Steve Steiner
@stevejsteiner.bsky.social
Love learning new stuff. Exploring emerging AI dev practice and tools. Enthusiast of 4E Cog Sci, Story structure, process ontology.
Pinned
GitHub - KnowSeams/KnowSeams: Fast, accurate sentence detection for English narrative with proper dialog handling.
Fast, accurate sentence detection for English narrative with proper dialog handling. - KnowSeams/KnowSeams
github.com
I want to analyze novels, and pysbd didn’t split dialog the way I want it split, so I wrote my own splitter.
I kind of rabbit holed on optimizing this until I could get better than 1000 Mb/sec including the disk reads and writes.
I kind of rabbit holed on optimizing this until I could get better than 1000 Mb/sec including the disk reads and writes.
“Once you become obsessed with the problem of process, you either become a complex systems nut or a cybernetics nut, and no one listens to either.”
www.argmin.net/p/staging-in...
www.argmin.net/p/staging-in...
Staging Interventions
Actions are fundamentally different than predictions, but it's hard to write this distinction in math.
www.argmin.net
November 9, 2025 at 4:10 PM
“Once you become obsessed with the problem of process, you either become a complex systems nut or a cybernetics nut, and no one listens to either.”
www.argmin.net/p/staging-in...
www.argmin.net/p/staging-in...
Next bookclub book.
November 6, 2025 at 2:44 AM
Next bookclub book.
Reposted by Steve Steiner
What happens when we model the detective archetype at scale? 🕵️♂️📚
Our new paper, accepted for #CHR2025 combines literary history and computational modeling to trace how the figure of the detective evolves across 150 years of French fiction.
arxiv.org/pdf/2511.00627
Our new paper, accepted for #CHR2025 combines literary history and computational modeling to trace how the figure of the detective evolves across 150 years of French fiction.
arxiv.org/pdf/2511.00627
November 4, 2025 at 5:36 PM
What happens when we model the detective archetype at scale? 🕵️♂️📚
Our new paper, accepted for #CHR2025 combines literary history and computational modeling to trace how the figure of the detective evolves across 150 years of French fiction.
arxiv.org/pdf/2511.00627
Our new paper, accepted for #CHR2025 combines literary history and computational modeling to trace how the figure of the detective evolves across 150 years of French fiction.
arxiv.org/pdf/2511.00627
I had Chatgpt analyze our favorite tv shows and suggest an ideal series.
With The Diplomat, Patriot, Lupin, Tokyo Vice, Killing Eve, Minx, you’re consistently drawn to:
“people with skills navigating systems — stylishly, emotionally, and intelligently.”
Now Watching:
With The Diplomat, Patriot, Lupin, Tokyo Vice, Killing Eve, Minx, you’re consistently drawn to:
“people with skills navigating systems — stylishly, emotionally, and intelligently.”
Now Watching:
The Bureau (TV series) - Wikipedia
en.wikipedia.org
November 3, 2025 at 5:03 AM
I had Chatgpt analyze our favorite tv shows and suggest an ideal series.
With The Diplomat, Patriot, Lupin, Tokyo Vice, Killing Eve, Minx, you’re consistently drawn to:
“people with skills navigating systems — stylishly, emotionally, and intelligently.”
Now Watching:
With The Diplomat, Patriot, Lupin, Tokyo Vice, Killing Eve, Minx, you’re consistently drawn to:
“people with skills navigating systems — stylishly, emotionally, and intelligently.”
Now Watching:
Claude just loves creating layer violations.
Fortunately it’s not allowed to automatically change .proj files so I can just say don’t make a layer violation and it fixes it.
Fortunately it’s not allowed to automatically change .proj files so I can just say don’t make a layer violation and it fixes it.
November 1, 2025 at 11:38 PM
Claude just loves creating layer violations.
Fortunately it’s not allowed to automatically change .proj files so I can just say don’t make a layer violation and it fixes it.
Fortunately it’s not allowed to automatically change .proj files so I can just say don’t make a layer violation and it fixes it.
