there’s a sweet spot for thinking effort that makes models not over-engineer everything, I swear Composer 2.5 outputs more readable code than Opus 4.7 or GPT 5.5 sometimes.
@larsencc optimizing for the quick, “Lovable”-like apps is what brings the easy VC money though, people never consider scale, security and complexity until it hits them in the face with a sledgehammer.
fun to see vibecoded apps getting 0 traction as it turns out they solved no problems.
@aspiwack sadly, the cost of onboarding and training juniors is not really worth it for start-ups and SME, spoke to multiple founders/leads about this. the consequence is that no one will grow to be a “senior” from the generations that are in uni right now.
@leetllm you miss the fact that the LLM is confident in writing SQL because it has seen probably 10,000x more SQL than actual ORM usage... you can probably infer the exact multiplier by running a GitHub search... the LLM does not magically transpile all the ORM code into SQL...
most of the consequences of supply chain attacks could be avoided if people would develop inside docker containers but vibecoding has eroded good security practices sadly…
@levelsio@Airbnb I’ve always had good experiences with Airbnb, they refunded multiple trips where there were ant infestations or misleading pictures, didn’t even need to argue with the support, just refunded me straight away and gave me a 10% off voucher on the same day, probably 30m after.
@taimurabdaal I think flight itineraries still make sense for long and complex work trips or holidays in multiple places but booking a single flight is usually a 2 min process.