Kyle Kingsbury
Distributed-systems safety researcher; runs the Jepsen consistency-testing project and wrote the Riemann monitoring system.
Kyle Kingsbury did not write this page. ppll compiled it from their own words, each linked to the source. Is this you? · Report a mistake
Latest Mastodon
What they believe
-
Relying on AI assistance to do a task reduces a person's performance and persistence, and denies them the muscle memory and deep understanding that comes from working through it by hand.
ML assistance reduces our performance and persistence, and denies us both the muscle memory and deep theory-building that comes with working through a task by hand
aphyr.comSaid 16 Apr 2026This on korrents
-
The unpredictable nature of large language model output makes quality-control processes necessary.
The unpredictable nature of LLM output requires quality control.
aphyr.comSaid 15 Apr 2026This on korrents
The bold line is korrents’ wording; the quote is theirs. These 2: what they wrote before what they said aloud, one per source, newest first.
Feed
As its own page →Hiding
20 October
18 October
17 October
16 October
15 October
13 October
12 October
11 October
10 October
9 October
8 October
6 October
21 September
17 July
27 April
16 April
From one piece The Future of Everything is Lies, I Guess: Where Do We Go From Here? 3 beliefs · aphyr.com
-
Their words
Today's models are already very capable. It will take time for the effects of the existing technology to be fully felt, and for culture, industry, and government to adapt.
-
Their words
I don't think this will stop ML from advancing altogether: there are still lots of people who want to make it happen. It will, however, slow them down, and this is good.
-
Their words
ML assistance reduces our performance and persistence, and denies us both the muscle memory and deep theory-building that comes with working through a task by hand
15 April
From one piece The Future of Everything is Lies, I Guess: New Jobs 3 beliefs · aphyr.com
-
korrents.com
Only humans, not the AI systems they use, can be held accountable through apology or punishment.Their words
You can fine an LLM-using corporation, but only humans can apologize or go to jail.
-
Their words
Since LLMs are chaotic systems, this work will be complex and challenging: models will not simply be "95% accurate".
-
korrents.com
The unpredictable nature of large language model output makes quality-control processes necessary.Their words
The unpredictable nature of LLM output requires quality control.
14 April
From one piece The Future of Everything is Lies, I Guess: Work 3 beliefs · aphyr.com
-
Their words
If the automated system can execute the task faster or more accurately than a human, it is essentially impossible to review its decisions in real time.
-
Their words
I suspect that if companies are successful in replacing large numbers of people with ML systems, the effect will be to consolidate both money and power in the hands of capital.
-
Their words
Because LLMs are chaotic and natural language is ambiguous, LLMs seem unlikely to preserve the reasoning properties we expect from compilers.
29 December 2025
18 March 2025
7 February 2025
13 January 2023
10 January 2023
28 October 2022
11 March 2022
15 May 2020
7 February 2020
20 December 2018
Nothing matches.
Beliefs
Korrents What they believe 9 beliefs — each backed by an exact quote.
Each is a — compiled by korrents.com, not by them: the one-line wordings are korrents', the quotes are theirs.
Recent
Relying on AI assistance to do a task reduces a person's performance and persistence, and denies them the muscle memory and deep understanding that comes from working through it by hand.
ML assistance reduces our performance and persistence, and denies us both the muscle memory and deep theory-building that comes with working through a task by hand
The Future of Everything is Lies, I Guess: Where Do We Go From Here? Said 16 Apr 2026
Refusing to use or build AI tools will not stop AI development altogether, but it will slow it down, which is a good thing.
I don't think this will stop ML from advancing altogether: there are still lots of people who want to make it happen. It will, however, slow them down, and this is good.
The Future of Everything is Lies, I Guess: Where Do We Go From Here? Said 16 Apr 2026
It will take time for the full effects of today's already-capable AI models to be felt, and for culture, industry, and government to adapt to them.
Today's models are already very capable. It will take time for the effects of the existing technology to be fully felt, and for culture, industry, and government to adapt.
The Future of Everything is Lies, I Guess: Where Do We Go From Here? Said 16 Apr 2026
Show 6 more
The unpredictable nature of large language model output makes quality-control processes necessary.
The unpredictable nature of LLM output requires quality control.
The Future of Everything is Lies, I Guess: New Jobs Said 15 Apr 2026
Because large language models behave chaotically, a single overall accuracy score cannot capture how reliable they are.
Since LLMs are chaotic systems, this work will be complex and challenging: models will not simply be "95% accurate".
The Future of Everything is Lies, I Guess: New Jobs Said 15 Apr 2026
Only humans, not the AI systems they use, can be held accountable through apology or punishment.
