"Yeah, my OpenClaw runs entirely on talkie-1930-13b-base, I connect to it only through telegram. No, not that telegram."
@emollick
-
Vintage LLM Runs on Device: Pre-1931 Siri Era
By
–
The new LLM trained only on pre-1931 text is small enough that it can potentially run on device, so, with the right tools, you can get a fully vintage version of Siri, but from the era of Downton Abbey. Here, I asked for it to arrange for sushi delivery in Philadelphia. Hmmm…
-
AI Model Knowledge Cutoff and Scientific Bias Analysis
By
–
This is an incredibly cool experiment
— Ethan Mollick (@emollick) 28 avril 2026
It is also fascinating that the model knows information up to 1931, but, at least in some science topics, seems very stuck in the early 1900s. For example, it defends the lumiferous aether hypothesis & has a distrust of special relativity https://t.co/xTVF7tkIw8 pic.twitter.com/UvcUFuJb90This is an incredibly cool experiment It is also fascinating that the model knows information up to 1931, but, at least in some science topics, seems very stuck in the early 1900s. For example, it defends the lumiferous aether hypothesis & has a distrust of special relativity
-
Incentive Misalignment: Quantity Over Quality in AI Systems
By
–
The problem is that the incentives push for "more" over "better" Paper: https://
pubsonline.informs.org/doi/full/10.12
87/orsc.2026.ed.v37.n3
… -
AI in Science: Quality Over Quantity in Research Systems
By
–
Very cool analysis of the submissions to a major management journal that shows how much the system of science, built for humans, is under strain as a result of AI. AI can be used to do better science or it can be used to just do more stuff. The danger is that "more" is winning
-
GPT-5.5 Creates Tabletop RPG Guide Through AI Playtesting
By
–
GPT-5.5 in Codex made a surprisingly solid table top RPG game masters guide & player guide, which it "playtested." It leans into the storytelling aspect, and still has some very LLM-y elements, but it is a novel setting. PDF: https://
drive.google.com/file/d/10QKnfj
JaWHxsTu4fo_dgMU6pAJXxuw3t/view?usp=sharing
…
More: https://
oneusefulthing.org/p/sign-of-the-
future-gpt-55
… -
Papers Must Better Justify Their Underlying Assumptions
By
–
I don’t think most papers I have seen think hard enough about these assumptions or are clear about why they pick particular ones. (Present company excluded)
-
AI Labor Substitution Models and S-Curve Adoption Patterns
By
–
I think this is an underlying assumption built into every model: when you’re assuming what types of labor AI could substitute for and in what time frame you are implying a certain S-curve shape.
-
AI capability and speed: the two fundamental questions shaping predictions
By
–
Every AI discussion ultimately rests on two questions: how good can AI get? And how fast? They are predictions about the s-curve shape. Everything else (job impact, potential risks, etc.) is downstream of those questions. I think it would be useful to focus on them more often.
-
AI Frontier Gaps: Where Human Help Remains Essential Today
By
–
The only way to fully appreciate the jaggedness of the AI frontier is up close. When you use it for a task you know well you find tons of tiny points where AI requires human help. Some are tedious (move a thing) & some profound (is this idea good)? But there are many, for now.