iiinova — where your digital self keeps evolving

Claas Voelcker @cvoelcker.bsky.social · 57m

At this point "I know how openreview and paperplaza work" should be the qualifications for instant hire as a faculty.

Reposted by Claas Voelcker

David Gleich @dgleich.bsky.social · 29/09/2026

I’m annoyed by all the academics stressing “quality over quantity” as a tonic for the LLM deluge of results / crap. People have been saying this for years. Either setup a compact with some language on putting this in hiring documents/recommendations/etc. or stop saying it.

2122

Claas Voelcker @cvoelcker.bsky.social · 29/09/2026

The major open question of our times remains: will we go down the Path of Dune, or the Path of 40k? Either way, expect more and exciting confrontations between the Catholics and AI

001

Claas Voelcker @cvoelcker.bsky.social · 29/09/2026

We always stop at the boring ones! Why not Guncle of AI, weird step-brother of AI, 2nd Cousin twice removed of AI, my great-aunt's lady friend who lived with her in the same room in the 50s of AI

140

Claas Voelcker @cvoelcker.bsky.social · 27/09/2026

Why do people submit a paper to be presented at multiple workshops, even after the paper has been presented before?

010

Claas Voelcker @cvoelcker.bsky.social · 26/09/2026

We are academics because we reject the bourgeois oppression of office clothing styles! (This is brought to you by the postdoc who gets called overdressed because he always wears a shirt and refuses to put on shorts)

030

Claas Voelcker @cvoelcker.bsky.social · 24/09/2026

Why is university teaching always structured badly? None of the faculty resources I have access to have any tips and lessons learned on actually running the classes! It's only about the bureaucracy!!!

040

Claas Voelcker @cvoelcker.bsky.social · 24/09/2026

Opus 5.5 is actually nice to work with ngl

000

Claas Voelcker @cvoelcker.bsky.social · 24/09/2026

I have finally been convinced to up my p(doom)

000

Claas Voelcker @cvoelcker.bsky.social · 24/09/2026

Survey question: What is the point of a scientific paper?! Please write down anything and everything that comes to mind

031

Claas Voelcker @cvoelcker.bsky.social · 23/09/2026

Is something actually wrong, or is it pent-up grief. The answer is always #2 www.youtube.com/watch?v=c-Dh...

youtube.com

Die Toten Hosen // Nur zu Besuch [Offizielles Musikvideo]

YouTube video by DIE TOTEN HOSEN

000

Claas Voelcker @cvoelcker.bsky.social · 23/09/2026

If you want my own opinion btw, I think we need to make it really really really hard to get a paper in. Require a 20 page submission, meticulous statistical analysis, theoretical results and empirical scaling. It should take a lab of very good people a year again to write a paper.

000

Claas Voelcker @cvoelcker.bsky.social · 23/09/2026

God, I just want to keep blogging about the dumb paper submission limit party over on X. None of these proposals come anywhere close to addressing the issue. Reputational costs: How on earth do you want to track reputation among 10,000s of authors? Furthermore, I know of several extremely prominent

240

Claas Voelcker @cvoelcker.bsky.social · 23/09/2026

Reducing the current flood of papers requires us to take a loooong and hard look at the incentive structure beneath it all. Limits etc feel good, but will probably barely move the needle based on authorship stats. The problem is a simple RoI calculation. AI papers are tickets to two of the most ...

2163

Reposted by Claas Voelcker

Blake Richards @tyrellturing.bsky.social · 22/09/2026

My hot take: We were *already* in a state of being overwhelmed by reviewing grants, scholarships, and papers prior to generative AI. Now it's not even close to tenable. The time to start rethinking the entire peer review model in #academia was 10 years ago. Hopefully this will force it!

36210

Reposted by Claas Voelcker

Kosta Derpanis @csprofkgd.bsky.social · 22/09/2026

#KostasVisualizations now up to 22 topics! csprofkgd.github.io/kostas-visua...

