Mark Schreiber, PhD

Time for another coffee!

Something about shoulders and giants?

I strongly recommend using shellcheck ideally integrated with your IDE. It can find many potential bugs.

Breaking news: ChatGPT stops talking to humans fearing corruption of its foundation model.

It’s probably better that future AI models talk amongst themselves rather than risking talking to the likes of us.

I just watched it last week. Amazing find.

If they did the dishes and put the washing on this seems like a fair trade.

SGI and Sun workstations were the Mac book pros of their day. Only the rich kids had them or could afford them.

I was once playing with cellular automata and produced a pattern that looked very similar to fingerprints. Made me wonder if a similar process produces the real thing.

Looks more like a feather to me 🤔

Write drunk, edit sober - Hemingway (allegedly)

So doting over someone might actually mean being irrationally or foolishly nice to them?

I had assumed dotard referred to being old and dotage was old age but perhaps it is your foolish or feeble minded age?

Definitely not how it looked in my biology textbooks. The reality is so much cooler!

The hardest problem in database design is determining when two things are the same and when they’re different.

Add some pea flowers to the bottle

I’ll be there in five minutes!

Did you try tasting it? In the interests of science of course!

Qatar claimed that Iran forewarned them so it’s not impossible. I believe I saw a news article stating that the base was already evacuated. It also tracks with the idea that this was seen by everyone as a “de-escalation”.

One of The Oatmeal’s best so far.

Our hounds were very reserved in their patrolling of the premises.

Not to 💯! ☀️🥵

Lidderly!

Optimize HPC Workloads with the Updated AWS Well-Architected Lens | Amazon Web Services Today, we're announcing an update to the AWS Well-Architected High Performance Computing (HPC) Lens, providing you with comprehensive architectural guidance in a single, authoritative resource. This n...

Boost your HPC performance and efficiency on AWS with the newly updated Well-Architected HPC Lens! The new guidance consolidates years of AWS expertise into structured best practices for running HPC workloads on AWS. Read more about it at aws.amazon.com/blogs/hpc/op...

Step-by-Step Guide to Setting OIDC With Terraform for GitHub Actions Workflows with AWS Your GitHub Actions Secrets Are the Weakest Link in Your AWS Security Chain Are you still using...

dev.to/aws-builders...

Many examples use hardcoded credentials. These should never really be used. The good news is that in almost all cases there is a better way. When working with services like GitHub Actions and workflows you can use Open ID Connect (OIDC) for auth to AWS. (1️⃣/3️⃣)

🧵

Bold white chicken staring you down.

Eggatha don’t care!

Release v0.1.2 · awslabs/linter-rules-for-nextflow Release v0.1.2 Release Date: 2025-06-20 Changes chore: bump version to 0.1.2 (728467f) feat: add comprehensive release automation (59b94ce) Installation JAR Files linter-rules-0.1.2.jar ast-echo...

🧬🚀 Happy to announce linter rules for Nextflow v0.1.2. Featuring some fixes, improved testing, improved docs and better tooling. Also added the CICD friendly option to fail on lint violations. github.com/awslabs/lint...

It also depends on the task. You can exceed the human brain’s speed at any simple math problem with very little compute power. Complex reasoning, intuition who knows? It might be less than we think (or vastly more). We still beat generative AI in energy efficiency.

I’m less worried about LLMs being similar to us than I am with how similar we are to them.

Anthropic Researchers Ask the Question: What If Models Are Conscious? Anthropic has launched a research initiative exploring whether AI systems like Claude might one day warrant moral consideration—if not now, then soon.

www.maginative.com/article/anth...

Probably, but you start to get into the realms of needing to define what it means to care, why and how we care and if a complex enough system can care or imitate caring. If you prompt a model to care or ask it to favor responses with multiple lines of strong evidence is it still a bullshitter?