July Edition

🎉Our holidays welcome event is here!

We’re glad to announce our next Claude Code Barcelona meetup! Don’t miss these two talks plus a great networking session, all thanks to our new host Runroom!

⚠️ Spots are limited. We recommend reserving your spot as soon as possible. If you’re unable to attend, please let us know in advance so we can release your seat.

📅 Jul 8th | Doors open at 6:00 PM, kick off at 6:15 PM
📍 Carrer de Santa Eulàlia, 5-9, 3ª planta, Gràcia, 08012 Barcelona

Agenda
➡️ 6:15 PM — Check-in
🏢 6:30 PM — Host company presentation
🎤 6:40 PM — Code is cheap. Now what? by Chris Ford, in English
🎤 7:20 PM — The parrot is eating the harness: how LLMs are being trained and what it means for your AI applications by Gabriel Mesquida, in English
🤝 8:00 PM — Networking

***

🎙️Talks

Code is Cheap: Now what? (In English)

Coding agents like Claude Code make a big difference to how individual developers do their job. Some activities, in particular those relating to making code, are a whole lot faster.

Coding agents also make a big difference to how we run engineering organisations, because they have traditionally been shaped by the constraint that code was very expensive to produce. Examples of how organisation-level processes change include how we treat legacy code, how we design architecture and how we decide whether to buy something off-the-shelf or build it ourselves.

The changing economics of software will reshape the context within which every software developer works. If you understand these dynamics, you have a better chance to take advantage of the opportunities.

By Chris Ford, Technology Director at Thoughtworks. He's the author of the upcoming O'Reilly book 'Agentic Engineering at Scale', which is all about getting value from coding agents beyond developers' personal productivity.

The parrot is eating the harness: how LLMs are being trained and what it means for your AI applications (In English)

In this presentation we will review what LLMs are and how they have been evolving. From the initial transformers to the latest evolution trained with Reinforcement Learning and with more efficient latent attentions. At some point, they needed to be incorporated into a wider, integrated system to make them useful, managed by code.

And that’s when harnesses were born: crewAI, strands, openclaw and also our beloved Claude Code. They all have something in common: they ride the parrot so that the LLM does what we want, in the way that we want, adding instructions, identity, guardrails, memory, persistence. Include in the list of harnesses any AI application that your startup, or your company is building on top of an LLM. But, where did harnesses come from and what do they do?

Now for the main premise: training by Reinforcement Learning is so powerful that, in the last few months LLMs have been evolving incorporating functions that harnesses used to do for them. What has happened exactly and what does this mean for us, AI developers and AI entrepreneurs?

By Gabriel Mesquida, course facilitator and teaching assistant at Stanford Engineering Center for Global & Online Education

***
We’d like to thank our sponsor, Runroom 🙌, for providing the space, snacks, and drinks! 🍕🧃

Don't miss it! 🎉
----