Marius Soutier | Data Engineer
@mariussoutier.bsky.social
Data Engineer | Machen Sie mehr aus Ihren Daten! | Selbständiger IT-Berater | Scala, Kafka, Spark, Akka
"Serverless is coming to Scala" on Databricks! I was worried they had abandoned Scala for good and were focusing only on Python.
www.youtube.com/watch?v=ndX4...
www.youtube.com/watch?v=ndX4...
Use Scala to develop and deploy Lakeflow Jobs on serverless
YouTube video by Databricks
www.youtube.com
November 11, 2025 at 8:07 AM
"Serverless is coming to Scala" on Databricks! I was worried they had abandoned Scala for good and were focusing only on Python.
www.youtube.com/watch?v=ndX4...
www.youtube.com/watch?v=ndX4...
Slack AI summary of a recent discussion at work:
"Unclear Discussion Topics. The conversation is highly fragmented, with the participants jumping between various unrelated topics without clear context or purpose.
...
"Unclear Discussion Topics. The conversation is highly fragmented, with the participants jumping between various unrelated topics without clear context or purpose.
...
July 23, 2025 at 7:48 AM
Slack AI summary of a recent discussion at work:
"Unclear Discussion Topics. The conversation is highly fragmented, with the participants jumping between various unrelated topics without clear context or purpose.
...
"Unclear Discussion Topics. The conversation is highly fragmented, with the participants jumping between various unrelated topics without clear context or purpose.
...
Reposted by Marius Soutier | Data Engineer
I also think it’s very dangerous to assume that actual humans will keep contributing useful information and code to Stack Overflow and GitHub at the *same current rate that LLMs were designed to depend on*, in a world where LLM tools increasingly dominate the Internet and the tech economy
there’s this assumption that actual humans will keep posting useful human-generated information on Reddit forever in places where we, as other humans, can easily find it - but I do not think this is actually the case
Relying on people passively posting random material on the internet about their vacations faces a whole lot of fundamental filtering problems.
And it’s also crucial that the LLMs themselves have been trained massively on highly-detailed travel blogs produced by people who do that professionally.
And it’s also crucial that the LLMs themselves have been trained massively on highly-detailed travel blogs produced by people who do that professionally.
July 21, 2025 at 3:54 PM
I also think it’s very dangerous to assume that actual humans will keep contributing useful information and code to Stack Overflow and GitHub at the *same current rate that LLMs were designed to depend on*, in a world where LLM tools increasingly dominate the Internet and the tech economy
AGI maybe in 2250; LLMs aren't thinking, only inferring; a tool doesn't have to be a human to be useful. Great interview by @joereis.bsky.social
www.youtube.com/watch?v=E-nF...
www.youtube.com/watch?v=E-nF...
Beyond Generative AI: John K. Thompson on AGI, Data Modeling, and the Future of AI
YouTube video by Joe Reis
www.youtube.com
July 9, 2025 at 1:03 PM
AGI maybe in 2250; LLMs aren't thinking, only inferring; a tool doesn't have to be a human to be useful. Great interview by @joereis.bsky.social
www.youtube.com/watch?v=E-nF...
www.youtube.com/watch?v=E-nF...
Indeed, LLMs helped me fix some nasty JVM-Scala-interaction bugs in my code. On the other hand, I couldn't figure out a simple error message where I was using a dependency built with Scala 2.12 in a Scala 2.13 project.
The fact that I can basically just toss extremely annoying typescript "the thing works but it won't technically build" errors at LLMs and they can almost always fix them now is actually making me do web side projects now. I would have never wanted to put up with that shit in the past
June 30, 2025 at 4:59 AM
Indeed, LLMs helped me fix some nasty JVM-Scala-interaction bugs in my code. On the other hand, I couldn't figure out a simple error message where I was using a dependency built with Scala 2.12 in a Scala 2.13 project.
IMHO, AI-based autocomplete results in small productivity gains. However most of your time is spent thinking, where AI can help you, or also distract you by giving you false information.
June 29, 2025 at 7:16 AM
IMHO, AI-based autocomplete results in small productivity gains. However most of your time is spent thinking, where AI can help you, or also distract you by giving you false information.
Latest episode from Infra Pod is really great, a very humbling view on security and leadership. Also puts AI into perspective (less hype, more realism).
open.spotify.com/show/7qZ7hDm...
open.spotify.com/show/7qZ7hDm...
The Infra Pod
Podcast · The Infra Pod · The Infra Pod brings you insightful and thought-provoking discussions on the world of infrastructure software. This podcast is started by two engineers, Ian Livingstone (tech...
open.spotify.com
June 24, 2025 at 10:51 AM
Latest episode from Infra Pod is really great, a very humbling view on security and leadership. Also puts AI into perspective (less hype, more realism).
open.spotify.com/show/7qZ7hDm...
open.spotify.com/show/7qZ7hDm...
Oh my God, yes :-)
June 23, 2025 at 5:22 AM
Oh my God, yes :-)
Reposted by Marius Soutier | Data Engineer
left: the interview questions
right: the actual job
right: the actual job
June 22, 2025 at 3:20 PM
left: the interview questions
right: the actual job
right: the actual job
Reposted by Marius Soutier | Data Engineer
AI investing is crazy.
- everyone knows AI infra is rebuilding microservices patterns
- mcp is a mess
- valuation is off the charts
- space is changing so fast
- OpenAI is a product company (takes consumer)
- Microsoft takes enterprise
- building on LLMs is like building on mobile (good for indy)
- everyone knows AI infra is rebuilding microservices patterns
- mcp is a mess
- valuation is off the charts
- space is changing so fast
- OpenAI is a product company (takes consumer)
- Microsoft takes enterprise
- building on LLMs is like building on mobile (good for indy)
June 18, 2025 at 7:25 PM
AI investing is crazy.
- everyone knows AI infra is rebuilding microservices patterns
- mcp is a mess
- valuation is off the charts
- space is changing so fast
- OpenAI is a product company (takes consumer)
- Microsoft takes enterprise
- building on LLMs is like building on mobile (good for indy)
- everyone knows AI infra is rebuilding microservices patterns
- mcp is a mess
- valuation is off the charts
- space is changing so fast
- OpenAI is a product company (takes consumer)
- Microsoft takes enterprise
- building on LLMs is like building on mobile (good for indy)
Reposted by Marius Soutier | Data Engineer
Little Bobby Tables had a brother
June 12, 2025 at 1:13 AM
Little Bobby Tables had a brother
Is it just me or is Databricks a very convoluted way of using Spark? I now have multiple bash-based cluster init-scripts to only to configure very basic stuff (replace outdated libraries, customize logging, adding a Spark listener, and so on).
April 3, 2025 at 7:13 AM
Is it just me or is Databricks a very convoluted way of using Spark? I now have multiple bash-based cluster init-scripts to only to configure very basic stuff (replace outdated libraries, customize logging, adding a Spark listener, and so on).