Welcome to Transcedentrepreneurship! If you’re keen on Tech, Product and Business like me, you’re in the right place! No fluff, just my philosophical blend with a pinch of quirky allegories!
P.S. I’m preparing a free workshop on May 15th, check it out!
https://www.tisacode.com/startup-workshop
> Yoo, take it easy
> Put that down (the panic, the cynicism, something else? — all of it.)
> Allow me to elaborate…
> And maybe, reframe a few things.
• • •
> AI does magic when it comes to abstraction and naming things- one of the hardest tasks in software.
> And with the rich palette of IDEs and tools, allowing anyone to produce code, from PMs, marketing associates, to any developer, is sky really the limit? Or have we just moved the limit to somewhere invisible — like comprehension, architecture, or ethics?
> It all sounds amazing
>And on the other side, it sounds terrible- it sounds like layoffs
> However…
> If we were to zoom in, a whole new universe would appear before us
> It’s a universe where the IDE is your copilot, where product specs talk back, where codebases explain themselves like seasoned mentors…
> Now, let’s analyse a potential scenario where two people with different backgrounds try to solve a tech problem using LLM
> For example, let’s imagine we need to build a system that lets users upload CSVs of customer data, processes them (asynchronously), and then shows reports of any errors and imported rows. This should work for large files and be secure.
> An experienced dev’s prompt would potentially look like this:
"We need a CSV ingestion pipeline. Frontend should support chunked uploads or use something like tus.io for large files. On the backend, files should be validated and queued for async processing — preferably using a job queue (e.g., BullMQ, Sidekiq). We’ll need schema validation per row, and error logging tied back to the original line number. For the report, show processed vs. failed rows with detailed messages. S3 for storage, signed URLs for access. Also, input sanitisation and access control are non-negotiable."
> In contrast, e.g. a marketing manager with no coding experience whatsoever could prompt it like:
(equally valid, just shaped by intuition and storytelling)
“Okay, picture this: someone drops a huge spreadsheet into our system — like a big ol’ dump of their contacts. The system should go, ‘Cool, I got this!’, and quietly work in the background. If something’s wrong, like a weird email or missing name, tell them exactly where it happened — row 47, column B or whatever. Then show them a nice summary: ‘Hey, 982 rows imported, 18 had issues.’ Basically, magic, but trustworthy.”
> Guess whose solution would, by any standard, be better, more future-proof, and ultimately come as cost-effective
> Two prompts produce two different outcomes
> And I hope you, too, measure success by outcomes
> What’s the difference between the two prompts, you might ask
> They both share the same idea. The first one is just more techy.
> Let the “AI” handle it (Use this new game-changing LLM from planet Mars)
> However, that techiness is the essence
> Let’s get back to digging deeper
> For any system to exist, some fundamental laws have to allow it
> No system emerges from chaos without order beneath.
> In the context of software, there is one specific law I want to highlight:
> Context is the king
> Context is the a priori
> Context isn’t just surrounding text. It’s your data model, your user expectations, your infrastructure, your edge cases, your constraints — all of it.
> For LLMs data is crucial
> Good code is scarce
> Software engineering was a popular career choice
> Quality is for discussion…
> You can hear this everywhere: Garbage in, garbage out
> A cliché, but only half-right
> What it should be: poor context in, garbage out
> Garbage out isn’t just a bug — it’s a mirror.
> Context increases the value and relevance of the result produced by an LLM
> And the lack of context directly impacts our P&L
> Misunderstanding is expensive.
> Additionally, that means that you, my software aficionados (if you’re reading this), are safe
> …kinda…
> That safety is directly proportional to your relationship with the vast spectrum of software engineering concepts, practices, and theories, plus the ‘soft’ part.
> So the question is: Are you deepening that relationship? Are you just coding, or are you cultivating fluency across the stack — human and technical?
> Communication. Empathy. Systems thinking. Meta-awareness.
> Maybe the future isn’t “code” as we know it. Maybe it’s logic as language, architecture as design, orchestration as creativity.
> Engineering teams are getting smaller.
> Many will need to rethink their careers
> Productivity is skyrocketing, on the other hand
> Opportunities are many, so are the concerns
> It’s always about the perspective
>I hope we good now, or at least a bit more equipped to surf this wave — not just survive it.
> I’d like to hear your thoughts in the comments
> Until next time,
> Vibe coding is the future, but context is still the king.
Cheers,
DB