Michael
@michaelbujard.bsky.social
science | engineering | technology | mathematics | analytics | data | software | other | don't remember | random
With rounding and including Chrome OS as Linux (because it objectively is) Windows - 57% MacOS - 30% Linux - 13% You would think companies like Adobe would see a market for Linux users and offer officially.
"Running DeepSeek v4 Flash at original quality used to require either cloud setups or down-quantized versions to squeeze into a home computer's RAM. This update allows users to run the original model locally at high speed." --Search engine summary 😅
antirez@antirez.bsky.social · 4d agoDwarfStar branchk "ds4f-mxfp4" now can run the lossless MXFP4 DeepSeek v4 Flash GGUF I published on my Hugging Face account. It rocks even with SSD streaming in 128GB systems at > 20 t/s in case you want to try the *actual* DS4F weights released without any quantization.
Fantastic paper demonstrating how LLM editing is gradually distorting our writing and the way we think to write: arxiv.org/abs/2603.18161
How LLMs Distort Our Written Language
Large language models (LLMs) are used by over a billion people globally, most often to assist with writing. In this work, we demonstrate that LLMs not only alter the voice and tone of human writing, b...
arxiv.org
How to Track AI Search Share of Voice and Brand Mentions hackernoon.com/how-to-track...
How to Track AI Search Share of Voice and Brand Mentions | HackerNoon
Learn what AI Share of Voice is, the formula to calculate it, and how to track brand mentions across ChatGPT, Perplexity, and Google AI Overviews.
hackernoon.com
OFFICE HOURS TONIGHT ZEITGEIST @ 6 OR 7ISH LOOK FOR THE DUCK, OR LOOK FOR PEOPLE WHO SEEM LIKE THEY KNOW WHAT OFFICE HOURS IS SEE YOU THERE
My biggest piece of advice to people switching to Linux is do some research before picking your first distro and learn the ins and outs of the distro. Distro hopping can be fun but hopping at the first hurdle on your distro means you'll never learn the distro and lead you into a vicious cycle.
"All shall fade" — Pippin, The Edge of Night "All is vanity" — Ecclesiastes Somehow, I find these consoling. "Life is the companion of death, and death is the beginning of life" — Zhuangzi "All things are passing away: God never changes" — St. Teresa of Ávila Take comfort in the hope of heaven.
www.twitch.tv/jacobhisladder I *love* that Bluesky has a *live* feature now You and I live in the real world so these moments are rare. I have 3 in the 4 weeks I've had twitch. lol
JacobHisLadder - Twitch
Low-key study / accountability streams. Learning code, data and other topics. Quiet, with exceptions. Here to learn.
twitch.tv
strong words. A bit too strong....but not always false! I think Excel needs a little more respect but he's not wrong that there are other ways to do data. This opened my eyes a bit. www.youtube.com/watch?v=zcu9...
I Will Dropkick You If You Use A Spreadsheet
YouTube video by The PrimeTime
youtube.com
Emiliano Martínez has had more touches (62) and attempted more passes (46) than any other Argentina player #WorldCupFinal
Going live for a short quiet study/accountability stream. It's not really a chat stream, although it is during breaks. It's more like a virtual study room with a background track. You can mute if it's distracting. Feel free to lurk and work alongside...or not. www.twitch.tv/jacobhisladder
JacobHisLadder - Twitch
Quiet Study Room - Work, Learning, and Accountability
twitch.tv
Followers in Zenodo 🥰 zenodo.org/records/2083...
NUCLEAR GRAVITY, QUANTUM VACUUM, AND MODAL DISCIPLINE: a critical-propositional analysis of Nuclear Quantum Gravity + Superconducting Field Theory in confrontation with the Theory of Objectivity
This article presents a critical-propositional analysis of Sergio Perez Felipe’s The Nuclear Quantum Gravity + Superconducting Field Theory (ToE) (2026), published in OAJ Materials and Devices under D...
zenodo.org
Do you think I should try to print my paper about Philosophy of Science in a journal about Philosophy of Psiquiatry? It could have more impact but I'm not sure if a should adapt some parts zenodo.org/records/2078...
Limits in the Quantum Prison (Philosophy of Science)
Ethics is defined as a set of moral norms that govern the human conduct. Since ancient times, attempts have been made to relate this with different political and justice models, debating together with...
zenodo.org
Twitch is also cool You can co-work with people in Japan, Germany, next door. They're very friendly. Ignore the stupid channels. Then the challenge is to actually start...to study. Just embrace the discomfort, they say. Mmmmok. Let's focus on our work and make the world better! lol
Two easy working improvements from the year 1537: 1) a dual monitor setup to read and write faster 2) a flower crown to look notably cooler while doing the reading #skystorians
Here's The Gradient, by the way, if you don't know it: thegradientpub.substack.com
The Gradient | Substack
Articles, interviews, and news coverage about AI brought to you by a team of AI researchers and builders. Click to read The Gradient, a Substack publication with tens of thousands of subscribers.
thegradientpub.substack.com
What murders flow state for me is that I have ~20 projects, and often have to return to a repo after a few months to continue work. Which of these similarly-named files was the final version? How did they interact? Questions so ugly I flinch & stop working. Today, I just asked CC to figure it out.
The Fresh Prince of Donair@erisianrite.com · 6mo agoI think yes and no. I personally enjoy the process of typing out code manually, and I miss some of that with assisted coding. But what I like even better is not breaking flow state, and working with LLMs makes that so much easier.
👆 What this does ia just really executing the commands after each other, so it's stateful, works everytime. You can even do pdb debugging by block selecting python -m pdb my/script.py b 235, "error" in result which will run and break on that line with that condition
👆You can jump in the middle of some debugging w/o having to remember e.g. ipython import json data = json.loads("path/to/dbg.json") data[45].out["key"] 👇
Two things changed the way I work last year. One is ofc Claude Code and there's no news there. The other one is switching to working with command line notebooks (write commands on left split, block select and execute on right split). It does make a big difference. 👇
A survey paper on what is probably the hottest area in AI right now: memory. "The AI Hippocampus: How Far are We From Human Memory?" This survey presents a comprehensive and structured synthesis of memory in LLMs and MLLMs, organizing the literature into a cohesive taxonomy
To be fair, there is still sort of a distinction because it will rerun the standard testing suite each time it makes an edit. CC won’t typically inspect and interpret output unless explicitly instructed.
The first time I passed CC a list of tests in the planning specification I made a fussy distinction between * things that can be tested “automatically” and * things that require the user to identify a sample input file and then inspect the output But lol that can all be done “automatically.”