| 10:21 |
s1gnp0st |
test from WIP pestron. |
| 10:28 |
s1gnp0st |
I put 0xFA in AGENTS.md and bellowed at gpt-5.1-codex until this popped out the other side. |
| 10:28 |
s1gnp0st |
(started yest, lmao) |
| 10:30 |
signpost |
nice, basic interop with blatta seems like it works. |
| |
~ 1 hours 26 minutes ~ |
| 11:56 |
asciilifeform |
signpost: neato! |
| 11:58 |
asciilifeform |
signpost: what's it made of ? |
| |
~ 33 minutes ~ |
| 12:31 |
awt |
signpost: wow! |
| 12:32 |
awt |
been using claude to fix some bugs on akris |
| 12:36 |
awt |
raip was written completely by claude. Been considering making an "archiver" plugin for akris to preservie links/images. |
| 12:38 |
asciilifeform |
awt: what kinda bugs did bot find ? (and what does the process look like?) |
| 12:38 |
* |
asciilifeform singularly unlucky to date w/ getting anyffing useful outta the gptbots |
| 12:39 |
asciilifeform |
(understatement, actually: typically they produce mind-bogglingly overcomplicated AND subtly-broken proggies, in asciilifeform's experience) |
| 12:40 |
awt |
asciilifeform: ui bugs - like hitting enter doesn't send the message in some environments, or focus doesn't switch to the text field. |
| 12:42 |
awt |
asciilifeform: process was just entering: "why does enter not send the message" basically - then claude just shows me the diff and I try it. |
| 12:43 |
awt |
For raip, I had claude generate a doc for the akris client api, then use that to "build a bot that connects to pest and accepts commands for querying a local llm" then iterating on that until it worked. Haven't even looked at the code - I assume it's garbage. |
| 12:49 |
awt |
Hopefully somehow we can get everyone on pest stations that support multipart - makes it much more convenient for bots to dump more info into the chat instead of having to link to paste bins. |
| |
~ 3 hours 12 minutes ~ |
| 16:02 |
signpost |
http://logs.nosuchlabs.com/log/pest/2025-11-15#1035556 << -std=c11 and sqlite |
| |
~ 2 hours 29 minutes ~ |
| 18:31 |
signpost |
weeeee shall see whether it did any good, but the spec is so nicely written it was straightforward to make the golem lay down tests per MUST. |
| |
~ 41 minutes ~ |
| 19:13 |
crtdaydreams |
someone might find some interest in some of these LLM integrations/experiments by Joe Marshall, one of the authors of the LMI lambda lispm https://funcall.blogspot.com/2025/08/llm-in-debugger.html |
| 19:14 |
crtdaydreams |
Talking to him, I believe asciilifeform was the one that reached out and asked him to post the remaining source code from it. In recent months he's been posting quite regularly and at great length about his exploration of LLMs in code-genration |
| |
~ 3 hours 49 minutes ~ |
| 23:04 |
awt_akris |
@!ask Summarize discussion around nat penetration and proposed solutions. |
| 23:07 |
raip |
awt_akris: https://paste.rs/Fqh9C.md |
| 23:07 |
awt_akris |
Hooked raip up to the pest logs ^ |
| |
~ 44 minutes ~ |
| 23:52 |
signpost |
awt, neat, this is a RAG thinger hooked up to logs? |
| 23:53 |
awt_akris |
signpost: yes. hooked up an mcp server that queries the local sqlite db. |