Sign in

Lachlan Kermode

@ohrg.org
156 followers 208 following 257 posts

PhD Candidate at Brown University / software at liminal-lab.org. Freeing computers one terminal at a time.

PostsRepliesMedia
Reposted by Lachlan Kermode
Ranjodh Singh Dhaliwal @ranjodhdhaliwal.com · 13h
And this Friday at contralab.online, a closed workshop (funded, in part, by SNSF) on what might count as explanatory knowledge of/after large language models, ...
111
Lachlan Kermode @ohrg.org · 26/09/2026
I need to return to it! I remember finding the genealogy from Hume through to set theory very helpful, as a lot of programming quality discourse has often seemed to me a sort of schizophrenia between data empiricism and high modernism
021
Lachlan Kermode @ohrg.org · 26/09/2026
Potential research for this take:
press.uchicago.edu
Axiomatics
The first history of postwar mathematics, offering a new interpretation of the rise of abstraction and axiomatics in the twentieth century. Why did abstraction dominate American art, social science, a...
100
Lachlan Kermode @ohrg.org · 23/09/2026
We don't have any browser plugins or the like (equivalent to Org-capture). I haven't used those so much, but would be curious to understand them better
000
Lachlan Kermode @ohrg.org · 23/09/2026
Rookery is built in a way where you can litter ideas wherever in a folder, and everything can be picked up in the 'global' views. This is the associative archiving philosophy talked about in the docs.
100
Reposted by Lachlan Kermode
dan @danabra.mov · 14/09/2026
i still don't fully understand how to use them correctly but i just tell llm "make this shit monadic" and it does exactly what i want
2312
Lachlan Kermode @ohrg.org · 21/09/2026
To stay tuned for these updates and to follow rookery/Rheo's development, follow along through the socials of the Free Computing Lab: freecomputinglab.ohrg.org
freecomputinglab.ohrg.org
Index
000
Lachlan Kermode @ohrg.org · 21/09/2026
Rookery is also the basis for a more structured and local-first way of doing context engineering, the vision for which I wrote about on my blog: www.ohrg.org/ideas/birdke...
ohrg.org
Birdkeeping
100
Lachlan Kermode @ohrg.org · 21/09/2026
When combined with Rheo, rookeries are incrementally compiled in response to live edits, and browser pages are updated in place.
rheo.ohrg.org
Introduction | Rheo
110
Lachlan Kermode @ohrg.org · 21/09/2026
I built rookery to migrate my knowledge management system (read: scores of plain-text files in a folder) from Org-mode to Typst. It has now become the basis for my todo tracking, my book/paper/meeting notes, my brainstorming, my slideshowing, and---most importantly---my writing.
130
Lachlan Kermode @ohrg.org · 21/09/2026
Can't sleep at night for want of knowledge gardens (Zettelkasten-style) in Typst? Stay awake no longer because rookery is here: rookery.ohrg.org
232
Lachlan Kermode @ohrg.org · 20/09/2026
A much better example of a (tasteful) vibe-coded system than Yegge's
110
Lachlan Kermode @ohrg.org · 15/09/2026
I do love the FT: "if Anyhropic causes an extinction event, obviously that's bad for investors"
000
Lachlan Kermode @ohrg.org · 06/09/2026
The social contract between reader and writer in code Is different. That's not to say there aren't certain kind of code in circulation-- PRs to an open source repository, say-- that might hew more closely to prose's social contract.
000
Lachlan Kermode @ohrg.org · 06/09/2026
I also think this might be the foundation for more reasonable thinking on why writing CODE with LLMs might be more acceptable than writing prose with them.
100
Lachlan Kermode @ohrg.org · 06/09/2026
Started in 1994, still raging: www.versobooks.com/products/88-...
versobooks.com
Read My Desire
In Read My Desire, Joan Copjec stages a confrontation between the theories of Jacques Lacan and those of Michel Foucault, protagonists of two powerful modern disciplines—psychoanalysis and historicism...
