David Manning
dlmanning.bsky.social
David Manning
@dlmanning.bsky.social
No one of consequence
Or people don’t understand the distinction between model undergoing a bunch of RL training for specific tasks and fundamental model capability.
November 28, 2025 at 6:36 PM
I don’t know… if it were as easy to tell that someone was publishing work stolen from uncredited grad students I’d be okay dismissing anything they said out-of-hand simply as a means of disincentivizing such behavior.
November 28, 2025 at 6:32 PM
That is an 89-page press release masquerading as an academic publication.

When you look at the individual examples it’s all some version of “the text predictor phrased a thing in a way that gave a human an idea.”
November 27, 2025 at 5:54 PM
The more you know about Thanksgiving’s location the less you know about its velocity.
November 27, 2025 at 1:41 PM
Does this mean upcoming models will have more cargo space?
November 26, 2025 at 7:40 PM
My TV is trying to push an update with a chatbot in it.
November 26, 2025 at 1:36 PM
Truth
November 25, 2025 at 1:43 PM
Philosophers in the 19th century: “Infinite quantitative change does not produce a qualitative distinction.”

Tech bros in the 21st century: “3% improved on SWE Bench! All Hail the Machine God! My life for the Omnisiah!”
November 25, 2025 at 1:34 PM
Your mom seems nice.
November 25, 2025 at 12:29 PM
Yes
November 23, 2025 at 11:59 PM
To me, storage implies things like durability. Probabilistic interpolated output that sometimes resembles training data is a different thing. E.g. Regression models neither store nor retrieve the data used to build them.

I understood Paul to be differentiating between storage and “lossy database.”
November 23, 2025 at 2:44 PM
For my part I only meant that LLM’s are not properly understood as search engines.
November 23, 2025 at 2:34 PM
I think you’re meaning to respond to Paul here.
November 23, 2025 at 2:04 PM
Warum sieht die Switch so riesig aus?
November 23, 2025 at 1:51 PM
A close second is the idea that they are “search engines.”
November 23, 2025 at 1:49 PM
Is that the Curb Your Enthusiasm guy?
November 18, 2025 at 9:05 PM
Has anyone done a startup that uses AI to predict how much funding stupid ideas can get in a seed round?
November 18, 2025 at 4:19 AM
Yeah I mean that’s absolutely true.
November 16, 2025 at 12:37 PM
The most baffling thing to me about the rush to embrace “AI” is how corporate IT departments have seemingly jettisoned four decades of cultivated paranoia in the span of a couple years.
November 16, 2025 at 12:16 PM
*yawn*
November 12, 2025 at 7:50 PM
Yep. Since they have no actual path to profitability, the game is to compete with one another over-leveraging themselves to see who can win a government bailout by becoming a threat to the national (and global) economy.

Evolution of “get acquired by Google” as a business plan.
November 11, 2025 at 5:25 PM