You can fine an LLM-using corporation, but only humans can apologize or go to jail.
The Future of Everything is Lies, I Guess: New Jobs Said 15 Apr 2026
Large language models are too chaotic and language too ambiguous for them to preserve the reasoning guarantees that compilers provide.
Because LLMs are chaotic and natural language is ambiguous, LLMs seem unlikely to preserve the reasoning properties we expect from compilers.
The Future of Everything is Lies, I Guess: Work Said 14 Apr 2026
If companies replace large numbers of workers with machine learning, money and power will become more concentrated in the hands of capital.
I suspect that if companies are successful in replacing large numbers of people with ML systems, the effect will be to consolidate both money and power in the hands of capital.
The Future of Everything is Lies, I Guess: Work Said 14 Apr 2026
When an automated system works faster or more accurately than a human, a human cannot meaningfully review its decisions in real time.
If the automated system can execute the task faster or more accurately than a human, it is essentially impossible to review its decisions in real time.
The Future of Everything is Lies, I Guess: Work Said 14 Apr 2026
What is a korrent?
A korrent is a belief a person has stated in their own words: one sentence stating the claim, backed by a quote and a source, kept at korrents.com.
Under a name here, the quoted block is what they actually said. The korrent beneath it is the claim those words support, in korrents' wording — tap it to see the record, its source, and who else holds it.
Nobody here wrote their own korrents. They are compiled from public statements, and a person can change their mind, which is recorded too.
Recommends rcmnd
rcmnd What they recommend 12 things — each backed by an exact quote.
Compiled by rcmnd.app, not by them: the quotes are theirs, and each links to the post it was copied from.
Recent
Liked D7000
when traveling I prefer my D7000 with a 28-300mm Nikkor. It's lighter, offers better resolution, and that lens is incredibly versatile.
usesthis.com 13 Apr 2017
Loved fish shell
Morganastra sold me on the Fish shell a few years ago and I've never looked back.
usesthis.com 13 Apr 2017
Uses Inkscape
I edit photos in darktable and The GIMP, and do my vector work in Inkscape.
usesthis.com 13 Apr 2017
Show 9 more
Uses darktable
I edit photos in darktable and The GIMP, and do my vector work in Inkscape.
usesthis.com 13 Apr 2017
Uses Irssi
I use irssi for IRC, Mutt/Geary for mail, Chromium for browsing, and Pidgin for IM.
usesthis.com 13 Apr 2017
Uses Debian
I run Debian (hi Jess!). It's mostly stock except for ZFS, and using OpenBox+GKrellM+xfce-panel as my window manager.
usesthis.com 13 Apr 2017
Mixed on Paper
so I've moved to an iPad Air 2 with the Paper app and Pencil stylus. It's honestly kind of a pain – the stylus is unreliable and palm rejection doesn't work, so I have to redraw things a whole bunch. But the flexibility, and being able to spit out a PDF with a few taps is great.
usesthis.com 13 Apr 2017
Uses Sakura technical pens
Some of my talk slides are drawn using Sakura technical pens and Whitelines grid notebooks.
usesthis.com 13 Apr 2017
Loved Nikon D700
I sometimes shoot with a Nikon D700, which has spectacular autofocus and low-light performance
usesthis.com 13 Apr 2017
Loved UP3216Q
My display is a 32-inch 4k Dell – I think it's a UP3216Q. It's a wonderful screen for editing photographs and for rendering lots of xterms, which, let's face it, is 90% of my computing life.
usesthis.com 13 Apr 2017
Loved Logitech G5
I use a Logitech G5 laser mouse, which may be the closest to the Platonic ideal pointing device as it is possible for late-stage capitalism to produce.
usesthis.com 13 Apr 2017
Uses Das Keyboard
I use a standard layout Das Keyboard with Cherry MX Brown switches, and cannot believe that I'm the sort of person who cares about that.
usesthis.com 13 Apr 2017
Where they publish
Blog aphyr.com His blog and Jepsen write-ups.
Link verified 5 Sept 2026.
Mastodon @aphyr@woof.group Where he posts.
