Piloting Agentic Engineering

✈️ Piloting Agentic Engineering👀 Using Claude Code: The Unreasonable Effectiveness of HTML🤖 How LLMs Work — A Visual Deep Dive🧐 Treat Agent Output Like Compiler Output🐐 The Story of Mel, a Real Programmer

Have a great week.

Interesting, I experienced the same atrophy of skills because I use a lot of agents with Cursor. I love the parallel with the auto-pilot and the pilot, even with an auto pilot the pilot needs to be able to fly in manual. In software engineering, we need to use theses new tools but not every day and all the time because we can loose our skills if we offload it entirely to the AI.

Instead of keeping every plan, documentation etc… in markdown file, Thariq transform them into HTML files to have a better readbility and make them interactive. See the example.

Very cool guide which explain the 7 steps to build an LLM like ChatGPT / Claude / Gemini.

I am conflicted because, it’s true in a sense, we can’t review the enormous output of the AI but thinking we just need to build better tooling in order to let the AI writing code freely is too soon for me. I still love writing good and performant code by hand but yeah I agree on the tooling part.

It’s ridiculous to know that this article is written in 1983 ( 53 years ago !!! ), in this time, Fortran is all the rage and engineer still programm in hexadecimal code !!! I really love this story because if someone know the in and out of a machine, they can do some incredible things. In comparaison with the precedent article, I find it sad now we just want to generate a lot of code, not looking ate the output and a lot of people don’t care about performance anymore. But there always be people like Mel to optimize and caring about their craft :).

Web tools and Library