Beyond the prompt | S01 E04 - Conclave edition
May 31, 2025 2:11 pm
Hi there,
Welcome back to Beyond the Prompt!
Things are getting busier at Agilytic HQ.
Why, you ask? Because a lot of organizations have been waiting, and waiting since January. And now that summer holidays are approaching, things need to get done by tomorrow.
OK, I might be exaggerating here, but by how much? Let me know. 😇
🤗 Before we start: we're (still) hiring!!!
Besides the usual Data Science and Engineering gigs, we have two exciting opportunities open right now.
- An Administrative and Finance Manager (part time) to help me steer the ship in an efficient way.
- A Data Engineering Manager to grow this thriving practice alongside our brilliant experts.
💡 From the case studies
- Geographical Information Systems are rightfully getting a lot of traction. The tech is mature, and the use cases are convincing. We're totally here for it, this time for a major Telecom institution.
- Ad tech has been doing optimization since day 1. We were happy to be part of this never-ending challenge around audio content.
📜 From the blog
- We are very disciplined in knowledge management. Here's why, and a bit of "how", courtesy of Clément.
- There are so much options in choosing LLMs that Yoann gifted us this quick guide to help you decide between open and closed-source.
- The text version of my brilliant (hum...) appearance on our podcast.
🎙️ From the podcast
This month, Caroline discusses her background, managing data projects, and the role of Chief of Staff.
🧪 From the lab: What we're cooking
- Chatbot Development: Centralized, Scalable, and Secure: we're experimenting with a centralized chatbot solution, making chatbot creation and sharing easier than ever. It includes no-code simplicity, scalability, security, and transparency.
- We're reviewing MLflow to supercharge our machine learning workflows: experiment tracking, model comparison, and version control.
- Quick Roadmap Methodology 2.0: we're preparing a revised methodology for quick roadmapping projects, focusing on client needs and budget alignment.
🌐 From everywhere else
News from the big players
- OpenAI drops over $6 billion to acqui-hire Jony Ive's hardware design capability. There's a lot of hype around this one, but keep in mind it's $6bn of equity, a.k.a. "paper money". Tech darwinism favors the phone format, but I'd be happy if I'm proven wrong. It's been a while since the hardware scene had a truly game changing innovation.
- Claude 4 is released: Big Number release means it's a meaningful upgrade, but I think the bottleneck now lies in realizing the use cases. Last year's models were already good enough for a lot of ideas.
- Microsoft Build conference was heavy on Copilot Studio and Fabric.
- Google I/O: if a tech publication needs to summarize a keynote in "15 biggest announcements", you might have an overcrowded pipeline... all of it AI, of course. Android was relegated to a side show with a dedicated session.
Insurers at Lloyd's have launched a new product to cover losses caused by AI chatbot errors, addressing growing concerns over the risks associated with these technologies.
The Guardian: ‘It cannot provide nuance’: UK experts warn AI therapy chatbots are not safe. Well, ya think?
ChatGPT's emergence has fundamentally reshaped the field of natural language processing, prompting both existential reflection and rapid adaptation among researchers. Also, I have a soft spot for "oral history" articles and books.
Habemus papam. Bocconi researchers used network science to analyze the dynamics of the papal election, revealing how status, information, and alliances might have shaped the conclave's decision-making process.
Damien Charlotin is compiling legal decisions in cases where generative AI produced hallucinated content – typically fake citations, but also other types of arguments. The list keeps on growing...
👋 Parting thoughts
As we near the summer break, remember: every project deadline is really just a mini-conclave. Senior stakeholders parade their scarlet slide decks, informal alliances form over coffee, and somewhere in the corner a data scientist lobbies for Bayesian priors like a forgotten cardinal from Avignon. When the dashboards finally align and the status light turns green, that’s our fumata bianca—the blessed white smoke of “approved, ship it.”
Until then, may your factions stay friendly, your lobbying stay ethical, and your chatbots produce fewer imaginary papal bulls. Pax et queries vobiscum!
All the best,
Julien
_
PS: Feedback is a gift. Tell me what you liked (and liked less) simply by replying to this email.