February 14, 2026
When vibes ship, apps slip
Do Not Outsource Judgement
Engineers rage over “vibe‑coded” mega commits as blame game explodes
TLDR: An engineer warns: AI can help write code, but humans still own the results—and over‑trust is the real danger. The community split between panic, tough‑love accountability, and blaming leadership, with stories of massive AI‑written code dumps and even AI‑generated defenses fueling a fiery debate over who owns the risk and why it matters.
Engineer Dan Crews lit up the dev world with a blunt PSA: AI is a power tool, not a replacement for your judgment. He says the real bubble isn’t AI—it’s over‑trust—and warns that slick, confident outputs can hide subtle mistakes. His nightmare scenario? Junior devs look 10x faster while reviewers crawl 100x slower cleaning up “vibe‑coded” messes. Translation for non‑tech folks: shiny robot helper writes code fast, but the adults still have to check the homework.
The comments? Absolutely on fire. One camp is pure tough love: user wellf says a 15,000‑line “vibe” code dump should be ignored and the culture fixed—up to firing if needed. Another vibe is sheer doom, with am17an’s deadpan “we’re cooked” becoming the thread’s catchphrase. Others got philosophical: myvehicle cites a scholar’s difference between judgment (human context) and reckoning (cold calculation), arguing that AI is great at the latter but not the former. Then came the contrarians: deserts says pushing risk upward isn’t always bad and blames sloppy code on bad hiring and leadership. The spiciest tale? dgxyz claims teammates worship “machine gods,” shipping trash and even pasting AI‑written defenses with “reasoning at the level of an elderly rat.” Chaos, memes, and stunned silence ensued.
Key Points
- •AI-assisted tools can speed development but do not replace engineer responsibility for shipped code.
- •Committing code implies understanding, thorough review, and confidence in correctness, security, and maintainability.
- •AI increases the need for diligence because its outputs may appear correct while being subtly wrong.
- •The real “AI bubble” is over-trust: treating generated code as fine, skipping deep reviews, and offloading judgment.
- •AI can create speed mismatches where contributors appear faster while reviewers slow dramatically to ensure quality.