The limits of vibe coding
Nice piece on the "dangers" of vibe coding by John Mc Keown for Tekedia: 'When AI is asked about the number of Rs in “cranberry”, “elderberry” or “barberry”, different answers might be given each time, ranging from two to four, and only sometimes three. This inconsistency isn’t a glitch, but rather a fundamental aspect of the transformer architecture that underlies language models.
'Unlike humans, these models don’t count letters; instead, they rely on probability distributions to sample tokens. So, an inquiry about the number of Rs in a word prompts the system to predict the typical token sequence that follows that question pattern, without examining the word letter by letter.'
Many years ago in my 3rd year in FUT Owerri, Prof J.I. Ejimanya who taught us EEE 304 Telecommunication Principles began that class by writing on the board:
Register for Tekedia Mini-MBA edition 19 (Feb 9 – May 2, 2026): big discounts for early bird.
Tekedia AI in Business Masterclass opens registrations.
Join Tekedia Capital Syndicate and co-invest in great global startups.
Register for Tekedia AI Lab: From Technical Design to Deployment (next edition begins Jan 24 2026).
You received two messages
1. “$4 00,0 0”
2. “Trav l to Lagos tomo row”
He then asked, what do you think the sender was trying to communicate? Quickly, he noted that not many people will care in #2, but #1 was complicated. From there, he emphasized the importance of reliable communication systems.
That takes me to the points this article is making. Can vibe coding deliver when needed?
