OxyCon 2026: Key Takeaways

Welcome Word & Industry Overview

Vytautas Savickas, Group CEO @ Oxylabs

The day opened the way you'd hope it would – warm, direct, and with a clear sense of why everyone had actually shown up. Vytautas made the case for why the day's topics matter right now, backed by fresh numbers from a June 2026 global survey of 500 senior business decision-makers, conducted by Censuswide: 68% expect their web data needs to grow over the next year.

Your Next Dataset Starts with a Prompt

Andrius Kūkšta, Tech Lead @ Oxylabs

A strong, practical opener. Andrius walked through the all-too-familiar spiral of a "small" dataset request turning into a multi-day slog of site analysis, selector-hunting, and infrastructure headaches, and asked whether an LLM can really shortcut all that. His answer: only if you already know enough scraping to guide it. That's the gap Oxylabs' new AI-Guided Datasets solution closes, baking pagination logic, hidden API calls discovery, and selector-writing know-how directly into the model. Live on stage, we saw how an unfamiliar site can go from empty project to finished dataset in about half an hour.

Web Hydration: Loading Dynamic Content Without a Browser

Etienne Ellie, Lead Developer @ Scraping Bee

Etienne argued that the most expensive line in a scraping stack is often the one that launches a browser – not because it's slow, but because that cost repeats across every single page, at scale. He laid out a "cost ladder" for handling dynamic content: extract straight from the HTML when you can, replay the underlying API call when you can't, hydrate the page yourself with a lightweight DOM when there's no clean endpoint, and only reach for a full browser when you genuinely need a rendering engine.

Where AI Actually Helps in Production Scraping

Giedrius Šteimantas, Director of Scraping Engineering @ Oxylabs

Giedrius took on the "just use AI" reflex head-on. At Oxylabs, AI doesn't touch extraction itself, but works around the pipeline: keeping access reliable, helping parsers adapt as sites change, and clearing repetitive work off engineers' plates. He was candid that AI adoption can go two ways: real gains or a flood of "slop" – and that the fix starts with choosing the right metrics and listening to engineers rather than chasing raw output.

The Cost of Solving the Wrong Problem: Data-Driven Decisions for Web Access

Kieron William Spearing & Juan Manuel Perez, Engineers @ Centric Software 

Running thousands of crawlers means every hour of engineering time has to earn its place. Kieron and Juan shared their investigative mindset – a repeatable, almost forensic process for isolating exactly where and why a spider gets restricted, tested down to individual cURL requests, and used Akamai and DataDome as case studies for deciding where R&D effort actually moves the needle across a whole crawler fleet.

Agentic Workflows: Patterns Behind Successful AI Automation

Yenny Cheung, Vice President of Engineering @ Bluefish AI

Building for Fortune 500 clients leaves little room to experiment carelessly, so Yenny shared the framework Bluefish uses to bring AI into its dev lifecycle deliberately: automate the genuinely monotonous, keep humans in the loop where judgment matters, and keep people fully in charge where creativity and accountability count most. Her broader point – AI can't buy you taste, and as coding becomes commodity, planning becomes the step that determines quality.

The Adaptive Web: Turning Industry Challenges into Competitive Advantage

Panel: Pierluigi Vinciguerra, Andreea Stroe, Hocine Amrane, Andrey Gourine; hosted by Marija Gecaitė

A panel that came in ready to spar but landed on a surprising amount of agreement. Host Marija pushed the group on provocative framing, including whether "the agentic web" is more buzzword than reality, and the discussion settled into common ground more often than conflict, with a shared, cautiously optimistic read that plenty of "AI problems" can be solved with better automation and prompting alone.

We Made Our Web Search Worse on Purpose

Emilis Strimaitis, Head of Innovation @ Hostinger

Hostinger's five million customers mostly aren't engineers – they just want a working website and some competitor insight. Emilis explained how the team's launch instinct to feed their agent more (more search results, another tool, or additional agents) actually backfired: a model's context window is a fixed desk, and stuffing it with unrequested information just crowds out what the customer actually needs. Cutting the many agents down, as well as cutting search results down, both made the product simpler and more effective.

Is the Web Closing? What It Could Mean for Businesses, Journalists, and Researchers

Panel: Carl Miller, Paul Bradshaw, Krishna Sood; hosted by Denas Grybauskas

One of the most anticipated sessions of the day, given how many audiences it touches and how much uncertainty surrounds it, yet the panel came in confident and largely aligned. Krishna pointed to a growing asymmetry in data access across both policy and technology; Carl described a research field watching access shrink just as its analytical tools finally caught up; Paul reframed the debate entirely, arguing it's less about whether the web is closing and more about which parts need to stay open.

Best Open-Source Scraping Repositories of 2026

Tadas Gedgaudas, Founder @ TopYappers 

Closing out the night, Tadas put HTTP, browser, and agent-based scraping engines through their paces head-to-head, separating what holds up from what just demos well – capped off with a first look at his new project, ScrapingArena.org, a live leaderboard for scraping tools.

Was there a common thread?

Look back at the event as a whole, and the three themes this year’s OxyCon was built around (AI-native extraction, adaptive infrastructure, and governance) showed up everywhere, just rarely in the ways you'd expect. AI kept appearing as a tool that works around the hard problems rather than a shortcut past them, the fastest systems were usually the ones doing the least unnecessary work rather than the most, and the panels made clear that questions of web access and openness aren't going anywhere soon. 

Thank you, OxyCon 2026

A huge thank you to our 18 speakers, the guests who flew in from all over the world, the internal teams who pulled the event together, and our newest online audience members who kept us on our toes just as much as we hoped to impress them.

Keep an eye out for the on-demand videos, coming soon. And if you've got thoughts, suggestions for future events, or just want to tell us what this year's OxyCon was like from where you sat, drop us a line at events@oxylabs.io.