Konrad kording, @Penn Prof, deep learning, brains, #causality, rigor, https://xitter.cf/t.co/tTJW05RRfa, https://xitter.cf/t.co/qf7ZHxjaK1, Transdisciplinary optimist, Dad, Loves outdoors, 🦖

Philadelphia, PA
Joined November 2012
Intelligence saturation. Everyone called us crazy ;)
OpenAI, Anthropic and other leading AI labs face a growing challenge: Better models aren't necessarily attracting more usage. A 10-point intelligence advantage now earns just 20% more developer traffic, down from nearly 69% in late 2025. Full story: thein.fo/4hGJ9jY
3
1
70
5,663
Why We Systems Neuroscientists Should (Usually) Be Quietists: The Philosophical Logic of NeuroAI -- response to David Barack kording.substack.com/p/why-w…
5
13
1
75
5,288
Kording Lab 🦖 retweeted
I created a mock-up for a preprint server called the 'Test of time' which serves as an evidence-record for each article in perpetuity. test-of-time-preprint.netlif… Given that science is speeding up, we need a place to track the strength of collective scientific evidence.
I'm actually saying that I think in the future the two aren't necessarily distinct processes. But there's no need to suggest that #2 should delay #1, as if these are entirely different dialogues. I understand your concern about not conflating technical review with PR/random hot takes. And that's a thing that I think we should be thinking about how to solve. I don't think that gating through journals is a solution that has in the past adequately addressed this. Nor do I think it will stand the test of time given the pace of modern technology and the fact that science is increasingly moving outside the walls of Academia
1
1
1
1,082
LLMs used to often be wrong. Now they are often bullshit. I preferred them when they were wrong. They literally use language where they can not be wrong. They start sounding like my most insufferable colleagues!
7
6
1
108
9,131
Me: What’s 2+2? Old LLM: 5. New LLM: The move isn’t to fixate on a number. It’s to ask what the number is doing. The real payoff comes from holding both twos in productive tension.
2
3
35
1,970
Kording Lab 🦖 retweeted
Grant Sanderson (@3blue1brown) is the most influential mathematician of the 21st century.
17
33
4
585
33,438
What about we don't put up laws indemnifying the AI companies and instead accelerate the mechanisms to sue them? Because it sure seems like this is the single most important "alignment" action, far more useful than any alignment work.
5
10
40
2,990
Kording Lab 🦖 retweeted
OpenAI’s faster Fourier transform is less a mathematical breakthrough and more a malicious exploit of mathematical traditions. argmin.net/p/faster-fourier-…
67
118
56
1,154
365,117
Kording Lab 🦖 retweeted
Spicy take incoming. 🔥🔥 The combined bomber offensive in WW2 finally crippled the Axis powers by destroying their fuel supply chain. Then it took the industrial might of the entire US to build tens of thousands of bombers. Now anyone between you and the Middle East can burn down a refinery, tank farm, pipeline, or tanker, with an easily assembled robotic flying bomb that costs less than my AI subscriptions. @PeterZeihan predicted "the tanker war" would happen in the Indian Ocean but it doesn't matter where. There's only one country that could plausibly enforce global peace on shipping lanes and only one industrial power that is self sufficient on oil supply, and it's the same country. Trump asks Iran and Zelensky to please stop blowing shit up for five minutes, but the primary victims of this phase of the kinetic conflict are China and Europe, both places that have spent decade(s) ignoring and/or undermining the Pax Americana and, in Europe's case, going out of their way to destroy their own economies. China wants to be a big boy on the global stage? Let them spend their own blood and treasure pacifying the Middle East. Europe wants to stick it to Trump? It's easier to tell the yanks to get fucked if you don't rely on them for literally every aspect of your industrial infrastructure and economic stability. I've been in rooms with senior energy ministers from Europe and China. The Chinese are rational and pragmatic, the Europeans are functionally comatose, but neither party has a better plan than getting into a bidding war to buy whatever LNG the US can spare, forever. It is blindingly obvious to me that the least unstable future equilibrium is one where oil users localize supply using solar synthetic fuel tech, essentially a silico-industrial form of farm-to-table agriculture. Short supply chains, locally secured. It is equally obvious to me that a generation of people who grew up in the "developed" world are going to endure a decade or more of severe hardship before they, collectively, accept the L and move to the next step. Most people would rather watch their neighbor freeze than admit they were wrong.
If you can't defend your entire oil supply chain from anyone who could possibly have access to a $2000 drone and could possibly have any reason to use it with zero consequences, you are one third party misunderstanding away from abruptly re-entering the pre-industrial age.
70
131
21
1,931
166,487
Most research problems I see start before the research: the question isn't sharp yet, and everything downstream is thus murky. So I built Question First, a free tool that gets your question and logic clear before you start. questionfirst.org 🧵
2
15
100
6,920
It's free. Try it on the project you're stuck on right now: questionfirst.org Teaching a methods course or onboarding new students this fall? Send it to them. That's who I built it for. 🙏
1
1
525
If you used this under its old name, Plan Your Science: same login, same projects, nothing to do. Old links still work.
1
1
480
Longterm high quality performance of a BCI. Not glamorous. But essential. Great to read this!
Performance or longevity. For years, it seemed like the field of BCI would have to choose between the two. Our accumulating preclinical data suggests otherwise: four years of stable neural recording and decoding in large animal models, with steady spike SNR and no electrode array migration. Read the full announcement here: paradromics.com/news/paradro…
6
10
2,498
Kording Lab 🦖 retweeted
Check out this loser! >uploads a draft to Chat on Sept. 8th >gets scooped by Chat on Oct. 6th "The result released by OpenAI (#311) on October 6, 2026... identical to a preliminary version of the present paper that was uploaded to ChatGPT on September 8, 2026" LMAO get rekt
66
266
93
3,460
358,279