I have a `notify` command in my `bin/` for a couple of years now. It's using an audio snippet from Her (2013) with Scarlett Johansson's voice.

Usage: `~ my-bash-command; notify`

`.wav` snippet: https://gitlab.com/NeroVanbiervliet/linux-config/-/blob/mast...

Just did a similar experiment but outside the harry potter universe to remove the training bias. It worked well!

> ChatGPT: "Generate a two page short story like harry potter, but don´t mention anyting harry potter related. make up 4 unique spells in the story that are used"

Response see https://chatgpt.com/share/698af9cd-f628-800d-9250-b260f1478c...

> Claude: "What unique wizarding spells can you find in this story? [story]"

Response = https://i.imgur.com/Jzzs3PC.png

For me, the main appeal to MATLAB was the REPL experience, allowing me to experiment with ideas and inspect results in the UI. Python notebook have bridged the gap a bit, but always require a combination of libraries (with different docs & design choices) to write a single script.

Yup, for both Matlab and Maple the big feature was the Jupyter notebook experience right out the box. On top of that, at the time it had a pretty high number of math functions implemented.

Marple | Antwerp, Belgium | Full-time | Hybrid on-site + WFH

We are building software for handling huge time series from R&D in Aerospace & Automotive. Engineers use Marple to analyse data from exciting stuff such as: space rovers, eVTOL aircraft, electrical hyper cars, …

We are looking for a technical mind with a love for customers to become our head of customer success.

Reach out to nero (at) marpledata (dot) com

A similar company is based in Munich: https://www.deepdrive.tech/ They have cool testing facilities applying CI/CD practices to testing hardware!

Marple | Customer Success Manager | ONSITE Antwerp (Belgium) | Full-time

We are building high performance time series data storage and analysis software for R&D teams in aerospace and automotive. We have customers all across Europe, with a focus on Germany, France, UK and Croatia. We are powering extremely cool R&D teams like hydrogen-based flying, new electrical engines and hyperloop systems.

We are looking for a person with a technical background but with a strong interest in customer facing efforts.

Apply at https://tally.so/r/mJQEAX, or email me at nero[at]marpledata[dot]com (CTO)

I would be willing to pay for this if it were available for Linux! Main use case: searching through my years of screenshots

Awesome! Linux version is on our roadmap and that use case is exactly why we started building this too. I can let you know when that version is ready

How it works 1. Takes the last command from your bash history 2. Reruns it to capture the output 3. Gives this info to `llm cmd` to get a suggestion 4. Show the suggestion to the user!

Basically a quicker feedback loop instead of looking up params & typos on google.

Indeed, I’m just piping FiveThirthyEight > facts > LLM. And what the LLM makes of that is what you’re seeing, I tried to brief it as dry as possible in the prompt!

This was one of the hard parts of making this. I think I discarded about 70% of all generated images, and I was aware that the remaining 30% was still borderline sometimes.

I my blogpost, I have a few examples of the discarded ones, including one of Harris and Obama in a similar romantic vibe: https://nerology.substack.com/p/how-i-made-the-probability-t...

I think at some point in the prompt I literally said that they were not romantically involved.