Krzysztof K. Zdeb

Every role wants to own the agent workflow. That's the problem. Why? Read here: ai.gopubby.com/beyond-vibe-co…

I treat every 🤖 agent mistake as a signal to improve a skill, not as random bad luck. Why? Read here: ai.gopubby.com/beyond-vibe-co…

Some think vibe coding is a skill. I think it's the absence of one. Why? Read here: ai.gopubby.com/beyond-vibe-co…

I just published Five Things I Expect in Data & AI in 2026 medium.com/p/five-things-…

If you think Delta Deep Clone is “just replication”, here’s a twist: it behaves a lot like a Git squash. It's a neat analogy I unpack in the Fun Fact of my DR for Azure Databricks article. Worth a read if you design data platforms 👇 blog.dataengineerthings.org/the-hard-art-o… #DataEngineering

Disaster Recovery for Azure Databricks is hard. And most teams only notice it when it breaks. I wrote a hands-on guide covering Deep Clone replication, Unity Catalog metadata sync, streaming recovery, URL switching, and clean failover/failback. Read here: blog.dataengineerthings.org/the-hard-art-o…

Read the article here: ➡️ blog.dataengineerthings.org/stop-partition…

Date partitioning in #Databricks is the way to go in many cases. But if done incorrectly, it can become a bottleneck. My latest article dives into common pitfalls and smarter methods you can apply.

@geekwhocodes There's an ongoing incident in West Europe: status.azuredatabricks.net/pages/incident…

is #Azure #Databricks down? status.azuredatabricks.net says up and running isdown.app/status/azure-d…

Surrogate keys generated through hash functions bring efficiency but also come with the risk of hash collisions. In @kzdeb's article, dive into how common hash functions like MD5 and SHA-256 can lead to collisions and explore the birthday paradox. towardsdatascience.com/collision-risk…

Learn about the potential risk of hash collisions when using hash functions like MD5, SHA-1, and SHA-256 to generate surrogate keys in data warehouses and lakehouses in @kzdeb's latest article. towardsdatascience.com/collision-risk…

Are the reasoning capabilities of current OpenAI LLMs good enough to play a classic guessing game? Krzysztof K. Zdeb shares insights based on his experiments with playing forehead detective with several GPT models. buff.ly/4dNPoOv

@alexpotato @canalCCore2 There is. LLMs work well with ternary weights. See: arxiv.org/abs/2402.17764

@canalCCore2 Isn’t there evidence that you can make neural nets with integer weights? If yes you wouldn’t even need the variable intensity: just on or off.

If we built a neural network where the weights were lenses instead of vectors, for instance, and the input was light-shaped, the inference cost would be zero.

Here's my new publication on dimensional modeling of hierarchies for Data Warehouses. Happy reading!

Best Practices for Technical Columns in Database Design - When architecting a transactional database or a data warehouse, it’s important not to forget about various types of technical columns. 🖋️ by Krzysztof K. Zdeb buff.ly/4byev6p

It all started with 🤖 GPT having an input context window of 512 tokens. After only 5 years the newest LLMs are capable of handling 1M+ context inputs 🤯. Where’s the limit? Read full article: 👉 medium.com/towards-data-s…

It all started with GPT having an input context window of 512 tokens. After only 5 years the newest LLMs are capable of handling 1M+ context inputs. Where’s the limit? by Krzysztof K. Zdeb buff.ly/4bfsQEM

"Whenever I have to assign a classword to a column name, it forces me to pause and think deeply about the data that column will hold. What does this data truly represent?" by Krzysztof K. Zdeb buff.ly/3wDVTCG

The AI Community-led survey sheds light on AI's impact on PM. A standout finding is the eagerness of mid-career professionals (45-54 yo) to adopt AI, with 83% ready to lead AI-driven projects. Full survey report: pmi.org/-/media/pmi/do…

Join the bustling AI Community at PMI, where nearly 7k subscribers exchange knowledge & experiences. Since 2016, the community has seen a sharp increase in activity with 350+ threads last year alone (vs 40 ever before). Engage with fellow PM enthusiasts at projectmanagement.com/topics/artific…

A as proud member of @PMInstitute since 2011 (and holder of PMP & ACP certifications) I'm thrilled that PMI is at the forefront of the #GenAI revolution. Let's dive in! 🌐✨