We benchmarked 10 LLMs on writing infrastructure code: 90 OpenTofu modules for AWS, GCP and Azure.
28% failed a mandatory security item. 0 of 30 S3 modules produced working access logs. Top four within 3 points at a 9x cost spread.
Rankings and code: https://t.co/j2dTLmlHfc
@FredLambert Homeboy couldโve made an F150 competitor, same price range, same target customer and probably crushed, but decided to make a truck for people that donโt need trucks at a price that people that need a truck canโt afford