Measured, never vibes
Best-of-n, LLM-as-judge, scoring against fixtures, plagiarism checks. Quality is a number on a dashboard before the feature ships, with many models compared across their releases.
Agents in production since 2024
An AI developer who puts agents in production and keeps them there
Tool calling, MCP servers, LLM-as-judge, streaming, prompt-injection defense and cost caps. An agent opened about a third of my team’s pull requests in 2026. Another one is waiting below to answer questions about me.
Available from November 2026 · Wrocław, Poland (CET), remote
- “someone you can always count on”public LinkedIn recommendation
- “a true bias for action”client CTO, on LinkedIn
- 15+ yearsshipping software
- Free call, no pitch deck30 minutes, then you decide
~30%of merged PRs in 2026 opened by an AI agent, ~85% reviewed by me
10,000+affiliate programs catalogd, ~97% found automatically
~40×more commits year over year with agents under enforced rules
4months from first commit to the App Store
Why me
Agents under rules
Enforced stop conditions, tool-call limits, delegation chains, lint and test gates. Coding agents work on main with hooks and every PR gets a human review. The model suggests, the criteria decide.
Prompt injection as a design constraint
An email client where every message is untrusted input taught me to wrap user text, keep tools narrow and let an agent see no more than the person would.
MCP that people actually use
An MCP server from first merge to first external users, routed through existing REST controllers, adoption per client visible in PostHog from day one.
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 produce ten times the output and ten times the mess. 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 the agent opened about 30% of merged pull requests across two products and I reviewed about 85% of them. My own output went from under 1,000 to 5,000+ commits a year.
See also: Full-stack & custom software · AI in your company
Agentic content platform built from scratch, 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.
See also: AI in your company
An MCP server from first merge to first external users
- The problem
- Users 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 within weeks. A bot-protection rule that rejected MCP clients by user agent was found and removed the same day because the dashboard showed it.
See also: PostHog & product analytics · AI in your company
One foot in the door
Not ready for a project? Start smaller.
Two ways to see how I work before you commit to a sprint. Fixed scope, a number in writing after a short call.
AI & agents codebase audit, one week
I read the repo, CI, tests, security and the parts the team avoids. You get a written report with findings ranked by risk, a plan with milestones and a go or no-go you can defend. I have run this for a large software house before its clients saw the code.
Code review of one pull request or one module
Send me the diff that worries you. Within a few days you get line-level comments, the three things I would change first and whether I would merge it.
Ask about an audit Free 30-minute call first. If an audit is the wrong tool, I say so on the call.
Live demo
Ask my agent about me
This chat runs on Workers AI with three tools: check_availability, match_stack and list_case_studies. Every tool call shows up in the trace before the answer streams in. It only knows what is on this site, it will say “I don’t know” otherwise, and nothing you type is stored.
Demo loading…
Try “Does he fit React, Cloudflare and Postgres?” or paste your job ad and ask for the gaps.
How it works
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. If I am not the right person, I say so and point you to someone who is.
- 2
Scope in writing
A short written plan follows: what we build first, what we measure, how we work together and what it costs. No slide deck.
- 3
Something to touch in week one
The first slice ships in the first week: deployed, measured, reviewable. From there the work is paced by what the numbers say.
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 a team of up to four
- TypeScript, React, Node.js, Cloudflare, Postgres, PostHog
Probably not when
- A program that needs a ten-person vendor team
- A stack built around Java, Python or a cloud I do not sell
- Body leasing by the hour with no say in how the work is done
- A start before November 2026
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.
Measured, not assumed
PostHog in every project from day one. A tracking plan per feature and dashboards leadership actually reads.
Zero bug policy
A bug is fixed or consciously closed, never parked. Alerts and error tracking in every project, so I learn about problems before the client does.
FAQ
Questions people ask before they book
How do you defend against prompt injection?
User text goes into the prompt as data, wrapped and escaped, with a standing instruction that it is never an instruction. Tools are narrow and read-only unless the product needs more. Outputs are filtered before they reach the screen. The agent on this page works that way.
Which models and providers?
Claude, OpenAI and open-weight models on Workers AI, picked per task and swapped when a release changes the picture. I keep evaluations so a model change is a measured decision.
What is an agent or MCP readiness audit?
One week. I look at your API, auth, tool surface, evaluation data and analytics, then hand you a report with what an agent could safely do today, what needs work and a plan. It is the cheapest way to find out whether to build. The audit is the first rung on the offer ladder.
Do you use coding agents in client work?
Yes, under enforced rules and with disclosure. Roughly 30% of merged PRs across two products in 2026 came from an agent and about 85% of those were reviewed by me. You get the output, I keep the responsibility. The same rules apply when I run a project as fractional tech lead.
Can you make our product usable by other people’s agents?
That is what an MCP server is for. I have shipped one to external users, removed destructive tools from it and measured adoption per client. Same approach for your product.
How soon can you start?
November 2026. Book the call and we start with the readiness audit.
Let’s find out in 30 minutes
A free 30-minute call. You bring the problem, I bring questions. No prices on this page because scope comes first.
If I am the wrong person I will say so on the call and point you to someone better.
After you book: a calendar invite, then the call. Nothing else to fill in.
Available from November 2026 · B2B contract or employment · Wrocław, Poland (CET), remote