to December 2027

Chatbot disclosure still applies from August 2026

Reviewed AI text needs no label

Claude Code skips AGENTS.md when CLAUDE.md exists

Codex stops reading at 32 KiB

Questions people ask before they book

A dad joke, on the house

In 30 minutes you will know if we are a fit

Related services