# Session title

WebExpo 2026 Wrapped: Lucerna Cinema

## Session URL

https://webexpo.net/blog/webexpo-2026-wrapped-lucerna-cinema/

## Content

Part 2 of 4



From the Great Hall's case for human judgement, we move into Lucerna Cinema. This was the room of working systems: containers, MCP servers, product maps, typography, authorisation, research engines, Git, contract testing, funnels, refactoring, search visibility and AI adoption.



The common thread was clarity. Not tidiness for its own sake. Clarity as a standard operating procedure: local environments that resemble production, product teams that know what they are optimising for, text people can read (yes, still), contracts that machines can enforce, and AI initiatives grounded in purpose rather than panic.















Making systems less surprising



The opening cinema session was the kind of technical talk that made the theme immediately concrete. In "Going from containers, to pods, to Kubernetes", Cedric Clyburn started from the familiar pain of local setups that do not match production. Containers, Podman, rootless execution and Kubernetes manifests were not presented as cool infrastructure, but Cerdic’s demos showed how developers can move from local containers towards Kubernetes without treating production as a separate universe.



That same bridge-building continued in Ladislav Prskavec's "Under the hood of AI: Building your own MCP server in Go". If Cedric connected local development to production, Ladislav connected AI models to organisational tools and data. He described AI as "powerful but blind", then showed how MCP can act as a standard connector. The caution stuck: input validation, authentication and prompt injection risk have to be part of the server design.



Product teams got their version of this discipline in Christiane Moser's "Product experience mapping". Her promise was turning chaos into structure. The talk was not about prettier canvases. It was about forcing teams to name users, goals, needs, experience principles and metrics before they react to every stakeholder signal. Unclear systems produce unclear decisions.



At the level of interface detail, Oliver Schöndorfer made typography impossible to dismiss as polish. "How bad typography kills UX" moved through readability, hierarchy, contrast and brand expression, showing how poor type choices quietly damage comprehension. "Accessibility is not a bonus" broadened the argument. If users cannot read it, scan it or understand it, the product is already failing. His feedback was ALL CAPS in the best possible way. People loved it. 🙌







AI with boundaries



AI came back at midday, with strong human and security guardrails. Daria Rudnik's "Partnering with AI" opened with a blunt statistic: "95% of organisations are seeing zero return on investment from AI initiatives”. Her operating rule, "a human take always comes first", resisted the idea that automation alone creates value. Teams need to decide which relationships, decisions and quality standards remain human-owned.



Sohan Maheshwar made the boundary problem technical in "How to prevent AI Agents from accessing unauthorised data". As agents move through applications, old access-control weaknesses get dangerous. His case for relationship-based access control, including ideas from Google Zanzibar and SpiceDB, made one thing clear: AI agents need deterministic authorisation, not vague trust in context. Huge thanks to Sohan for making it to Prague on crutches after recent surgery. Nothing could stop him.



Growth teams got a similarly practical warning from Vladana Bačová. "Fixing a broken SaaS funnel" challenged the assumption that more acquisition solves everything. Her funnel examples showed leaks after sign-up, in qualification, conversion and retention. "What happens after the sign-up?" became the useful question because it forced attention onto the parts of the customer journey that are easiest to ignore.



Then Lucky Nkosi went bananas. 🍌 "Flying a drone with gestures, bananas & Web APIs" used voice, gesture control and banana inputs to make the room laugh, but the underlying point was serious enough: web technologies can open up new interaction modes when developers are willing to experiment with the physical world.















Decisions, agreements and team memory



Day two started with decision-making rather than tooling. In "Building the engine that drives better product decisions", Julian Della Mattia argued that teams often lack the right evidence at the moment a decision is being made. His research engine was not a repository with a nicer name. It was a culture, infrastructure and communication loop designed to get evidence into the room before opinions harden.



Katherine Corneilson's "Hourglass analysis" offered a method for moving through messy qualitative material. When research gets too top-down, it needs raw detail. When bottom-up analysis becomes endless, it needs structure. Her advice to flip modes when stuck gave teams a practical way out of analysis paralysis.



The technical equivalent came from Robin Pokorný in "Contract testing for teams that want to move fast". Distributed systems fail when teams carry different assumptions about the same interface. "A pact is an agreement between humans that a machine can enforce," Robin said, making contract testing sound less like test overhead and more like a shared language between teams.



Sergès Goma turned Git into shared memory. "How to Git away with murder" sounds silly (we think this was a joke), but the message was serious: commits, branches, rebases and reflogs are how teams preserve evidence, recover from mistakes and explain change. Clean history is not vanity. It is communication. But it wasn’t ‘murder-serious’, for the record…



The room then turned from team systems to team culture. Irena Zatloukalová's "How designing for autism improves your teams" argued for moving from culture fit to culture add. Clear agendas, explicit expectations, focus time and direct communication are inclusive practices that benefit everyone. Her message was simple: if teams rely on "implicit communication", they are in trouble.







Visibility, refactoring and adoption



External clarity became the subject of "From SEO to AIO", where Aneta Holá and Aleš Moravec described search moving from lists of links towards AI-shaped answers. "Today, the search engine basically becomes a decision engine," they explained. Visibility is no longer just about ranking. It is about whether AI systems understand, trust and represent a brand correctly.



In "Refactoring an entire product from scratch", Jan Toman brought the conversation back inside the product. Moving from Flutter to React was only part of the story. The harder work was prioritisation: understanding product depth, resisting the temptation to port everything, and deciding which happy path deserved the most care.



Lucerna Cinema closed with a business case for measured AI adoption. Tomáš Braverman and Romana Trusinová's "From threat to tool" showed how Slevomat approached AI through existing strengths, customer experience, outsourcing decisions and internal upskilling. Their advice to "stop, map what's already strong, and build up on that" felt like the room's final operating principle.







Key lessons from Lucerna Cinema




Clarity starts before production, whether in containers, contracts or customer journeys.



AI needs access rules, team norms and clear ownership.



Research matters most when it reaches decisions in time.



Readability, accessibility and explicit communication are product quality, not polish.



Good refactoring means choosing what not to carry forward.




Fewer hidden assumptions, fewer mismatched expectations, fewer expensive surprises. Where the Great Hall asked what humans still need to judge, Lucerna Cinema showed how teams turn judgment into repeatable practice.



Next: Marble Hall. If Lucerna Cinema was about making work clearer, Marble Hall asks what happens when AI, design systems, accessibility, pricing, DevOps and futures thinking all start moving faster at once.