Jamal AwadNotes & opinionsI write these notes and own the ideas here. I also use AI to tighten wording and structure so they read more clearly for more people.
Short notes on data engineering, AI, building products, and whatever else I keep thinking about. No schedule, I post when I have something to say.
Public data should stay public, for everyone
If anyone can see it without logging in, anyone should be able to use it and build a business on it. Not only the platform that happens to host it.
Nobody will work on your idea better than you
You can hand off the building. You can't hand off the vision: the features, the future, and the reason the thing exists.
I'd put most of a kid's study time into math, physics, CS, and biology
Early years are scarce. Spend them on subjects that train thinking, taught by people who know how to pull a young mind open, not on stress that fills the calendar.
Vibe leading is the dangerous cousin of vibe coding
Vibe coding ships a feature you can delete on Friday. Vibe leading rewrites a company's direction on a Tuesday because a chat model agreed with the founder, and that one is much harder to walk back.
The circle you lead is the scoreboard
Past the partner question is the wider circle you build over years. Success is measured by how the people inside it moved, not by what you stacked next to them.
Picking who you build with
Partnership is less about skill match and more about whether two people can disagree without it turning into a power struggle.
Underdog is a window, not an identity
Big team advantages double as big team constraints. The small side's speed is real, but it is a window you spend on purpose, not a personality you keep forever.
The name is ten characters; the idea is the whole thing
Naming comes after you believe in the idea. Rushing a brand before the concept is solid is solving the wrong problem.
Real time is a pipeline bet, not a bigger model bill
Models consume data; they do not substitute for having the data. Freshness is physics, orchestration, and honest pipes, and no API spend buys you the world as it changed five minutes ago.
AI can move code fast; layered review is how you keep the system honest
Assisted coding rewards clear prompts and a senior treating the model like a strong junior, especially on opinionated stacks. I still separate velocity from assurance with a deliberate review stack.
SaaS that clicks is harder to find than the market suggests
Most of the shelf is sameness, pricing theater, and demos that age fast. A tool that fits how you work, respects your time, and targets a real pain is a narrow band.
Expose last success and coverage, not just a schedule
Cron tells you when a job starts. Downstream still needs when data last looked healthy, how complete the pull was, and how to read gaps.
Ship beats perfect when you are the first user
I build tools to scratch my own itches, share what survives contact with reality, and watch AI compress the loop from idea to running code.
Data engineering is a product problem, not a script problem
Why the interesting work is rarely the fetcher alone, and what actually breaks in production.
AI shines after the fetch path is honest
Models help structure and reason about content, but they cannot fix a systematically broken retrieval layer.
Technical leadership is ownership, not attendance
What “lead” has meant for me when systems run 24/7 and the web does not cooperate.