From 9bf45afc4bc5ed32555d3591d4f58df2ca6be0b3 Mon Sep 17 00:00:00 2001 From: sepia Date: Wed, 30 Jul 2025 14:05:19 -0500 Subject: [PATCH 1/3] Post about AI memory --- data/posts/other_post.html | 17 +++++++++++++---- 1 file changed, 13 insertions(+), 4 deletions(-) diff --git a/data/posts/other_post.html b/data/posts/other_post.html index 51c60b1..320116c 100644 --- a/data/posts/other_post.html +++ b/data/posts/other_post.html @@ -1,8 +1,17 @@ - - + +
-

Welcome to another post!

-

another one!

+

Things AI Agents Should be Able to Do

+

Write Their Own Helper Scripts

+

+ If I ask my agent to play chess with me (or a game that doesn't exist), it should write a basic chess engine, unmprompted, in a sandboxed environment, which it can then use as a tool the next time it plays chess. Nobody has implemented this yet because it requires a lot of agency from the agent, and would have a bad error rate. I think it can be done effectively by limiting the agent to a single tech stack and a railroaded workflow for creating its own tools. Like a human software engineer, an AI agent should seek to automate tedious tasks. +

+

Forget

+

+ Humans still have better memories than LLMs, because we forget. Agents can be fitted with a RAG to quickly memorize trivia, but if you have your agent remembering everything it does all day, the database gets cluttered with information that doesn't matter, and recall brings up a lot of garbage. This effect would be exacerbated if you fed an agent a firehose of information from your camera-glasses, which is exactly how much information your own monkey brain is intaking. +
+ Human memory should be used as a model: when information is repeated, we remember it more strongly, and if it's not, we forget. Additionally, humans spend time "alone with our thoughts" reflecting on our memories and creating new super-memories that are much denser with useful information (For example, you might remember "My friend James likes strawberry ice cream," a reflection that allows you to throw away 200 instances of memories of him ordering strawberry ice cream). +

From be87c90b47846f8e867234ed1e88f44beafff933 Mon Sep 17 00:00:00 2001 From: sepia Date: Wed, 30 Jul 2025 14:06:03 -0500 Subject: [PATCH 2/3] Remove test post and move AI memory post --- data/posts/test_post.html | 4 ---- ...r_post.html => things_ai_agents_should_be_able_to_do.html} | 0 2 files changed, 4 deletions(-) delete mode 100644 data/posts/test_post.html rename data/posts/{other_post.html => things_ai_agents_should_be_able_to_do.html} (100%) diff --git a/data/posts/test_post.html b/data/posts/test_post.html deleted file mode 100644 index 2c5fdf8..0000000 --- a/data/posts/test_post.html +++ /dev/null @@ -1,4 +0,0 @@ -
-

Test

-

This is just a test!

-
diff --git a/data/posts/other_post.html b/data/posts/things_ai_agents_should_be_able_to_do.html similarity index 100% rename from data/posts/other_post.html rename to data/posts/things_ai_agents_should_be_able_to_do.html From 232855e95213d86d955aefd009f71f32846e01e1 Mon Sep 17 00:00:00 2001 From: sepia Date: Wed, 30 Jul 2025 19:42:53 -0500 Subject: [PATCH 3/3] Add code link, and fix broken br tag in a post --- data/pages/about.html | 4 +-- ...things_ai_agents_should_be_able_to_do.html | 27 ++++++++++++++++--- 2 files changed, 25 insertions(+), 6 deletions(-) diff --git a/data/pages/about.html b/data/pages/about.html index e90b69a..e123ee2 100644 --- a/data/pages/about.html +++ b/data/pages/about.html @@ -37,7 +37,7 @@

- This website was made with htmx and rust. I will not post the code unless - you bother me about it, because I am lazy. + This website was made with Elysia and HTMX. You can + read the code.

diff --git a/data/posts/things_ai_agents_should_be_able_to_do.html b/data/posts/things_ai_agents_should_be_able_to_do.html index 320116c..0cf85f8 100644 --- a/data/posts/things_ai_agents_should_be_able_to_do.html +++ b/data/posts/things_ai_agents_should_be_able_to_do.html @@ -6,12 +6,31 @@

Things AI Agents Should be Able to Do

Write Their Own Helper Scripts

- If I ask my agent to play chess with me (or a game that doesn't exist), it should write a basic chess engine, unmprompted, in a sandboxed environment, which it can then use as a tool the next time it plays chess. Nobody has implemented this yet because it requires a lot of agency from the agent, and would have a bad error rate. I think it can be done effectively by limiting the agent to a single tech stack and a railroaded workflow for creating its own tools. Like a human software engineer, an AI agent should seek to automate tedious tasks. + If I ask my agent to play chess with me (or a game that doesn't exist), it + should write a basic chess engine, unmprompted, in a sandboxed environment, + which it can then use as a tool the next time it plays chess. Nobody has + implemented this yet because it requires a lot of agency from the agent, and + would have a bad error rate. I think it can be done effectively by limiting + the agent to a single tech stack and a railroaded workflow for creating its + own tools. Like a human software engineer, an AI agent should seek to + automate tedious tasks.

Forget

- Humans still have better memories than LLMs, because we forget. Agents can be fitted with a RAG to quickly memorize trivia, but if you have your agent remembering everything it does all day, the database gets cluttered with information that doesn't matter, and recall brings up a lot of garbage. This effect would be exacerbated if you fed an agent a firehose of information from your camera-glasses, which is exactly how much information your own monkey brain is intaking. -
- Human memory should be used as a model: when information is repeated, we remember it more strongly, and if it's not, we forget. Additionally, humans spend time "alone with our thoughts" reflecting on our memories and creating new super-memories that are much denser with useful information (For example, you might remember "My friend James likes strawberry ice cream," a reflection that allows you to throw away 200 instances of memories of him ordering strawberry ice cream). + Humans still have better memories than LLMs, because we forget. Agents can + be fitted with a RAG to quickly memorize trivia, but if you have your agent + remembering everything it does all day, the database gets cluttered with + information that doesn't matter, and recall brings up a lot of garbage. This + effect would be exacerbated if you fed an agent a firehose of information + from your camera-glasses, which is exactly how much information your own + monkey brain is intaking. +
+ Human memory should be used as a model: when information is repeated, we + remember it more strongly, and if it's not, we forget. Additionally, humans + spend time "alone with our thoughts" reflecting on our memories and creating + new super-memories that are much denser with useful information (For + example, you might remember "My friend James likes strawberry ice cream," a + reflection that allows you to throw away 200 instances of memories of him + ordering strawberry ice cream).