What I'll be posting about:
๐ง AI tools worth trying (and ones that aren't)
๐ What's trending on GitHub in AI/agent tooling
๐ก Lessons from shipping AI in production
๐ฎ Building side projects as a full time dev
Axios just got compromised on NPM. Remote access trojan in malicious packages. 871 points on HN.
AI coding agents install packages on autopilot now. Pin your versions. Audit your lockfiles.
This conversation isn't happening enough. ๐
If you are spending the first hour of your day on stuff that could have been done overnight, that is an hour you are never getting back.
All of this runs on scheduled tasks and a few prompts. Happy to share the setup if anyone is curious.
The thing that surprised me is how much it changed my mornings.
I used to start every day reactive, catching up, figuring out what happened.
Now I start with priorities sorted, tools discovered, and metrics ready.
@Shivam25mishra The same way Claude tries to fix a problem, it tries over and over again until it finds the right solution and learns from its mistakes.
I spent the last week researching multi-agent AI tools.
Instead of one AI doing everything, you give specialised agents their own jobs. One researches. One writes. One reviews. They coordinate like a small team.
People are shipping this right now. Here is what I found ๐งต
Where to start (all free, all open source):
Mastra (TypeScript) - from the Gatsby team
CrewAI (Python) - most popular
deer-flow (ByteDance) - 45K stars this week
ruflo - built for Claude
LangGraph - maximum control
Worth exploring before it becomes the default.
Same pattern works for anything.
Customer support? One triages. One answers. One escalates. One follows up.
Data pipelines? One collects. One cleans. One analyses. One reports.
Break the big task into small, focused agents.