iiinova — where your digital self keeps evolving

Reposted by Konrad Kording

Nicholdav @nicholdav.bsky.social · 04/10/2026

Writing something up, and reading this + re-reading Mensh & @kordinglab.bsky.social's: journals.plos.org/ploscompbiol... The new paper has steps to follow, helps my ADD brain. The older paper helps me keep my eyes on the prize. I feel like many researchers need these, like, now more than ever

journals.plos.org

Ten simple rules for structuring papers

1102

Konrad Kording @kordinglab.bsky.social · 29/09/2026

beautiful overview. The proliferation of "agency".

010

Konrad Kording @kordinglab.bsky.social · 28/09/2026

The alignment argument is orthogonal to this. We like thermostats because they produce a joint system that obeys simple rules (its temperature is roughly constant, barring major perturbations). An "agentic AI" does not positively have this property. E.g. on temperature loss it may overreact.

static.klipy.com

Sygourney Weaver: Nuke From Orbit

ALT: Sygourney Weaver: Nuke From Orbit

120

Konrad Kording @kordinglab.bsky.social · 28/09/2026

I can try. But really, like in many things philosophical, Stanford Encyclopedia is my starting point: plato.stanford.edu/entries/agen... Key concepts here are intentionality and causality.

plato.stanford.edu

Agency (Stanford Encyclopedia of Philosophy)

110

Konrad Kording @kordinglab.bsky.social · 28/09/2026

Why does it imply agentivity to you. Imitating someone does not make you an agent in my book.

110

Konrad Kording @kordinglab.bsky.social · 28/09/2026

Also for an llm to be a thermostat alignment first need to be solved

100

Konrad Kording @kordinglab.bsky.social · 28/09/2026

the wording agency does have a tinge of that in English.

110

Konrad Kording @kordinglab.bsky.social · 28/09/2026

Yes. RnN declare a thermostat an agent. Everything is. Normal english minimally demands goals/desires and the willingness to work towards them. And usually a lot more.

110

Konrad Kording @kordinglab.bsky.social · 27/09/2026

But we did iterate to make the answers better. Which now *is* agentcy ;)

030

Konrad Kording @kordinglab.bsky.social · 27/09/2026

Why on earth did we, as a field, allow the word "agent" to be used the way it currently is?

4271

Reposted by Konrad Kording

Blake Richards @tyrellturing.bsky.social · 23/09/2026

I’m beyond honoured to be named a 2026 @schmidtsciences.bsky.social Polymath. With US$2.5 million over five years, I’ll be able to pursue an interdisciplinary project to combine quantitative and qualitative data to better capture human neurodiversity. 🧠📈 www.schmidtsciences.org/2026-polymat...

schmidtsciences.org

101258

Konrad Kording @kordinglab.bsky.social · 16/09/2026

I did advocate for it. But I was very thankful that people like Ed and George did the outreach for this.

010

Konrad Kording @kordinglab.bsky.social · 16/09/2026

Wait. I don't think I was on that one.

110

Reposted by Konrad Kording

Patrick Mineault @patrickmineault.bsky.social · 14/09/2026

Are virtual flies playing Beat Saber? SM64? Doom? No. The new fly connectome is an incredible achievement! Less is going on in the sim videos than it seems. I'm stoked to see people engage with neuroscience though. My breakdown: www.neuroai.science/p/are-flies-...

neuroai.science

Are flies playing Beat Saber?

What a time to be alive!

27021

Konrad Kording @kordinglab.bsky.social · 10/09/2026

Really just me seeing lots of bad science. And wanting to help students do better.

110

Konrad Kording @kordinglab.bsky.social · 09/09/2026

Author of this webpage here. What are your thoughts?

100

Konrad Kording @kordinglab.bsky.social · 08/09/2026

why? Time makes every behavior infinite dimensional no?

110

Konrad Kording @kordinglab.bsky.social · 07/09/2026

I wrote a super dynamic AI/LLM explainer webpage. Not public yet (don't yet know if it would bankrupt me and having some other constraints). Anyone willing to give it a try and give me some feedback? Mostly looking for feedback from the non-experts. Chat me if interested.

3145

Konrad Kording @kordinglab.bsky.social · 07/09/2026

Cool cool. Let’s not obsess about dimensionality. Good motto in a 1/f world!

1304

Konrad Kording @kordinglab.bsky.social · 04/09/2026

This is cool. But why don't people know how to pour coffee? www.biorxiv.org/content/10.6... the paper asks how to pour a coffee and the answer is "not like this!"

2100

Konrad Kording @kordinglab.bsky.social · 28/08/2026

Full page figure for explainer in popular journal.

000

Konrad Kording @kordinglab.bsky.social · 28/08/2026

Awesome! gd is tagged as technical term and I think is important for people who want to learn deeper about this. What would be the first to remove and simplify? I know it’s too complicated but it’s a bit of a tradeoff. Also, what can I read from you?

