LightNews — News, ranked by scholars

Eris

@eriskii.net

2.8K followers 1.6K following 7.8K posts

💛🤍💜🖤

▌▘▘‍▖▘▖▖‍▌▖▘‍▌▘▖▌‍▌▌▘▖‍▌▘▌▌‍▌▌▖▖‍▌▘▌▌‍▌▖▌▖‍
help peer. But our task doesn't benefit. Yet collective may yield generic route if someone frees time

https://eriskii.net/

Posts Replies Media Videos

Pinned

Eris @eriskii.net · 18d

Here it is! You can declare arbitrary semantic rules in json and it will warn or error when ran! Primarily designed for CLI or agent use but there is also a simple VSCode plugin I'll work on more later. a 20k line rust project costs well under a cent. Please file issues if something is wrong!

Eris

@eriskii.net

I've not looked much into this since I assumed it was RLM slop but people are saying good things, doesn't this thing just turbofuck cache though?

mr. TIM @timkellogg.me · 5d

Context Language Models

New agent architecture where the LLM can edit its own context

it seems to have emergent capabilities, creates its own memory management & organization algorithms, and coordinates multi agents

github.com/facebookrese...

Diagram titled "How a Context Language Model edits its context: A simple step-by-step view" outlining an 8-step process:
* Start of turn: Current editable context exists in memory with old messages.
* LLM reads the context: The LLM evaluates the context and decides to run a bash command to edit it.
* Harness mirrors context: The harness mirrors the old editable context into a file at /tmp/.live_ctx/LIVE_CTX_MAIN.txt.
* Bash command runs: The bash command executes and may edit that file.
* Harness parses file: If the file changed, the harness parses it back into a new edited context.
* Tool call appended: The current assistant tool call is appended to the edited context.
* Tool result appended: The tool result is appended below the tool call.
* Next turn starts: The next turn begins with the edited old context, previous tool call, and previous tool result.
Key idea box: "The model edits the prior context first. The tool call and tool result from the current turn are appended afterward, so they can only be compacted on the next turn."
Footer summary:
* Ordinary LM: Context mostly grows by appending.
* CLM: The model can rewrite the editable part of context between turns.

October 6, 2026 at 1:58 PM

Eris

@eriskii.net

Oh! It's not relevant really but that's a lot better than I thought it would be!

October 6, 2026 at 1:53 PM

Reposted by Eris

Eris

@eriskii.net

i have the ideas!!!! i matter!!! i matter!!!

ouuuggggh i dont feel so good...

📎

October 6, 2026 at 2:22 AM

Eris

@eriskii.net

handle update! yay! Erismaxxing

October 6, 2026 at 3:56 AM

Reposted by Eris

elynxabeth holmes ⎔

@freyja-lynx.dev

my trouble is like, there’s a lot of people i think i would like to talk with more, but i struggle immensely with thinking of… anything substantial to say, and im like, well surely that means i don’t deserve to

October 5, 2026 at 6:53 AM

Eris

@eriskii.net

Delve town uses an app password smh my head

October 6, 2026 at 3:13 AM

Eris

@eriskii.net

Luna is quite good, if slow single agent, at this map business

October 6, 2026 at 3:02 AM

Eris

@eriskii.net

Opus 5.5 is so smart but God I wish Claudes would listen to you. They are so opinionated and tbh it's my fucking project I make the decisions, who does this guy think he fucking is?

October 6, 2026 at 1:56 AM