iiinova — where your digital self keeps evolving

Reposted by Rasmus Aagaard

NobodyWho @nobodywhooo.bsky.social · 29/09/2026

NobodyWho in 25 seconds 🎬 Run open-weight AI models fully offline: 💬 Text generation 🖼️ Image + audio ingestion 🔊 STT, TTS & VAD 🔎 Embeddings & RAG Python, Flutter, RN/Expo, Swift, Kotlin, Godot... No cloud. No API keys. Open source. docs.nobodywho.ooo

021

Reposted by Rasmus Aagaard

Rotten Tomatoes @rottentomatoes.com · 23/09/2026

All-new trailer for Nathan Fielder and Lance Oppenheim's 'You Can See Everything.' The Elizabeth Holmes documentary is in theaters on October 16.

3241592

Rasmus Aagaard @rasgaard.com · 22/09/2026

As I'm approaching the end of my 1st year as a PhD student I'm struggling a bit with how coding agents affect my learning. In that light I found Tim Dettmers' new post very good and inspiring! "Let go of how you work. Not who you are." timdettmers.com/2026/09/21/d...

010

Reposted by Rasmus Aagaard

Kenneth Nørgaard @kennethnrgaard.bsky.social · 12/09/2026

Kan vi blive gladere af at forbruge mindre? Seniorforsker Toke Haunstrup mener, at nøjsomhed kan være en vej til både større trivsel og et mindre pres på kloden.

dr.dk

Kan vi blive gladere af at forbruge mindre? Det mener forsker

Seniorforsker Toke Haunstrup mener, at nøjsomhed kan være en vej til både større trivsel og et mindre pres på kloden.

10483

Reposted by Rasmus Aagaard

Simon Eiriksson @simoneiriksson.bsky.social · 24/08/2026

Happy to announce the sixth session in Copenhagen Advanced Machine Learning Study Group (CAMeLS-Group 🐪). If you live in Copenhagen you are welcome to drop by and read the magnificent paper Bayesian Learning via Stochastic Gradient Langevin Dynamics by Max Welling and friends. luma.com/mhattbxq

luma.com

CAMeLS-Group: sixth session in Copenhagen Advanced Machine Learning Study Group · Luma

Welcome to the sixth instalment of the Copenhagen Advanced Machine Learning Study Group (CAMeLS-Group). We meet every second Monday to read and discuss a…

041

Rasmus Aagaard @rasgaard.com · 16/08/2026

Presented our work on layer pruning of the Whisper encoder. We show that if we’re careful about which layers we prune away, we can remove 6 (18.5%) of them without major performance drops and even recover performance through simple distillation.

020

Rasmus Aagaard @rasgaard.com · 16/08/2026

Was at @ijcai.org yesterday for the GLOW workshop glow-ijcai-2026.github.io/glow-ijcai-2... :)

160

Rasmus Aagaard @rasgaard.com · 14/08/2026

Had a really good time at the Danish Data Science Academy's Pre-Graduate retreat :) Was invited to participate in a panel discussion about the different ways of doing a PhD, representing the industrial PhD programme. Can highly recommend it if you’re a MSc student!

ddsa.dk

DDSA Pre-Graduate Retreat 2026

Are you a bachelor’s- or master’s student with data science as a key component of your study programme? Would you like to know more about pursuing a PhD in data science?

030

Rasmus Aagaard @rasgaard.com · 14/08/2026

Maybe @kasperjunge.bsky.social has thoughts on this or seen something out there in the wild? 👀

120

Rasmus Aagaard @rasgaard.com · 13/08/2026

Any examples of orgs creating an internal SWE-bench? Curious about having ways to measure when local models reach a “good enough”-stage to be viable of x% of relevant coding tasks and this seems to be a promising way to communicate capabilities internally

362

Reposted by Rasmus Aagaard

Kenneth Enevoldsen @kennethenevoldsen.eurosky.social · 05/08/2026

Embedding models keep getting cheaper - what required ~300M parameters three years ago now requires only 11M. In practice, that means that what runs on the server today could run in the browser tomorrow.