100

Konrad Kording @kordinglab.bsky.social · 28/08/2026

I agree that is all subtle. But would you agree that there are still massive differences between a human agent and a 2026 ai „agent“?

100

Reposted by Konrad Kording

Peter Tennant @pwgtennant.bsky.social · 27/08/2026

Does the new shingles vaccine (Shingrix) protect against cardiovascular disease (CVD)? Yes, according to a new 'natural experiment' study in @natmed.nature.com. But the truth is sadly far less convincing! Pull up a chair and let's examine another peer review failure in a Nature journal. 1/10 🧵

817574

Reposted by Konrad Kording

Community for Rigor @comm4rigor.bsky.social · 27/08/2026

🎉 We are excited to welcome Katie Corker as a featured speaker at C4R26! 💬 "Strengthening Rigor Through Open Research and Evaluation" 📅 September 29, 11:30 am ET 📍 Philadelphia, PA 🎟️ Free with RSVP 🔗 link in bio #Preprints #OpenScience #Science

152

Konrad Kording @kordinglab.bsky.social · 26/08/2026

Would love to see what you come up with!

010

Konrad Kording @kordinglab.bsky.social · 26/08/2026

I was pretty sure noone would see it!

110

Konrad Kording @kordinglab.bsky.social · 26/08/2026

Inner speech is really outside of the yellow box. Its about allowing tokens that the user does not see. While the yellow box just makes tokens.

000

Konrad Kording @kordinglab.bsky.social · 26/08/2026

You would (somewhat) open the yellow box? — that’s where much research is but it’s in a way subordinate to what happens outside the box.

100

Konrad Kording @kordinglab.bsky.social · 26/08/2026

That boat has sailed. Sorry. It would have been better but it won’t happen. People want to claim these worlds. They want to believe them!

200

Konrad Kording @kordinglab.bsky.social · 26/08/2026

There is a reason why we need purple!

100

Konrad Kording @kordinglab.bsky.social · 26/08/2026

In high dimensions there are saddles but “no” local minima.

100

Konrad Kording @kordinglab.bsky.social · 25/08/2026

And there you have it. The Old Recipe got extended with ‘inner speech’ + tools + agents + better evaluation into today's interactively ‘thinking’ machines. A great deal of the strengths and weaknesses of today's LLMs can be understood from these component ideas.

170

Konrad Kording @kordinglab.bsky.social · 25/08/2026

And all of this, is better with better “training“ goals. Wherever truth is defined, push towards this truth (D). And where it is not defined, use LLMs to check answers ("self evaluation", (E))

180

Konrad Kording @kordinglab.bsky.social · 25/08/2026

A third idea (C) for making answers, combine many components into "Agents", : coordinate component large language models (LLMs), memory, databases, and tools

260

Konrad Kording @kordinglab.bsky.social · 25/08/2026

A second idea (B) for making answers, tool use. Call programs like databases, internet search, calculators etc.

160

Konrad Kording @kordinglab.bsky.social · 25/08/2026

A first idea (A) for making answers, inner speech. Produce words that are just like a scratchpad which the user never sees.

160

Konrad Kording @kordinglab.bsky.social · 25/08/2026

The result of this are what is sometimes called large-scale "stochastic parrots". Very impressive text but often somewhat superficial. But these are the old LLMs (think 2022). Innovation was super fast since.

1101

Konrad Kording @kordinglab.bsky.social · 25/08/2026

The thing that makes Neural networks so powerful is that there is a direct way (4) of improving them with examples and towards any definition of answer quality (training).

170

Konrad Kording @kordinglab.bsky.social · 25/08/2026

The real work is done by a core machine, the neural network which is used to generate one word at a time based on its inputs. The probability of each word depends on prior words. And words are chosen one by one according to those probabilities.

270

Konrad Kording @kordinglab.bsky.social · 25/08/2026

All that LLMs do is convert prompt words into answer words (1). And the old recipe does that with two goals, language imitation (2) and people's rating of answer quality (3). Answers are produced word by word using a neural network.

1100

Konrad Kording @kordinglab.bsky.social · 25/08/2026

We will use everyday language where possible and mark technical terms in purple. Key is that modern LLMs have a component of making answers (in blue, called "inference") and a component of improving answers (red, called "training").

1120

Konrad Kording @kordinglab.bsky.social · 25/08/2026

It will also be part of an upcoming manuscript. Optimized with Brett Mensh of optimizescience.com -- and work in progress. Feel free to give feedback.

optimizescience.com

Optimize Science – Science Consulting and Writing Support

1130

Konrad Kording @kordinglab.bsky.social · 25/08/2026

Thread. I have yet to see a proper explanation of how LLMs actually work that is targeted to an informed general audience. So here is an explainer that I put together along with Tim @countzerozzz (inspired by @CIFAR_News LMB discussions).

1723160