fmind-dev

I know several teams and big org struggling to leverage AI agents. One of the easiest path is to use UNIX (e.g., your laptop + a shell). All the tools are available, shared between humans and agents, and compose infinitively.

The point is exactly what you say: Not to have the latest tech product (e.g., grok bot, instinct, ...), but to just use powerful tools whose flexibility have been battle tested other the ages.

I'm a geek, a perfectionist, and I like to have the best tools when I work. If you take my previous dotfiles (100% human made): https://github.com/fmind/dotfiles-ansible, it had more than 1293 commits from all the tools and configs I tested. But this was draining my energy. I'm glad we now have LLM to automatically find the best tools/configs for you, and test if the settings are working.

My new dotfiles are much more powerful and efficient than ever before thanks to LLMs. For instance, I did not know about Zellij, mise, dprint, lazyvim or git-cliff before updating them. Having them is your toolbet helps me and my agents build better tools.

Indeed, you need to invest some tokens in configuring the tools, configs and skills. But once it's done, tools and configs do not consume more tokens, as they are automatically used by your AI Agents. Moreover, skills support progressive disclosure, and I try to keep a budget of max 5000 tokens for global + 5000 tokens for local setups (= Skills + AGENTS.md).

But yes, I burned too many tokens on this project. That's why I'll be happy if it can help others.

The article is more about justifing the investment. Dotfiles is the foundation for all your other AI projects. If you have good dotfiles, you can automatically steer all your other projects in the right direction.

I love the design, I'm sure there will be more knowledge framework coming in the near future.

I made a joke to my teammates today: "This will half the amount of AI startup in Europe ..."

Gemini Flash is one of the best "good-enough" models. I use this type of model daily, for automation and quick development iteration loops.

Unfortunately, it's often not strong enough for heavy refactoring and long running development loops.

I'm using Go more and more for my projects (AI, Web, CLI, ...). This is my go to for all my new projects.

My background is in data science and MLOps, where Python rules. But the focus is now less on building new AI models, and more on building the infrastructure and API calls with AI Agents. Go has a great async model, stellar performance, amazing tooling ecosystem, and far less ways of doing things than Python.

I except grow to become more and more popular, as our LLMs are now writing most of our code. Between a 50 MB portable binary in Go with 10x performance, and a 5 GB venv in Python with lack of proper parallelism, the choice is easy.

I'd love Django, if they had a better async story ... I used it for a recent project. While the framework is overall amazing, I had to switch to Go for better performance and easier async support.

API prices are amazing, but hosting this on-premise will be real challenge.

I've switched from Python to Go + HTMX recently. It's a really nice combo, fast, efficient, and easy to maintain. HTMX should be part of the HTML spec.

From my first tests today, it is a workhouse. It can scan my whole code base, optimize every part, with a greater level of autonomy than other tools. This is insane, we are living at the best time.

Glad to see Meta back on track! Users will benefit greatly from this competition.

It reminds me of the time where I deployed Gitea for self-hosting my git projects. In the end, nobody wanted to use it beyond myself. I would love to have a true federation protocol for Git, to decentralize the solution further.

I'm glad we have another player remaining in the competition. More competition means lower price and more quota for us (hopefully).

It takes time and effort to build such benchmark. It works at Databricks scale, I'm not sure smaller companies are ready to invest on internal benchmarks.

But they are more vendor neutral, now they don't sell their own model. It's interesting from a benchmark point of view.

I wonder how Mistral will prioritize its robotic development against its LLM development. We have either players that prioritize both (Google, AMI), or players that prioritize coding and agentic (OpenAI, Anthropic, ...).