Meta Buys Half of Scale AI, Apple Polishes Glass and Veo 3 Crashes the NBA Finals

Zuck drops $14.8B on ‘superintelligence’, Apple skips Siri upgrades and AI-generated ads hit primetime TV

Fellow Data Tinkerers!

It’s time for another weekly round-up on all things Data and AI. But before that, I wanted to mention If keeping up with the latest news is not your jam, you can click on settings and turn it off (Data Bytes) and keep receiving emails for the other topics you are interested 👇

I’d rather you receive fewer emails that you are interested in than getting stuff you are not crazy about :)

Now, with that out of the way, let’s get to this week’s round-up!

---

The Buzz 🐝

• While it looks like Meta is struggling with keeping some AI talent, Zuck’s thrown $14.8B at Scale AI for a 49% stake. Their CEO, Alexandr Wang, is now heading up a new “superintelligence” lab at Meta. But the deal will rattle Scale’s other clients like Google and Microsoft who aren’t exactly thrilled about their go-to vendor cozying up with the competition. Oh, and Meta also launched V-JEPA 2 - a vision model that helps robots understand the real world. It’s meant to teach machines to think before acting. (Some would argue that’s advice Meta could’ve used earlier)

• Apple’s WWDC this year was more “refinement” than revolution. Highlights included live AI translations for calls and messages, a personalized Workout Buddy for Apple Watch and smarter Maps that learn your routines. Oh, and of course the biggest “breakthrough”, the Liquid Glass. It looks like Apple is having a Windows Vista moment!

  

  (Source: Twitter)

  But with no sign of the AI-powered Siri teased last year, some walked away feeling it all lacked punch, especially with competitors racing ahead on frontier AI models.

• Kalshi aired a fully AI-generated ad during the NBA Finals created in just two days for $2,000 using Google’s Veo 3. Directed by PJ Accetturo, the spot featured surreal scenes from beer-chugging aliens to egg-filled pools as a metaphor for a world gone mad. It’s probably a glimpse into the future of advertising: fast, cheap and easy to produce. You can check it below 👇

---

Data Science & AI

• Optimizing LLM-based trip planning
  Learn how Google combines Gemini LLMs with optimization algorithms to plan smarter and more realistic trips. The LLM captures user preferences while the algorithm fine-tunes the itinerary to fit real-world constraints like opening hours, travel time and scheduling feasibility.

  

  (Source: Google)

• How Uber Cut Invoice Handling Time by 70% with GenAI
  Uber used GenAI to tame its chaotic invoice mess: thousands of formats across 25+ languages. The new system cut processing time by 70% and costs by 30%, proving GenAI can do more than just generate hype.

• PandasAI
  PandasAI turns your pandas DataFrame (or SQL, CSV, Parquet source) into a chat interface. Just plug in an API key, run df.chat() and get instant answers or visualizations even across multiple DataFrames.

---

Data Engineering

• Databricks Data + AI Summit 2025
  Last week was Databricks annual summit and there are even more ‘Lake’ products now! They dropped Agent Bricks for building domain-specific AI agents and MLflow 3.0 with GenAI-native features. Lakebase, their new Postgres-compatible transactional database, and Lakebridge, a free warehouse migration tool, were also announced. Plus, Databricks Free Edition now lets students and newcomers get hands-on with data and AI tools at no cost.

• Lakebase from Databricks
  

  Daniel Beach
  does a good review of Lakebase, the new Databricks product which is a new fully managed Postgres OLTP engine built into Lakehouse, blurring the line between transactional and analytical workloads.

---

Data Analysis and Visualisation

• Keep your BI tool as dumb as possible
  Interesting article by

  Tim Hiebenthal
  about why your BI tool should stay dumb and just visualize

• AI Chatbots vs Search Engines
  Interesting visualisation of the web traffic

  

---

Crappy Data Magic!

(Source: LinkedIn)

---

---

Keep reading: