Claude Community Wrocław #2 [EN]

[Event in English 🇬🇧]
Welcome to the second Claude Community Wrocław meetup - a community-driven initiative built by builders, for builders https://claudecommunity.pl/

We share real stories of how we use Claude, Claude Code, and Claude Cowork in our daily work. No product pitches, no vendor talk - just honest experience from people who build with these tools every day.

This event is NOT affiliated with Anthropic. This is an independent community initiative.
Pizza and drinks are included and the event is free, sponsored by Algoteque International!
If you wish to speak at our future events, use this form: https://forms.gle/8fRNxrCG8JWJUkRQ9
🟢 Where and when?
Klub Wędrówki, Wrocław - August 18, 2026, 6 PM CEST
Agenda
➡️ 18:00 - Welcome & intro by Claude Community Wrocław organizers
🧑‍💻 18:05 - Software Engineering Practices for Agentic Development - Oleksandr Stefanovskyi
🧑‍💻 18:50 - Spec-Driven Development on a Higher Level: Behavioral Tests and Architecture as Guardrails for AI Agents - Maks Piechota
🍻 19:35 - Networking & Pizza
🧑‍💻 20:05 - Agentic Coding vs. Software Architecture: An Early Report - Tomasz Melcer

➡️ Software Engineering Practices for Agentic Development: Bringing engineering maturity with autonomous AI agents
Applying traditional engineering practices to fix current issues with AI agents like Claude Code - why deep CS knowledge still matters, and why hard skills remain essential.
🎤 Oleksandr Stefanovskyi - AI Solutions Architect at Intelliarts with a background in full-cycle software product development, cloud infrastructure, and machine learning leadership. He specializes in designing and refining SDLC processes that integrate AI capabilities into production-ready software architectures.

➡️ Spec-Driven Development on a Higher Level: Behavioral Tests and Architecture as Guardrails for AI Agents
A good spec can still fail you: you hand the agent a precise description, get back code that passes - and still don't know whether it's what you actually wanted. A spec describes intent, but it doesn't constrain the solution space; everything you leave open, the agent fills in for you.
I'll show how I extended Spec-Driven Development with two layers that take those degrees of freedom away: deterministic behavioral tests pin down what should happen as an executable acceptance contract, and C4 diagrams pin down within which boundaries - architecture as a constraint, not documentation. Together they raise the size of the task you can safely delegate. I'll walk through the mechanism on real examples from my own work, plus the tool I built to manage these specs and many parallel Claude sessions at once.
Concrete model to take home: spec = intent, tests = behavioral contract, architecture = structural boundaries.

🎤 Maks Piechota - Senior Engineer on the AI Platform team at FirstConnect Insurance Services. His whole career - from enterprise (Dolby Laboratories) to US startups - has centered on one thing: making software delivery faster without giving up quality, through demanding Engineering Excellence practices (TDD, BDD, Extreme Programming, Trunk Based Development, Continuous Delivery) backed by hands-on software architecture. He argues that the practices the industry spent decades building to minimize human-error risk are exactly what AI-generated code needs - and shares what he learns at meetups and conferences.

➡️ Agentic Coding vs. Software Architecture: An Early Report
The software industry runs on slogans, and agentic coding broke half of them. "Don't reinvent the wheel" was advice for a world where writing code was expensive - with AI we can spin up five wheels, benchmark them, and keep the roundest. "Talk is cheap, show me the code" has aged even worse: the code is what's cheap now, and the talk is the whole job. As an architect in a big-data shop I've been watching which parts of the job actually got cheaper - and which ones just became urgent problems.
🎤 Tomasz Melcer - Long-time software engineer and machine learning consultant turned enterprise architect. Contributed to science, open-source, and local tech communities. Still sometimes diving deep into coding - can't escape that! When in a more casual mode: learning (human) languages and playing board games.
Sponsor
Algoteque International (https://algoteque.com)