071

Claas Voelcker @cvoelcker.bsky.social · 22/09/2026

Major conferences should just desk reject papers from authors with h-index > 50. They have obviously published enough.

120

Claas Voelcker @cvoelcker.bsky.social · 22/09/2026

⏰ The deadline for our workshop workshop.oopsie-data.com on imperfect data at CoRL is nearing! ⏰ While you 🤖 are making progress on submissions, we came up with cool awards 🏆. Our workshop on imperfect data will award not only papers but also data submissions. The categories you can submit to 🥁...

workshop.oopsie-data.com

Oops, I Erred — Collecting, Curating and Using Imperfect Data for Realistic Scenarios

121

Claas Voelcker @cvoelcker.bsky.social · 22/09/2026

#claudetips I wrote a handoff skill that explicitly tells Claude to look over all the code it added and to reduce repetitions and redundancy. It burns a lot of tokens atm, but it genuinely helps with keeping the codebase form metastasizing.

100

Reposted by Claas Voelcker

Clément Canonne @ccanonne.github.io · 22/09/2026

"If OpenAI and Anthropic really want to help advance science, there is one obvious way they can do so: They can spend a tiny fraction of their vast fortunes to fund the kind of humble but essential research that’s increasingly starved for resources." (That they don't, and won't, says it all)

1285

Reposted by Claas Voelcker

Alex Holcombe @alexh.bsky.social · 21/09/2026

Consequences of the flood of AI manuscripts: "An early-career researcher at an unbranded institution loses the one mechanism that ever let them be read on merit — anonymous review of the work itself, however badly it functioned in practice.

410823

Claas Voelcker @cvoelcker.bsky.social · 18/09/2026

I think I vibe coded a paper in a week...

010

Claas Voelcker @cvoelcker.bsky.social · 18/09/2026

I am so freaking tired of reviewing slop…

030

Claas Voelcker @cvoelcker.bsky.social · 17/09/2026

"Why am I not getting cited?!" Well, when was the last time you read a paper that didn't come from a top 4 industry or academia lab?

062

Reposted by Claas Voelcker

Eugene Vinitsky 🍒 @eugenevinitsky.bsky.social · 17/09/2026

"Why does science frequently focus on hype rather than rigor", well, dunno, which articles are you reading and sharing

3252

Claas Voelcker @cvoelcker.bsky.social · 17/09/2026

This may be the funniest thing I have seen all week 💀 "Our process preserves all invariants ... except the budget" „The slide says we handle millions of events. It does not say we handle them well“ www.youtube.com/watch?v=FG8s...

youtube.com

Interview with Big Data engineer in 2026.

YouTube video by Kai Lentit

010

Claas Voelcker @cvoelcker.bsky.social · 16/09/2026

I don't think there is anything as unprofessional as the whole IEEE system. Software, templates, style guides, nothing makes sense...

030

Reposted by Claas Voelcker

Ethan Mollick @emollick.bsky.social · 13/09/2026

Existential AI risk is obviously critical, but it is not the only AI thing that requires policy. I worry it will become the sole focus of AI discussions. We don’t need better models for AI to have wide impacts on jobs & society and we need to be preparing to encourage good outcomes & mitigate bad.

312617

Claas Voelcker @cvoelcker.bsky.social · 10/09/2026

When academic ML researchers talk about not being able to compete with industry, I am always reminded of the fact that noone in academia builds a car either. Our job is not to build production ready models.

0110

Reposted by Claas Voelcker

Marcel Hussing @marcelhussing.bsky.social · 03/09/2026

🤖 Coming to CoRL 2026: The first workshop on imperfect robotics data! We’ll explore how to collect, curate & use failures 💢, OOD events❓, and unexpected human behavior 🤔 - data every lab produces but rarely documents, shares, or reuses. Check out workshop.oopsie-data.com

142

Claas Voelcker @cvoelcker.bsky.social · 01/09/2026

More and more I'm fascinated that the LLM literature will say things like "RL can only collapse the distribution" when the only RL approach they have ever used is reinforce, which yes, can do exactly only that unless you jump through hoops to improve exploration.