0222

Reposted by Rasmus Aagaard

Tima Kunayev @dj.kz · 04/08/2026

Me and my agents, just out of frame, are having fun in this harness, laughing

0251

Reposted by Rasmus Aagaard

NobodyWho @nobodywhooo.bsky.social · 03/08/2026

NobodyWho Chat is live on iOS & Android! 🥳 A fully offline AI assistant. No internet, no accounts, no cloud. Run #Gemma, #Ministral, #Granite, #LFM, #Qwen and more, entirely on your device. Vision and hearing support included. Fully open-source. Links in comments 👇

222

Rasmus Aagaard @rasgaard.com · 02/08/2026

It’s short for dynamic so dyna would make way more sense. I have found that climbers have a curious way of naming things:)

110

Rasmus Aagaard @rasgaard.com · 02/08/2026

I guess my account is now dedicated to posting my latest dyno #bouldering

130

Rasmus Aagaard @rasgaard.com · 31/07/2026

That first move took about a million attempts #bouldering

040

Rasmus Aagaard @rasgaard.com · 20/07/2026

Probably my hardest climb so far. Super stoked:) Guess the grade? 🧐 #bouldering

020

Rasmus Aagaard @rasgaard.com · 20/07/2026

Jeg kan anbefale Radio4s podcastserie “Magtfuld” for at få et hurtigt indblik :) Biografien The Contrarian af Max Chafkin er også god. radio4.dk/podcasts/mag...

radio4.dk

Magtfuld - Peter Thiel (1:4) - Dungeon Master

Den stenrige tekno-kapitalist og kongemager Peter Thiel er en af de mest indflydelsesrige folk omkring Trump regeringen. Han vokser op præget af stren

270

Reposted by Rasmus Aagaard

Ai2 @ai2.bsky.social · 02/07/2026

The Danish Foundation Models (DFM) project is adapting our modular FlexOlmo architecture into a lighter-weight system that runs on commodity hardware—putting collaborative model building within reach of smaller research groups & organizations. 🧵

1329

Rasmus Aagaard @rasgaard.com · 18/06/2026

Got an extended abstract paper accepted at the GLOW workshop at IJCAI :) Turns out you can just delete 6 layers from the Whisper encoder and it doesn’t hurt transcription performance all that much (but you have to be pretty specific about which layers)

040

Reposted by Rasmus Aagaard

Vicki @vickiboykis.com · 15/06/2026

new post: how I develop recently using local models. the tooling is now good enough to do agentic workflows and everyone should give them a try! vickiboykis.com/2026/06/15/r...

vickiboykis.com

Running local models is good now

Local agentic coding has gotten great over the past few months

1825737

Reposted by Rasmus Aagaard

Vicki @vickiboykis.com · 11/06/2026

The machine learning circle of life

1. This is too big, we need to compress it
2. We compressed it but now it's too small, we need it to generalize
3. It generalizes but it takes too long, we need to make it faster
4. We made it faster but now we have a distributed systems problem, we need to coordinate it
5. We've coordinated it but now we've created a monolith
6. This is too big, we need to compress it

2868

Reposted by Rasmus Aagaard

Gergely Orosz @gergely.pragmaticengineer.com · 10/06/2026

Seeing how SOTA models are evolving: becoming more restrictive in usage (decided by the company), less transparent (you cannot tell if the AI lab nerfed your model) + less private (your prompts are stored, no opt out) makes me much more interested in open models + local inference

1119027

Reposted by Rasmus Aagaard

Politiken @politiken.dk · 07/06/2026

Magtudredningen: De sociale medier udfordrer demokratiet - og politikerne er medskyldige

politiken.dk

Magtudredningen: De sociale medier udfordrer demokratiet - og politikerne er medskyldige

062

Reposted by Rasmus Aagaard

stellz @piss.beauty · 05/06/2026

getting your weights quantized probably feels amazing for the larger model

6984

Reposted by Rasmus Aagaard

Vicki @vickiboykis.com · 30/05/2026

Very strongly recommend breaking out of the token discourse and touching grass from time to time to maintain a grip on reality

What Is Art For?
BOWIE'S BOOKSHELF
THE HUNDRED BOOKS THAT CHANGED DAVID BOWIE'S LIFE
JOHN O'CONNELL
The Beauty of Everyday Things
Soetsu Yanagi
THE POCKET POETS SERIES
LUNCH POEMS
by
Frank O'Hara
SECOND EDITION
DICTIONARY OF SUBJECTS AND SYMBOLS IN ART
JAMES HALL Introduction by Kenneth Clark
ABER NINETEEN

7583

Reposted by Rasmus Aagaard

Vicki @vickiboykis.com · 11/05/2026

Legitimately feels like an unquantifiable vibe shift the last few weeks where the pendulum is swinging back to reasonable takes and people experimenting with model choice 🙏

Top HN headlines over the past 24 with reasonable takes like “Local AI needs to be the norm” and “I’m going back to writing code by hand”

51173

Reposted by Rasmus Aagaard

NobodyWho @nobodywhooo.bsky.social · 27/04/2026

Today we're announcing our #ReactNative bindings for NobodyWho! 🎉 Run #LLM models offline in your #react apps for #Android and #iOS. No internet. No API keys. No usage fees. Fully open-source. www.nobodywho.ai/posts/announ...

