tacone

I am already doing that. For performance, I am just caching the latest explanation alongside the code.

The patterns in the article might be a starter, but there's so much more to cover:

agents role (Orchestrator, QA etc.), agents communication, thinking patterns, iteration patterns, feature folders, time-aware changelog tracking, prompt enforcing, real time steering.

We might really need a public Wiki for that (C2 [1] style)

[1]: https://wiki.c2.com/

Starred.

I am looking to that exact concept - for a different mean - to develop my agent orchestration hobby project.

LLM working in a «feature folder» where it stores change-logs, documentation, summaries, requirements, attachments, and so on. What I will be looking into very soon, is also storing the session id, for summarisation, history context, and so on.

I will definitely steal some concept from your project.

For Claude the most pollution usually comes from Claude itself.

It's worth noting thet just by setting the right tone of voice, choosing the right words, and instructing it to be concise, surgical in what it says and writes, things change drastically - like night and day.

It then starts obeying, CRITICALs are barely needed anymore and the docs it produces are tidy and pretty.

Value makes the world a better place, not profits per-se. Sometimes the two things coincide, sometimes they don't.

For some reason I made my own skills framework.

The LLM instructed to run the initialization script as the first thing, before reasoning about the use request (this proved tricky to achieve). The scripts greps the content matter out of the skill files, along with the file path.

I have no clue if this outperforms an embedded index.

Most often product development involves a lot of legacy. In consultancy you get to start from scratch at least every once in a while.

I am asking my agents to use HTTPie this day, it's more understandable for them in case of error.

The real future: when Claude/ChatGPT/Gemini have banned you, you must leave society, you employment, the planet.

Side question, I am currently using Github copilot, what would be a good reason to switch provider? Looks like I am almost the only one I here using it.

I didn't read the whole article and constitution yet, so my point of view might be superficial.

I really think that helpfulness is a double-edged sword. Most of the mistakes I've seen Claude make are due to it trying to be helpful (making up facts, ignoring instructions, taking shortcuts, context anxiety).

It should maybe try to be open, more than helpful.

The way I see it, is that for non-trivial things you have to build your method piece by piece. Then things start to improve. It's a process of... developing a process.

Write a good AGENTS.md (or CLAUDE.md) and you'll see that code is more idiomatic. Ask it to keep a changelog. Have the LLM write a plan before starting code. Ask it to ask you questions. Write abstraction layers it (along with the fellow humans of course) can use without messing with the low-level detail every time.

In a way you have to develop a framework to guide the LLM behavior. It takes time.

Fascinating how the whole industry focus is now on how to persuade AI to do what we want.

Two AGENTS.md tricks I've found for Claude:

1. Which AI Model are you? If you are Claude, the first thing you have to do is [...]

2. User will likely use code-words in its request to you. Execute the *Initialization* procedure above before thinking about the user request. Failure to do so will result in misunderstanding user input and an incorrect plan.

(the first trick targets the AI identity to increase specificity, the second deliberately undermines confidence in initial comprehension—making it more likely to be prioritized over other instructions)

Next up: psychologists specializing in persuading AI.

Thank you for the beautiful story. I work as a developer and have experienced the same in my personal projects, linux setup and - in general - all the collaterals.

AI is eroding the entry barrier, the cognitive overload, and the hyper-specialization of software development. Once you step away from a black-and-white perspective, what remains is: tools, tools, tools. Feels great to me.

> [Danger: Libertarian detected, rational communications may not work]

The parity bit gets immediately unreliable.

Thank you!

Do we need some kind of standardized URL syntax (like # for anchor) to have browsers take you to the sub-content and highlight it?

I'd argue that it's easier to debug a few lines of magical code than 10x lines if non-magical code.

Omarchy is a nice experience for devs.

Funnily enough, though, you can get a very user friendly experience using Niri and Dank Linux (don't remember the exact name). It takes two 3 CLI commands to install, and the top bar incredibly cool, compared to the i3 defaults and even to what I remember of Gnome and KDE.

Next up: somebody comes up with a desktop environment called BTW.

I use a folder for each feature I add. The LLM is only allowed to output markdown file in the output subfolder (of course it doesn't always obey, but it still limits pollution in the main folder)

The folder will contain a plan file and a changelog. The LLM is asked to continously update the changelog.

When I open a new chat, I attach the folder and say: onboard yourself on this feature then get back to me.

This way, it has context on what has been done, the attempts it did (and perhaps failed), the current status and the chronological order of the changes (with the recent ones being usually considered more authoritative)

No, setting the temperature to zero is still going to yeld different results. One might think they add random seeds, but it makes no sense for temperature zero. One theory is that the distributed nature of their systems adds entropy and thus produces different results each time.

Random seeds might be a thing, but for what I see there's a lot demand for reproducibility and yet no certain way to achieve it.

You might try to find some local group or club about one of you interests. Having common ground might help.

I long for quantum computing where white space will be able to be a space and a tab at the same time.

Quite funny coincidence with 418 HTTP status code.

Problem is, Copilot CLI doesn't really supports free or mini models. You have very tight choice of models. This looks like product decision. I understand why they won't allow you to use the free models on CLI, but not being able to use the (pay for) mini models is beyond me.

Is it already there? Pretty cool.

I was thinking to start planking (not just for fitness but also as a mean for stress inoculation).

1. You can do it at home. 2. It's short: when starting out, you might just be capable of holding the position max 1 minute. 3. you don't need special dressing or tools, the floor is all you need (a yoga mat would make it nicer though) 4. when done on the elbows, you can even use your phone while you do it (provides you consider that healthy)

It really seems a no brainer form my angle.

I am also getting interest in Fascia training (that's for old men, mind you)

Documentation is never really in sync, thus it's an heuristic. Henceforth, you could setup a pipeline to use AI for that. Not everything will be correct, but it never really is (unless the project is very small and focused or is very old and changes slowly).

Depending on your needs it might be a viable solution.

No, unless they also use subagents.

> This project was developed with assistance from AI/LLMs (including GitHub Copilot, ChatGPT, and related tools), supervised by a human who occasionally knew what he was doing.

That's the good part of AI. Lowers effort and knowledge barrier and makes things possible.