Re: AGENTS.md

On 2026-08-06 10:27, Yuri Khan wrote:
> (I get the impression that Jean had started to use LLMs before the
> concept of a harness crystallized, and might be somewhat skewed in his
> assessment of how people at large use LLMs for coding.)

Maybe. So I have been using LLMs directly in Emacs buffer and still do. 
Last year August 6, I found Mistral Vibe... then I found it is probably 
hard coded to remote LLM. Then I realized many such software are coming 
but I found it so bloated and unfit for quick surgical problem solving. 
Then January 18 2026, I found "opencode" that can be used with local 
LLMs. It is useful (as root) to transfer hosted servers, set up SMTP, 
backup user's accounts, SSL certs, and all the admin stuff, and then I 
found Hermes, which is kind of useful to remotely access my databases. I 
have done serious projects with those packages, designing some machines. 
Though Emacs Lisp I have been solving mostly from within the buffer and 
web UI interfaces.

There may be emerging need for such software, yet personally, I have 
"memories" that I can use with the LLM any time I want, like:

95650 Dry deck ceiling, 📝 Note, 🧠 LLM Memory, so I don't personally need 
AGENTS.md and just as systemd, I have identified AGENTS.md (additional 
LLM context) as way for proprietary companies to dictate their way and 
make it good for them.

It increases the context length. More context length means more money 
for proprietary companies.

[2602.11988] Evaluating AGENTS.md: Are Repository-Level Context Files 
Helpful for Coding Agents?:
https://arxiv.org/abs/2602.11988

"while context files are useful for specifying non-standard coding 
practices, any attempts to improve performance should be rigorously 
evaluated before deployment"

That it is adopted by 60,000 software packages it means again more money 
for proprietary companies as let's face it, majority do use proprietary 
LLMs.

-- 
Jean Louis