020
Lachlan Kermode @ohrg.org · 06/09/2026
More sage thinking about LLMs and the symbolic function of writing from Cantrill: "to use an LLM to write is to void the social contract between writer and reader: we readers shouldn’t be expected to labor to understand a sentence that the writer themselves didn’t work to create."
121
Lachlan Kermode @ohrg.org · 11/08/2026
And my blog site has finally been updated with a way to subscribe to new content via email, if you want to stay abreast of the latest in token soup
000
Lachlan Kermode @ohrg.org · 11/08/2026
I've been stewing on what to say about the discomforting industrial metaphorics of Steve Yegge's beads and gastown for a while, and here it is: www.ohrg.org/birdkeeping
ohrg.org
Birdkeeping
110
Lachlan Kermode @ohrg.org · 06/08/2026
In addition to the process laid out in the blogpost, I've also started using jj workspaces as a way to isolate agents' work from each other while they are working on a bead. If you start using soroban, you'll get this workflow by default.
010
Lachlan Kermode @ohrg.org · 06/08/2026
The basic idea is that LLM agents do much better when you parcellate work for them in advance. This kind of planning also opens up more economical downstream agential workflows such as using a local model to implement work that a more capable model from one of the frontier labs has planned.
120
Lachlan Kermode @ohrg.org · 06/08/2026
I recently described this process at length in a blog post: www.ohrg.org/taming-jagge...
ohrg.org
Taming jagged intelligence with beads and jj
100
Lachlan Kermode @ohrg.org · 06/08/2026
I've just released github.com/freecomputin..., an extension set for your coding agent to work effectively with jj and beads.
github.com
GitHub - freecomputinglab/soroban: Work with beads and jj effectively in a coding agent
Work with beads and jj effectively in a coding agent - freecomputinglab/soroban
120
Reposted by Lachlan Kermode
Sophie Moullin @sophiemoullin.bsky.social · 31/07/2026
Struggling to diversify your panel? Use our humanoids.
011
Lachlan Kermode @ohrg.org · 17/07/2026
Wrote up my current context engineering setup (seeded with a few thoughts about LLM hysteria): www.ohrg.org/taming-jagge...
ohrg.org
Taming jagged intelligence with beads and jj
020
Lachlan Kermode @ohrg.org · 09/07/2026
Can't believe I missed this post when it happened. Outstanding
010
Lachlan Kermode @ohrg.org · 03/07/2026
You know you've been coding too long when you get up and, as you walk to the kitchen, start thinking about what should be included in the the prompt you're about to write to prepare dinner
000
Reposted by Lachlan Kermode
Gabriella "Biella" Coleman @biella.bsky.social · 30/06/2026
Does anyone have any opinions on these paperless drywall products? Please share. I doubt I will get much traction here, but it never hurts to check.
221
Lachlan Kermode @ohrg.org · 30/06/2026
(Sam Altman, Peter Thiel, Elon Musk, etc.)
000
Lachlan Kermode @ohrg.org · 30/06/2026
Let's bring back Weizenbaum's excellent phrase and accusation "the artificial intelligentsia"
110
Lachlan Kermode @ohrg.org · 28/06/2026
sehr gut!
010
Reposted by Lachlan Kermode
Seny Kamara @senykamara.com · 27/06/2026
This is a fantastic essay from @jessica.bsky.social about what it means to do real work/research open.substack.com/pub/letterst...
open.substack.com
on doing Real work
a working theory
2111
Lachlan Kermode @ohrg.org · 28/06/2026
The "third real" Jessica writes about here-- especially as it pertains to our work (labor) and its value-- is (I think) exactly what Marx would call communism, i.e. a social reality suffused with value(s) that meet our ethical standards, a reality that is "internally consistent".
000
Lachlan Kermode @ohrg.org · 28/06/2026
This essay really effectively pinpoints how the ideology of objectivity associated with computing ('the world is REALLY just data') reifies the world and consigns it to its current capitalist reality.
110
Lachlan Kermode @ohrg.org · 28/06/2026
The problem, in a nutshell, is the 'real' status of money. Money is a real abstraction, because it is a fictional contract that quantifies value; but it is a REAL abstraction because it's not just something you can choose to do without, as you might a standard 'abstraction'.