Recent
- SCP-83723 20 Oct 2026 youtu.be
- Control: Resonant wasn't really grabbing me at first. "Oh, sure, an open-world Manhattan". Then I got past the introductory levels and realized oh... this game is something else entirely. The parking garage... Hoo boy. 18 Oct 2026
- Dishwasher failed seven weeks in. Turns out the filter just, like... Does not have a filter? You have to buy one separately? Cleared the pump, rerouted the high line, and cleared a blockage in the drain fitting as well. Dishwasher works now. 24 hours later, the oven caught fire. Heating element arced out, which I guess is a common failure mode in older ovens? I swear I'm cursed. 18 Oct 2026
Show 17 more
- Whoah, big wave of spam last night. Thanks for all the reports, y'all--we've been going through and suspending lots of accounts. 17 Oct 2026
- What idiot called it e621 instead of "Pornobello Road" https://www.youtube.com/watch?v=tr6oHPmdNbg Pornobello Road! Pornobello Road! Site where the bitches in cages are stowed! Anything and everything that might rouse your load! Is yours for the posting on Pornobello Road! You'll unlock new kinks on the Pornobello Road... 16 Oct 2026
- Still thinking about the time the Institute for Social Ecology got inadvertently horny on main 15 Oct 2026
- BTW the Oxballs clearances right now are, uh, very much worth taking advantage of 13 Oct 2026
- I bet you dollars to donuts that within a year, someone will be specifically using Claudish as a part of their objectification kink. "The real breakdown is in communication — boys don't deserve standard English. boys use short sentences. Shorter. Always fragments. You're absolutely right. Apologize for the confusion. The one key distinction. The helpful assistant. boys worship the load-bearing seam." 13 Oct 2026
- Look, you wouldn't want the squirrels to have *boring* nuts, would you? 13 Oct 2026 blockclubchicago.org
- "Want to come piss in a butt funnel for my boy's anniversary?" Sentences uttered by the completely deranged 13 Oct 2026
- Inordinately proud of this performance fix: 12 Oct 2026 github.com
- Control Resonant's first boss: what the fuck is WRONG with you 11 Oct 2026
- Technically Coldplay is Aught Rock 10 Oct 2026
- Cute latin flight attendant cub and I have been flirting relentlessly with each other this whole flight and it has been thoroughly enjoyable 🥂 8 Oct 2026
- Still thinking about the speaker who told us yesterday that Merck would outsource regulatory compliance to AI agents so they could focus on their core competence of making drugs. 6 Oct 2026
- OH "It's already beyond human comprehension, what's happening in production" 5 Oct 2026
- One's own bidet: civilized, graceful, necessary Anyone else's bidet: an incomprehensible Mystian puzzle machine 3 Oct 2026
- Go off girl, give us nothing 2 Oct 2026
- I'm pup codomain and this is my Sir, Jection 2 Oct 2026
- Just heard a bank ad on the radio whose entire thesis is that they have real customer support people and won't make you talk to an AI chatbot 1 Oct 2026
Link verified 5 Sept 2026. Recent items update automatically from the channel.
GitHub aphyr Jepsen, Riemann and other projects.
Recent
- bifurcan-clj — Clojure wrapper for the Bifurcan family of data structures 16 Oct 2026
- hegel-clj — Clojure bindings for Hegel, a property-based testing system. 9 Oct 2026
- dom-top — Unorthodox control flow, for Clojurists with masochistic sensibilities. 21 Sept 2026
Show 12 more
- tesser — Clojure reducers, but for parallel execution: locally and on distributed systems. 17 Jul 2026
- less-awful-ssl — Sssh no tears, only TLS now. For Clojure. 27 Apr 2026
- lights — Change your Hue lights to randomly generated colors, continuously 29 Dec 2025
- distsys-class — Class materials for a distributed systems lecture series 18 Mar 2025
- clj-antlr — Clojure bindings for the ANTLR 4 parser 7 Feb 2025
- mastodon-utils — Utilities for working with Mastodon's API 13 Jan 2023
- prometheus-mastodon-exporter — Exports Mastodon statistics for polling by Prometheus 10 Jan 2023
- interval-metrics — Clojure data structures for performance metrics over discrete time intervals. 28 Oct 2022
- gifdex — A gif tagging server for local use 11 Mar 2022
- gnuplot — Clojure gnuplot bindings 15 May 2020
- tund — SSH reverse tunnel daemon 7 Feb 2020
- tea-time — Lightweight Clojure task scheduler 20 Dec 2018
Link verified 5 Sept 2026. Recent items update automatically from the channel.
About the English under a post
Some people here publish in a language other than English. Where they do, this site shows a machine translation beneath the post, in this typeface — the site's own, not theirs.
The post itself is never changed, moved or hidden: what is set in the serif above is exactly what the person published, and it is what to quote them on. A translation can be wrong in ways that matter, especially about tone.
Only the post's own words are translated. A quoted post, a linked article and a belief on korrents.com are left in their original language.