051

Reposted by Rasmus Aagaard

é. @urcad.es · 18/04/2026

We've recently implemented the fastest static embedding model in the world by an insanely large margin Read on for additional info ↓

3775

Rasmus Aagaard @rasgaard.com · 15/04/2026

Researchers identifies "Super Weights" which are at most a handful of weights (amongst billions!). Models end up generating complete gibberish if these weights are removed, underlining the brittleness of LLMs and importance and careful consideration of outlier features. arxiv.org/pdf/2411.07191

2172

Reposted by Rasmus Aagaard

Eugene Vinitsky 🍒 @eugenevinitsky.bsky.social · 15/03/2026

For a recent lab meeting, I wrote up a grab bag of ways to think about your development as a researcher during a PhD: emerge-lab.github.io/papers/an-un... Sharing in case folks find it useful or have feedback!

emerge-lab.github.io

610112

Reposted by Rasmus Aagaard

Sung Kim @sungkim.bsky.social · 17/03/2026

Flash-KMeans An IO-aware implementation of exact k-means that rethinks the algorithm around modern GPU bottlenecks. Flash-KMeans achieves 30x speedup over cuML and 200x speedup over FAISS. Paper: arxiv.org/abs/2603.09229 Code: github.com/svg-project/...

0296

Rasmus Aagaard @rasgaard.com · 26/02/2026

Claude Code Copenhagen meetup 🍻🚀

110

Reposted by Rasmus Aagaard

Alexander Doria @dorialexander.bsky.social · 23/02/2026

First blogpost of my cofounder post-Delhi on the deafening silence of public/political sphere as we come closer to another industrial revolution. anastasiastasenko.substack.com/p/the-countr...

37219

Rasmus Aagaard @rasgaard.com · 23/02/2026

Strongly agree! Sounds like a fantastic idea for a special course :) Maybe you can take inspiration from Danoliterate danoliterate.compute.dtu.dk which was a master's project at DTU

danoliterate.compute.dtu.dk

The Danoliterate Generative Large Language Model Benchmark

130

Rasmus Aagaard @rasgaard.com · 20/02/2026

Helt klart. Der er stadig tonsvis af quirks. Jeg er mest begejstret for tendensen der bliver vist. At software og hardware begge udvikler sig i retning af at lokal inference bliver konkurrencedygtigt ift. cloud.

000

Rasmus Aagaard @rasgaard.com · 20/02/2026

LLM cloud inference dominates usage, but should it? Local models and accelerators have improved massively over recent years. Perfect routing to best local model "reduce energy consumption by 80.4%, compute by 77.3%, and cost by 73.8% versus cloud-only deployment" arxiv.org/pdf/2511.07885

393

Reposted by Rasmus Aagaard

Cohere Labs @cohereforai.bsky.social · 17/02/2026

Introducing ✨Tiny Aya✨, a family of massively multilingual small language models built to run where people actually are. Tiny Aya delivers strong multilingual performance in 70+ global languages in a 3.35B parameter model, efficient enough to run locally, even on a phone.

29615

Rasmus Aagaard @rasgaard.com · 18/02/2026

Excited to be in fantastic company amongst the speakers for IDA Driving AI 2026 :) ida.dk/driving-ai/t...

150

Rasmus Aagaard @rasgaard.com · 17/02/2026

The paper suggests that poor reasoning abilities is one failure mode that comes from removing deeper layers. But yeah, wonder what else unexpectedly takes a hit

020

Rasmus Aagaard @rasgaard.com · 16/02/2026

Due to residual structures in Transformer-models it's possible that many layers contribute very little to the downstream performance of the network, allowing for removal those redundant layers with little impact. openreview.net/pdf/72eeca23...

0122

Rasmus Aagaard @rasgaard.com · 10/02/2026

Local AI inference through the browser is great. Transformers.js makes it feasible for someone like me who has little web dev experience :) Here's a simple transcription tool using Whisper-large-v3-turbo, making use of your local hardware and ensuring privacy. rasgaard.com/webai-stuff/...

031