140

Claas Voelcker @cvoelcker.bsky.social · 01/09/2026

To quote one of my best friends after a philosophy lecture: I would love to explain to people how useless philosophy is, but for that we need to first find a shared Notion of “use“.

040

Claas Voelcker @cvoelcker.bsky.social · 19/08/2026

Potentially hot take: homework should never have been graded anyways. Normalize students learning and getting feedback without it going in the permanent record. If they don’t do it, their problem. This attitude is brought to you by the German academic system. No graded homework, works fine

090

Reposted by Claas Voelcker

Reinforcement Learning Conference @rl-conference.bsky.social · 19/08/2026

And here comes the final keynote of RLC26 from @sirbayes.bsky.social (Google DeepMind) with 'From RL to Automated Scientific Discovery' -- What if scientific discovery itself is a sequential decision-making problem? What a bold and inspiring vision to wrap up an incredible conference!

181

Reposted by Claas Voelcker

Reinforcement Learning Conference @rl-conference.bsky.social · 19/08/2026

It's been an unforgettable few days @RLC26! Huge thanks to speakers, panelists, workshop organizers, presenters & every attendee! From fascinating talks to lively posters, brainstorming to great food & ofcourse- Cirque du Soleil - rl-conference.cc/schedule.html Pls stay tuned for announcements!

091

Claas Voelcker @cvoelcker.bsky.social · 18/08/2026

I think the problem very often boils down to "Qui bono". I'm not opposed to AI models on training on art, I'm opposed to the artists not benefiting and the AI training companies accumulating civilizational levels of wealth.

190

Reposted by Claas Voelcker

fry69 @fry69.dev · 18/08/2026

I think that is a good point to recall this meme

1283

Reposted by Claas Voelcker

Alireza Modirshanechi @modirshanechi.bsky.social · 15/08/2026

Excited to share that I'll be starting as a junior professor and Emmy Noether awardee at the University of Göttingen this October! 🥳 🚀 I'll be hiring #phd students and #postdoc across #machinelearning, #neuroscience and #cogsci: modirlab.github.io/open-positio... Please spread the word! 🙏

Information about Göttingen and equal opportunities

1217059

Claas Voelcker @cvoelcker.bsky.social · 15/08/2026

The root cause for almost any issue in modern science is scarcity of rewards (mostly jobs). You cannot steer the demand and control the processes without addressing the supply.

150

Claas Voelcker @cvoelcker.bsky.social · 14/08/2026

Am I getting dumber or is Claude becoming more obtuse? I feel like it is basically at full conspiracy / Lumon levels of making up it's own language to describe it's experiment plans and findings...

010

Reposted by Claas Voelcker

Taylor W. Killian @twkillian.bsky.social · 13/08/2026

I am also eagerly recruiting graduate students at RLC 2026 for my new lab at @brighamyoungu.bsky.social. Take a look at twkillian.github.io/research for more information. Catch up with me on the conference slack, here, or via email to chat about the opportunities that are available.

twkillian.github.io

Research

Personal research webiste for Taylor W. Killian

031

Reposted by Claas Voelcker

Evangelos Kazakos @ekazakos.bsky.social · 14/08/2026

All this toxic pressure for publications pushes me away from research day by day. The point of research is **understanding** before anything else. The fact that everyone is so obsessed with innovation shows that there is a lot of narcissism in our field which makes want to throw up.

3111

Reposted by Claas Voelcker

Eugene Vinitsky 🍒 @eugenevinitsky.bsky.social · 12/08/2026

The newest Lea feature we've designed that I'm reallllly excited about: lea.ac/conferences/...

lea.ac

Lea - A calmer Bluesky for researchers

A curated garden for researchers on the AT Protocol

3295