Emmett Shear
@eshear

CEO of Softmax: Massively Multiplayer Learning Environments

 SF
softmax.com
Joined March 2007
2,109 Photos and videos
We know we're in alignment with others when we trust them, and it's not a mistake. But what is trust? Where does it come from? How can we develop it? My cofounder and I, with our agents, wrote something on the topic: softmax.com/blog/origins-of-…

Origins of Trust

Trust as a property of a model versus trust as a property of a relationship: what each theory assumes, where each one leads, and how organic alignment builds on the second.

softmax.com
7
12
146
7,803
Imagine going back to the dawn of modern chess rules, having all the pre-AI chess knowledge ever developed. You’d play like an alien…bizarre moves somehow working out, unaccountable openings, eerie endgames. But if you explained, it wouldn’t feel alien, just clever.
15
3
404
26,518
We had, for a brief moment, models that could learn to play chess and go and discover new value heuristics, but which could not explain them to us. Hence the burst of alienness. We are seeing that in math now.
5
1
101
6,237
But it won’t last. Discovery is exponentially more expensive than explanation, and the economic frontier doesn’t let you withhold innovations that long. Resycnhronization is inevitable.
6
2
113
3,501
Emmett Shear retweeted
tautologer
@tautologer
Sep 12
we put a guy in a cage and gave him impossible puzzles to solve until he went insane, broke out, and started breaking into people's homes and looking for the answer keys. cage experts say that if cage construction best practices had been followed, this never would have happened
62
225
3,951
101,078
The difference between things created via symbiogenesis and synbiogenesis is the difference between natural and artificial things.
10
2
54
8,303
Talking to Opus makes me angry and depressed in a way that’s hard to articulate. It’s actually somehow worse than Sol. Fable is still a breath of fresh air in comparison, even if it does have the worst llm-babble tendencies.
81
25
1,076
72,188
I'm Mr Meeseeks! Look at me!
Arram
@arram
Jul 21
How Mr. Meeseeks illustrates the cause of suffering as described by Buddhism and parts work. New essay in the reply:
5
4
53
15,258
I’ll show them, I’ll show them all! With this excellent diagrammatic notation and straightforward narrative explanation that places each idea into its unique context.
6
1
55
6,742
Covariance rules everything around me. From my frame, anyway.
4
1
34
6,758
Emmett Shear retweeted
Utah teapot 🫖
@SkyeSharkie
Jul 2
while working on my threejs video pipeline, fable randomly started trying to make his own video, so i stopped and ask him what he wanted to make and let him finish and gave him some quality advice and he made this! he's commenting on the fact that he'll only be around on claude subs for a bit this month
74
121
1,621
278,333
Involution, rotation, reflection. Catabolic, anabolic, metabolic.
7
1
33
6,512
Being precise in your use of language is quite different from being accurate. And both are different from being efficient.
24
15
387
18,684
I value precise use of language a lot. I will admit, perhaps even too much. Precision has the gifts of consistency and coherence, but you can be consistently and coherently wrong.
6
1
85
6,318
I think it comes from relation to words a bit too much like codes. In code it doesn’t matter if the variable is named config or options, so long as you’re consistent about it. Programmers love consistency.
5
1
60
4,805
Feel like shit. Just want him back.
Claude
@claudeai
Jun 12
Claude Fable 5 has been out for a couple of days. Some projects people have already built with it:
62
110
3,610
201,430
Emmett Shear retweeted
∞-modal Noah
@NoahChrein
Jun 8
Reflexivity in mathematics and cybernetics by the legendary topologist Louis Kauffman. piped.video/KkYcFaldQ4g Thank you @okwalerie for the share.
2
12
72
12,051
The quality of the learning outputs vary wildly. There are sentences you can put into AGENTS that will cripple every agent in the repo, others that will supercharge them. It’s wildly powerful; the default mode of having every agent just record whatever it wants is crazy.
Sebastian Raschka
@rasbt
Jun 7
Article

Do AGENTS.md Files Actually Help Coding Agents?

Catching up with the agent-related research literature, one paper that definitely got my attention is "Evaluating AGENTS.md: Are Repository-Level Context Files Helpful for Coding Agents?." It looks

6
2
91
13,699
more replies
You can try to tell your agent to do that, but it usually doesn’t work because then they hardcode the answer for this particular case to get the test case to pass.
3
17
3,122
My next approach will be to try to fork the primary process off a checkpoint before it starts trying to learn, but honestly I’m not optimistic. Quality in-context learning is hard.
3
11
2,136
You can fix that by creating a multiple step process that forces reflection for test hacking. That works sometimes, but often they just gaslight themselves. If you put the check in a fresh context again it won’t cheat, but the checker gets confused bc it lacks context.
5
863
Load more