Hello everyone, Iโm Yazan Tarifi ๐
Iโm a Software Engineer focused on building real products, with a strong background in native mobile development.
Here, Iโll share what Iโm building, what Iโm learning, and my thoughts about software engineering and technology.
Iโm especially interested in:
โข Native Mobile Development ๐ฑ
โข Android, Kotlin & Compose Multiplatform
โข Backend Engineering, especially NestJS
โข Software Architecture & System Design
โข AI, LLMs & Coding Agents
โข Performance, scalability & developer tools
Always learning. Always building. ๐
The stable release of AndroidX Security State v1.1.0 and Security State Provider v1.0.0 are here. ๐
These libraries bring transparency to device security by letting developers and MDM solutions programmatically verify component-level protection, check pending updates, and move beyond monolithic patch levels โ https://t.co/lcfdKgpcSu
Because the users that vibe coding their projects does not even understand anything in IT or software development, they always say to ai pick the best one, yes, yes
Thats why they let the ai take all decisions , everything
But in real software development world we plan, study the technologies, decide the roadmap then we implement in instructions, a full roadmap and the specific development rules
But vibe coding code is a nightmare and he even does not know how the implemented code even working , i saw users telling AI to run the project, they even dont know what is the commands and what is the args ๐
Yes always dont let agents and models decide on its own and searching in all files
1- Build a perfect Agents.md file
2- Build a perfect Architecture.md file
3- Setup Nested Agents.md file in each level
4- Build Skills and agents and mention them in Agents file
Agents file should not be large, but clear to understand what is the project, which framework, what configurations, then mention inside it files to read when it needs something
Architecture file is the development rules, restrictions, do and dont list, folders and classes structure
Nested agents file is per feature, dont let the agent to read a lot of files to understand what is this feature, let it read one file per feature and its ready to start edit
Agents that can spawn automatically by your prompts
In this way models should not read the whole code just to understand where it should edit, insted it read one md file and know what is this feature and where to go exactly
One of the most underused features in AI is Agents.
Invest once in building specialized agents with clear rules:
1. When they should trigger
2. How they should code
3. Which architecture to follow
4. What they should never touch
Then something changes:
AI stops randomly generating code.
It starts working like an engineer inside your system. ๏ฟฝ๏ฟฝ
Build the agent once.
Reuse the intelligence everywhere.
To do both i recommend Gaming PC with normal keyboard not mechanical
So its easier to code on but also you can have both keyboards and switch between them
But gaming pc is the option because you can build and upgrade the components as you like to push it to the maximum performance
ChatGPT web search is seriously underrated.
It doesnโt feel like normal Google search where you open 10 results and dig through each one.
It searches deeper into pages, links, docs, and sources, then gives you the exact information with the exact link you actually need.
One search. Direct answer. ๐โก