Interesting: dsyme.net/2025/09/24/o...
Intent, meet Toolchain
[ notes made for a panel at AI Native Dev podcast ] LLMs and Coding Agents affect all aspects of software engineering – documentation, specifications, tasks, intent, summaries, code generatio…
dsyme.net
October 28, 2025 at 1:11 AM
Interesting: dsyme.net/2025/09/24/o...
Llms poor story telling seems to me at least complicated by the simple bias toward non-suspensive sentences.
October 27, 2025 at 2:04 PM
Llms poor story telling seems to me at least complicated by the simple bias toward non-suspensive sentences.
I gave Claude Code some ‘feedback’.
Claude Code: “Yes that’s ass backwards”
Claude Code: “Yes that’s ass backwards”
October 27, 2025 at 12:48 AM
I gave Claude Code some ‘feedback’.
Claude Code: “Yes that’s ass backwards”
Claude Code: “Yes that’s ass backwards”
Given we’ve tried and have given up multiple times, Chatgpt just suggested Slow Horse again, but added we should skip the first episode.
October 26, 2025 at 9:16 PM
Given we’ve tried and have given up multiple times, Chatgpt just suggested Slow Horse again, but added we should skip the first episode.
The Adjacent Possible: How Everything Evolves
Foundations #1
open.substack.com
October 26, 2025 at 3:54 AM
Reduction in overused phrases — arxiv.org/abs/2510.15061
Antislop: A Comprehensive Framework for Identifying and Eliminating Repetitive Patterns in Language Models
Widespread LLM adoption has introduced characteristic repetitive phraseology, termed "slop," which degrades output quality and makes AI-generated text immediately recognizable. We present Antislop, a ...
arxiv.org
October 24, 2025 at 4:18 AM
Reduction in overused phrases — arxiv.org/abs/2510.15061
Claude sonnet 4.5 in GH Copilot just dropped a FUCK YES when it solved the problem. Hmm ICL?
For no reason at all: arstechnica.com/science/2023...
For no reason at all: arstechnica.com/science/2023...
Do better coders swear more, or does C just do that to good programmers?
For open source C code, curses mean quality, a recent bachelor’s thesis suggests.
arstechnica.com
October 22, 2025 at 4:24 AM
Claude sonnet 4.5 in GH Copilot just dropped a FUCK YES when it solved the problem. Hmm ICL?
For no reason at all: arstechnica.com/science/2023...
For no reason at all: arstechnica.com/science/2023...
Reposted by Steve Steiner
𝑵𝒆𝒘 𝒃𝒍𝒐𝒈𝒑𝒐𝒔𝒕! A rundown of some cool papers I got to chat about at #COLM2025 and some scattered thoughts
saxon.me/blog/2025/co...
saxon.me/blog/2025/co...
COLM 2025: 9 cool papers and some thoughts
Reflections on the 2025 COLM conference, and a discussion of 9 cool COLM papers on benchmarking and eval, personas, and improving models for better long-context performance and consistency.
saxon.me
October 17, 2025 at 5:24 AM
𝑵𝒆𝒘 𝒃𝒍𝒐𝒈𝒑𝒐𝒔𝒕! A rundown of some cool papers I got to chat about at #COLM2025 and some scattered thoughts
saxon.me/blog/2025/co...
saxon.me/blog/2025/co...
I hear ‘grow an AI’ being used because training it is not ‘engineering an AI.’
This is a bad metaphor in a different direction.
‘Brew an AI’ vs ‘Bake an AI’ provides a more informative pair of metaphors about what’s happening.
This is a bad metaphor in a different direction.
‘Brew an AI’ vs ‘Bake an AI’ provides a more informative pair of metaphors about what’s happening.
October 17, 2025 at 2:48 PM
I hear ‘grow an AI’ being used because training it is not ‘engineering an AI.’
This is a bad metaphor in a different direction.
‘Brew an AI’ vs ‘Bake an AI’ provides a more informative pair of metaphors about what’s happening.
