@GDGBabcock From your experience in industry, how much does a master’s degree actually impact credibility and employability for software engineers compared to strong projects and real world experience?
@GDGBabcock 1. What mistakes do self-taught or university developers commonly make when trying to break into industry?
2. How much does personal branding or posting online actually help in opportunities?
@GDGBabcock With AI coding agents writing more of the code, should earlycareer developers still spend time learning syntax and frameworks deeply, or focus more on core concepts and problem solving?
Feels easy to become overreliant on AI without noticing the long-term effect on actual skills
Paused this project for exams
Came back after exams and finally shipped V1 today
Still a lot to improve, but getting something stable and usable into production feels different from endlessly rebuilding features locally.
Now comes the real test: users.
https://t.co/5bjkLWeHHB
One thing this ERP project taught me:
In real projects, “small” changes are never really small.
One feature touches inventory, analytics, orders, caching, and things you didn’t even expect.
That shift in thinking changed how I approach backend systems.
https://t.co/5bjkLWe9S3
Four weeks ago this was just a bunch of CRUD APIs sitting in Swagger.
Today, V1 is live.
Had to pause development for semester exams, then came back to finish the project.
Building something beyond tutorial-sized apps changes how you think very quickly.
https://t.co/5bjkLWeHHB
Hard truth: You can't polish a system while you're still adding legs to it.
This week is for tightening flows and stress-testing order paths. V1 drops by the end of the week.
Focus on the foundation first. The features can come later.
#SoftwareArchitecture#SoftwareEngineering
What feature should I add next? is often the wrong question.
Ask instead: What logic cannot afford to fail?
For Manifest V1, I’m locking features to harden the core. Predictable behavior beats a bloated feature set every time. Build for pressure, not just for looks.
#Deploy
An ERP isn’t a collection of pages it’s a state machine.
Many devs (myself included) start by building "Ship Order" buttons. That’s the easy part.
The real value is in the state transitions. If your logic isn't solid, your UI is just lying to the user.
#SystemDesign#ERP
Had an AI suggest a “clean” solution recently.
Looked solid.
Completely ignored a dependency edge case.
If I didn’t understand the flow first,
I would’ve shipped a serious bug.
AI doesn’t remove mistakes.
It scales them.
How I use AI without getting shallow:
Ask for the map, not the car
If I don’t understand the logic, I don’t generate anything.
Set constraints early
No constraints = hidden technical debt.
Rewrite core logic
If I can’t explain it, I delete and rewrite it.
Use AI as a critic
React Query was lowkey hurting performance.
Defaults were refetching on every click.
Set proper staleTime → navigation feels instant now.
Also split heavy chunks (charts, UI libs) with Vite.
Less JS upfront = faster load.
People call AI “vibe coding.”
That’s not the real problem.
The problem is workflow.
Once things get even slightly complex state, data flow, dependencies
you can’t just “generate and hope.”
That breaks fast.
If you don’t own the architecture, the AI is just guessing for you.