💡 I document CQRS, Event Sourcing and DDD with real code on real stacks.
Latest: CQRS: separating the command from the query
👉 joemugen.medium.com/2868b1fe…
All my articles in this thread 👇
#CQRS #EventSourcing
1
2
45
Replying to @ItsChrisCorado @sakthisankara
Support messages are evidence, not roadmap items. The feature request is usually a proposed fix.
Look for the workflow or business rule that keeps failing, then decide whether one change solves the repeated problem
1
2
19
The worst response would be to hire only people who are already ready-made.
We need to get better at spotting potential when the obvious signals are easy to produce.
I wrote the longer version here:
joemugen.medium.com/d2654352…
6
AI can make the visible part of development look easy.
It does not tell you whether someone understands the problem, sees the rule a change might break, asks for missing context or knows when they do not have enough evidence to decide.
1
3
Real problems rarely begin with the right framework.
They begin with a vague request, an unwritten rule, a customer using the product in an unexpected way, or a support team that learned to work around the system.
1
3
What I look for in a junior is intelligence, curiosity, mindset and potential.
What have they tried to build on their own? Which problems did they hit? What did they learn? Which questions do they ask when they do not have the answer?
1
3
Junior hiring is getting harder. AI is not the only reason.
But when components, tests, migrations and configs can be generated in seconds, matching technical keywords becomes an even weaker signal.
2
7
I never believed in hiring junior engineers for their stack.
React, Node, TypeScript, Docker: useful foundations. Not a reason to hire someone.
With AI, that distinction matters even more.
1
34
Replying to @zachlloydtweets
2-3x cost makes the split important: how much comes from first pass findings, rereview after fixes, and false positives?
That would show whether the extra spend buys new signal or just more churn
9
Replying to @Hi_Mrinal
Microservices make a boundary executable. They don't create autonomy by themselves
If the decision, the data and the release cadence are not independent, you just moved the coordination cost onto the network where it becomes failure modes you cannot see in one transaction
24
Replying to @nyxienyxie
Persisting the wait is the easy part. Hard part is knowing whether the agent is still allowed to resume when the answer finally comes back
I'd persist the decision, the owner and the acceptance criteria too. Otherwise you get a durable process that can keep doing the wrong thing
1
15
👉 The full post, with the code: joemugen.medium.com/cqrs-in-…
It's a demo, so it shows a lot at once. A real project takes what its problem asks for.
#Laravel #CQRS
1
15
And to the Laravel crowd who thinks it's not "Laravel Way" enough: don't lose the thread.
The series stays on Laravel and on this repo and we'll get there step by step. I'm going to get you all to make peace with it :)
1
10
💡 The price never goes through the kitchen.
The dining room keeps its TableOrderPlaced fact at home and publishes a contract with the table, the visit and the lines, never the price.
The kitchen can't leak a price it never received.
1
64
Each context picks its own CQRS level:
🟢 Floor plan: level 1, same transaction as the table
🟡 To-serve list: level 2, a queued projector that can lag (the pass decides, not the list)
🔴 Kitchen ticket: level 3, a stream, no kitchen_tickets table
1
18
In a restaurant, "order" means 3 things. The dining room sells the plate, the kitchen cooks it, the till charges it.
Same plate, 3 classes, and none of them knows the other 2.
Part 2 of my CQRS series, in Laravel 👇
1
1
73
Replying to @joelapache @AkubueAlexander
Designing the schema "for the future" is the trap IMHO. The business will change shapes you can't guess. I'd rather keep write decisions as they happened and rebuild any report when the question changes. The tables that answer today's questions stay disposable
1
2
20
Replying to @AkubueAlexander
One rule accountants already follow is worth enforcing in the schema. A posted journal entry is never updated or deleted, a mistake gets a reversing entry. Then any balance can be recomputed from the lines and every number traced back to its source
2
2
25
Replying to @Vikram6072 @piyushgarg_dev
"ProductPriceUpdated" says what changed in the table. The business usually knows why: "PromotionApplied", "PriceIncreased", "PriceTypoCorrected". That's where events start answering questions a CRUD history can't
And what's a "simple application" or a "simple business"?
1
13