This is a bad metaphor in a different direction.
‘Brew an AI’ vs ‘Bake an AI’ provides a more informative pair of metaphors about what’s happening.
Claude —
Ha! No, “truthical” is not a word - you’ve caught me in an inconsistency! I was being overly prescriptive about “dynamical” vs “dynamic”:
“Dynamical systems” is indeed the standard term in mathematics
But it’s not because there’s some ironclad rule that “-al” makes things more technical
Ha! No, “truthical” is not a word - you’ve caught me in an inconsistency! I was being overly prescriptive about “dynamical” vs “dynamic”:
“Dynamical systems” is indeed the standard term in mathematics
But it’s not because there’s some ironclad rule that “-al” makes things more technical
October 13, 2025 at 7:50 PM
Claude —
Ha! No, “truthical” is not a word - you’ve caught me in an inconsistency! I was being overly prescriptive about “dynamical” vs “dynamic”:
“Dynamical systems” is indeed the standard term in mathematics
But it’s not because there’s some ironclad rule that “-al” makes things more technical
Ha! No, “truthical” is not a word - you’ve caught me in an inconsistency! I was being overly prescriptive about “dynamical” vs “dynamic”:
“Dynamical systems” is indeed the standard term in mathematics
But it’s not because there’s some ironclad rule that “-al” makes things more technical
Sonnet 4.5's suggestion on fixing the error it made splitting one large file into 10 smaller files.
"let me propose a more efficient approach: git reset HEAD"
"let me propose a more efficient approach: git reset HEAD"
October 12, 2025 at 11:32 PM
Sonnet 4.5's suggestion on fixing the error it made splitting one large file into 10 smaller files.
"let me propose a more efficient approach: git reset HEAD"
"let me propose a more efficient approach: git reset HEAD"
This is neat:
github.com/oOo0oOo/lean...
github.com/oOo0oOo/lean...
GitHub - oOo0oOo/lean-lsp-mcp: Lean Theorem Prover MCP
Lean Theorem Prover MCP. Contribute to oOo0oOo/lean-lsp-mcp development by creating an account on GitHub.
github.com
October 12, 2025 at 10:20 PM
This is neat:
github.com/oOo0oOo/lean...
github.com/oOo0oOo/lean...
At some point he needs to say "Smash that subscribe button"
youtu.be/K5w7VS2sxD0?...
youtu.be/K5w7VS2sxD0?...
Kevin Buzzard - Where is Mathematics Going? (September 24, 2025)
YouTube video by Simons Foundation
youtu.be
October 12, 2025 at 9:48 PM
At some point he needs to say "Smash that subscribe button"
youtu.be/K5w7VS2sxD0?...
youtu.be/K5w7VS2sxD0?...
I was curious …
These findings establish robust evidence that viable microbes (or at least their DNA / reproductive capacity) can exist in upper-atmospheric aerosols. A domain we might call the “aerobiome.”
It’s an open question if these microbes merely survive or actively live at altitude.
These findings establish robust evidence that viable microbes (or at least their DNA / reproductive capacity) can exist in upper-atmospheric aerosols. A domain we might call the “aerobiome.”
It’s an open question if these microbes merely survive or actively live at altitude.
October 10, 2025 at 11:13 PM
I was curious …
These findings establish robust evidence that viable microbes (or at least their DNA / reproductive capacity) can exist in upper-atmospheric aerosols. A domain we might call the “aerobiome.”
It’s an open question if these microbes merely survive or actively live at altitude.
These findings establish robust evidence that viable microbes (or at least their DNA / reproductive capacity) can exist in upper-atmospheric aerosols. A domain we might call the “aerobiome.”
It’s an open question if these microbes merely survive or actively live at altitude.
Not good. We’ve swapped to using the AeroPress while waiting for a replacement.
October 9, 2025 at 4:28 PM
Not good. We’ve swapped to using the AeroPress while waiting for a replacement.
Things going well, then "I'll make 10K tests."
