🔥 The DB Grill 🔥

Ah, yes. I've just had the... privilege... of perusing this announcement from the "Tinybird" collective. It is, one must admit, a truly breathtaking document. A monument to the boundless optimism of those who believe enthusiasm can serve as a substitute for a rigorous, formal education in computer science.

One must applaud the sheer audacity of a "chat-first interface" for a database. What a truly magnificent solution to a problem that was solved, and solved elegantly, by Dr. Codd in 1970. To think, we spent decades building upon the bedrock of relational algebra and the unambiguous precision of formal query languages, only to arrive at the digital equivalent of asking a librarian for "that blue book I saw last week" and hoping for the best. The sheer, unadulterated ambiguity is a masterstroke of post-modernist data retrieval. It’s as if they decided the entire point of a query language—its mathematical certainty—was an inconvenient bug rather than its most vital feature.

And the engine of this... contraption? A "Tinybird AI to generate exactly the SQL you need." How utterly wonderful! A statistical parlor trick that vomits out SQL, likely with all the elegance and structural integrity of a house of cards in a hurricane. I find myself morbidly curious. Does this "AI" understand the subtle yet crucial difference between 3NF and BCNF? Does it weep at the sight of a denormalized table? I suspect not. Clearly, Codd's fifth rule—the comprehensive data sublanguage rule—is now merely a suggestion, a quaint artifact from an era when we expected practitioners to actually understand their tools.

"...Time Series is back as a first-class citizen..."

One is simply overcome with admiration. They've rediscovered the timestamp! What an innovation! It's almost as if a properly modeled relational schema with appropriate indexing couldn't have handled this all along. But no, we must bolt on a "first-class citizen," presumably because the first-year-level data modeling was too much of a bother.

But my favorite part, the true chef's kiss of this whole affair, is the triumphant return of "Free queries return for raw SQL access." It's a tacit admission of defeat, is it not? A glorious little escape hatch.

• "When our delightful conversational bauble inevitably fails to comprehend a non-trivial request..."
• "When the statistical noise generator produces a query that performs a full table scan on a petabyte of data..."
• "When you actually need a predictable, correct, and performant result..."

"...please, by all means, use the grown-up tool we tried so desperately to hide from you." It’s utterly charming in its transparency.

I watch this with the detached amusement of a tenured professor observing a freshman's attempt to prove P=NP with a flowchart. They speak of conversations and AI, yet I hear only the ghosts of lost transactions and data anomalies. One shudders to think what their conception of the ACID properties must be. Atomicity is probably just a friendly suggestion. As for the CAP theorem, I imagine they believe it's a choice between "Chatbots, Availability, and Profitability."

Mark my words. This will all end in tears, data corruption, and a series of increasingly panicked blog posts about "unexpected data drift." They are building a cathedral on a swamp, a beautiful, glistening facade that will inevitably sink into a mire of inconsistency and regret. It's a tragedy, really. But a predictable one. Clearly, they've never read Stonebraker's seminal work. Then again, who in "industry" reads the papers anymore? They're far too busy having conversations with their data.