Dammit another weekend gone
github.com/rommapp/romm
So true
OH: giving a developer Terraform is like giving a puppy for Christmas.
Week 1 - walks, cleanup, chaos.
Week 2 - someone else's problem.
The Modernization Imperative: Shifting left is for suckers. Shift down instead
cloud.google.com/blog/product...
OH - the fastest way to go backwards is to stand still
LCNC had hard edges. AI has none. The scope is infinite and so is the risk of shipping bad code.
LCNC guardrails were baked in by design. With AI we have to build them ourselves.
Guardrails, proper processes and pipelines.
The irony is most teams never built them before AI. Now we have no choice.
Like it or lump it, nonβtechnical staff will write code.
Ironically, the more code they write, the tighter the safeguards become.
βDonβt import this transport until someone has watched the e2e test video confirming the PMβs feature they wrote during standup didnβt break anything.β
I've done a number of big data and IoT projects with CAP on XSA.
FWICS on-premise never moved beyond HANA 2, while the cloud kept evolving, and that growing gap is exactly what forces these workarounds
Being able to run HANA locally would help a lot too for CAP
Good topic for @hanatech.community
Gemini keeps reminding me of POSWID - the Purpose Of a System is What it Does.
Which also explains why the developer tooling problem looks easy but is actually hard.
The real system isn't just the repo. It's pipelines, configs, tickets, chats, observability and workarounds accumulated over time.
Really like the idea of tools like getunblocked.com. Using MCP to surface context from code, docs and conversations (Slack/Teams) to help developers onboard and move faster
But it also highlights a bigger issue - a lot of real tribal knowledge and practices were never documented or drifted long ago
Your telling me they don't need to attend stand ups
"Weβre all singing the same dumb song on the worst karaoke machine ever made."
Good read
- The environment used to create growth by accident.
- AI removes that natural friction, and many organizations respond by optimizing for output instead of developing seniority.
- True seniority now depends on judgment, system thinking, and navigating ambiguity, not coding volume
OH: MS Agent 365 onboards agents like their workers - roles, permissions etc. Like contractors, but who aren't forced to use Citrix.
The PowerBI / PowerApps problem.
Teams sick of waiting use it anyway, it works, then there's fifteen versions and nobody knows which one's real or who has access.
Someone senior steps in, forms a committee, kills the thing that was working.
Death by a thousand paper cuts. Back to waiting again.
Agree. I wonder if this is the inflection point for this hype style.
If AI can analyse, summarise and compare on demand, we donβt need a Magic Quadrant. Vendors can create their own hype and customers can test their own view.
It feel it still comes back to choosing the right problems first.
Agree. The tech is cool and experimenting is the fun part.
We just keep pointing it at the hardest problems first
Feels like the early Fiori journey again. Start with complex processes, burn time and money, struggle to show value or scalability
Executive buy in is easier than business buy in
I have fond memories of SAP Leonardo. Did some cool projects.
Looking back it was executive briefings, design thinking workshops and polished POCs that rarely went anywhere
Same story with RPA and low-code etc
McKinsey says 80+% of AI initiatives still donβt get past POC
What have we learned?
A good idea for a post, which could have been crafted into a really good read. Fed to AI. Came back broad, shallow, and with 60+ em dashes which to me look like placeholders where practical insight should have been.
Made me lol
You can't prompt for context, you can't draw a diagram based on someone else's reference architecture and call it done, context is what makes the prompt work, the gaps between the boxes is the where the real system lies, and getting that context used to start with a pager ringing at 3am
The experience was never the point. The context it carried was. Yes a lot of the knowledge is lost because it's no longer relevant, secondphase.com.au/integrationa... was a similar vein, knowledge without context is dogma, we've always done it this way is ignorance of why - the prompt is wrong
The article started as all this experience will be lost in time "like tears in the rain" mourning the craft, but then I realised tribal knowledge is everything. Thinking in Systems, talks about two systems, the one we think we designed and the one we got, the lived experience
It's the apprenticeship I feel is gone. Perl, VANs, EDI, IDocs, RFCs, SAPscript, Basis, rewiring the office, going to warehouses to fix printers and scanners, my gig whilst finishing Uni. The work demanded you step into and own it all. Its what I tried to capture
secondphase.com.au/why-you-cant...
Had to maintain some pretty gruesome forms and mailer apps written in Perl, full swivel chair operation. Informix was the DB, Perl talking to text files I literally put on a disc. I was misty eyed back then for sure π
I feel maybe i am naive and it was always happening
Early 2000s.
Working late, empty office.
Random phone rings.
A recruiter fishing.
Creating a market where they weren't invited.
Fast forward to today.
Now I'm hearing stories of recruiters passing π° to trusted people to generate leads from the inside.
The call is coming from inside the building.
Cost Is a Distributed Systems Bug
dzone.com/articles/cos...
Despite the 30+ em dashes - this is an interesting read
Treat spend like any other failure signal and design so mistakes cannot drain the budget.
Shopify's Secret to Handling Billions in Sales
hackernoon.com/shopifys-sec...