Claude my friend - What happened to the thing we were supposed to be doing?
"Root Cause: Vocabulary Shift
Phase 1 uses proof vocabulary
Phase 2.3 uses testing vocabulary
The WU1.8 task document diverged from it during the Phase 2 detailed planning."
Claude my friend - What happened to the thing we were supposed to be doing?
"Root Cause: Vocabulary Shift
Phase 1 uses proof vocabulary
Phase 2.3 uses testing vocabulary
The WU1.8 task document diverged from it during the Phase 2 detailed planning."
October 9, 2025 at 3:26 AM
Things going well, then "I'll make 10K tests."
Claude my friend - What happened to the thing we were supposed to be doing?
"Root Cause: Vocabulary Shift
Phase 1 uses proof vocabulary
Phase 2.3 uses testing vocabulary
The WU1.8 task document diverged from it during the Phase 2 detailed planning."
Claude my friend - What happened to the thing we were supposed to be doing?
"Root Cause: Vocabulary Shift
Phase 1 uses proof vocabulary
Phase 2.3 uses testing vocabulary
The WU1.8 task document diverged from it during the Phase 2 detailed planning."
mixed_script_confusables?
Made some Lean 4 proofs before asking Claude Code to build the implementation in rust.
The rust toolchain was not having this Greek symbol in source code malarky -
let θ = test_thresholds();
Made some Lean 4 proofs before asking Claude Code to build the implementation in rust.
The rust toolchain was not having this Greek symbol in source code malarky -
let θ = test_thresholds();
October 9, 2025 at 1:13 AM
mixed_script_confusables?
Made some Lean 4 proofs before asking Claude Code to build the implementation in rust.
The rust toolchain was not having this Greek symbol in source code malarky -
let θ = test_thresholds();
Made some Lean 4 proofs before asking Claude Code to build the implementation in rust.
The rust toolchain was not having this Greek symbol in source code malarky -
let θ = test_thresholds();
I’m a fan of Joe Edelman. His Attentional Policies for eliciting a persons values was interesting before llms, and is incredibly more interesting now.
This new paper is a collaboration. The norms model in it also sounds interesting.
www.full-stack-alignment.ai/paper
This new paper is a collaboration. The norms model in it also sounds interesting.
www.full-stack-alignment.ai/paper
October 8, 2025 at 9:50 PM
I’m a fan of Joe Edelman. His Attentional Policies for eliciting a persons values was interesting before llms, and is incredibly more interesting now.
This new paper is a collaboration. The norms model in it also sounds interesting.
www.full-stack-alignment.ai/paper
This new paper is a collaboration. The norms model in it also sounds interesting.
www.full-stack-alignment.ai/paper
Reposted by Steve Steiner
🎙️ w/ @edelwax.bsky.social on “Full Stack Alignment: Co-Aligning AI and Institutions with Thick Models of Value.”
Pluralism as a core principle in social design, thick models of value, multipolar traps, moral graph elicitation, starting with membranes, Moloch-free zones, co-optation risks, & more.
Pluralism as a core principle in social design, thick models of value, multipolar traps, moral graph elicitation, starting with membranes, Moloch-free zones, co-optation risks, & more.
EP 325 Joe Edelman on Full-Stack AI Alignment - The Jim Rutt Show
Jim talks with Joe Edelman about the Meaning Alignment Institute's paper "Full Stack Alignment: Co-Aligning AI and Institutions with Thick Models of Value."
www.jimruttshow.com
October 8, 2025 at 12:59 PM
🎙️ w/ @edelwax.bsky.social on “Full Stack Alignment: Co-Aligning AI and Institutions with Thick Models of Value.”
Pluralism as a core principle in social design, thick models of value, multipolar traps, moral graph elicitation, starting with membranes, Moloch-free zones, co-optation risks, & more.
Pluralism as a core principle in social design, thick models of value, multipolar traps, moral graph elicitation, starting with membranes, Moloch-free zones, co-optation risks, & more.