100
Lachlan Kermode @ohrg.org · 28/06/2026
The discussion of what makes for 'real' work here really effectively shadows an academic debate going on in the Marxology world currently about Alfred Sohn-Rethel's (mid-century Marxist) diagnosis of capital as a "real abstraction".
101
Lachlan Kermode @ohrg.org · 24/06/2026
"To write a program is to legislate the laws for a world one first has to create in imagination. Only very rarely does any designer, be he an architect... have so coherent a picture of the world emergent in his imagination that he can compose its laws without criticism from that world itself"
010
Lachlan Kermode @ohrg.org · 24/06/2026
"This misses the truth that programming is, again like any form of writing, more often than not experimental. One programs, just as one writes, not because one understands, but in order to come to understand."
110
Lachlan Kermode @ohrg.org · 24/06/2026
"Sometimes we connect two clauses with the word "therefore," only to then see that our chain of reasoning is defective."
110
Lachlan Kermode @ohrg.org · 24/06/2026
"Our pen writes the word "because" and suddenly stops. We thought we understood the "why" of something, but discover that we don't. We begin a sentence with "obviously"; and then see that what we meant to write is not obvious at all."
110
Lachlan Kermode @ohrg.org · 24/06/2026
Surely at least this: "Programming is rather a test of understanding. In this respect it is like writing; often when we think we understand something and attempt to write about it, our very act of composition reveals our lack of understanding even to ourselves."
110
Lachlan Kermode @ohrg.org · 23/06/2026
@shriram.bsky.social maybe an idea for your course. It could even be an audiobook assignment IMO
120
Lachlan Kermode @ohrg.org · 23/06/2026
Does anyone know of an intro to CS that includes Weizenbaum's _Computer Power and Human Reason_? What with this AI business, it seems more essential than ever to get this in front of computer people as early as possible
130
Lachlan Kermode @ohrg.org · 17/06/2026
There's also often a step 1.5, which is to interactively refine the beads issues before handing them off. I have the following general directive for plan mode to ensure the right information is included: github.com/breezykermo/...
000
Lachlan Kermode @ohrg.org · 17/06/2026
4. Follow the implementation with abacus - github.com/ChrisEdwards... - and tuicr - github.com/agavra/tuicr
100
Lachlan Kermode @ohrg.org · 17/06/2026
Coding workflow: 1. Have Anthropic's Opus family design a feature or fix as a set of beads issues: github.com/Dicklesworth... 2. Using either Anthropic's Sonnet family or GLM 4.7 (an open weights model), turn on caveman mode: github.com/juliusbrusse... 3. "br/jj pair": github.com/breezykermo/...
100
Lachlan Kermode @ohrg.org · 13/06/2026
Have just discovered github.com/chojs23/conc.... What a relief to be able to pare down the interface and navigate with Vi-style shortcuts
github.com
GitHub - chojs23/concord: A feature-rich TUI client for Discord
A feature-rich TUI client for Discord. Contribute to chojs23/concord development by creating an account on GitHub.
000
Lachlan Kermode @ohrg.org · 09/06/2026
One option would be to paywall ALL preshow rants (as that's what we really come for) and drop the free user unawares directly into the real topic
010
Lachlan Kermode @ohrg.org · 09/06/2026
Now that you've got serious VC money as a podcast company it's good to see you thinking seriously bout how to monetize.
110
Lachlan Kermode @ohrg.org · 08/06/2026
Okay, update: it turns out the upstream beads has changed its architecture. Now we use github.com/Dicklesworth...
github.com
GitHub - Dicklesworthstone/beads_rust: Fast Rust port of Steve Yegge's beads: local-first, non-invasive issue tracker storing tasks in SQLite with JSONL export for git collaboration
Fast Rust port of Steve Yegge's beads: local-first, non-invasive issue tracker storing tasks in SQLite with JSONL export for git collaboration - Dicklesworthstone/beads_rust
000