A former AI advocate details disillusionment with large language models, citing reliability issues, regression between versions, broken enterprise workflows, and lack of accountability in AI systems deployed across critical industries.
I used to be all-in on large language models. Built automations, client tools, business workflows..... hell, entire processes around GPT and similar systems. I thought we were seeing the dawn of a new era. I was wrong. Nothing is reliable. If your workflow needs any real accuracy, consistency, or reproducibility, these models are a liability. Ask the same question twice and get two different answers. Small updates silently break entire chains of logic. It’s like building on quicksand. That old line, *“this is the worst it’ll ever be,”* is bullshit. GPT-4o workflows that ran perfectly are now useless on GPT-5.5. Things regress, behaviors shift, context windows hallucinate. You can’t version-lock intelligence that doesn’t actually understand what it’s doing. The time and money that go into “guardrailing,” “safety layers,” and “compliance” dwarfs just paying a human to do the work correctly. Worse, the safeguards rarely even function. You end up debugging an AI that won’t admit it’s wrong, wrapped in another AI that can’t explain why. And then there’s the hype machine. Every company is tripping over itself to bolt “AI-powered” onto products that don’t need it. Copilot, ChatGPT, Gemini - they’re all mediocre at best, and big tech is starting to realize it. Real productivity gains are vanishingly rare. The MASSIVE reluctance of the business world to say something is simply due to embarrassment of admission. CEO's are literally scrambling to re-hire, or pay people like ME to come in and fix some truly horrific situations. (I am too busy fixing all of the broken shit on my end to even think about having the time to do this for others. But the phone calls and emails are piling up. Other consultants I speak with say the same thing. Copilot easily being the most requested to be fixed). Random, unreliable, and broken systems with zero audit requirements in the US. And I mean ZERO accountability. The amount of plausible deniability massive companies have to purposely or inadvertently harm people is overwhelming. These systems now influence hiring, pay, healthcare, credit, and legal outcomes without auditability, transparency, or regulation. I work with these tools every day, and have from jump. I am confident we are at minimum in a largely stalled performance drought, and at worst, witnessing the absolute floors starting to crumble.
An opinion piece arguing that AI systems, especially large language models, are fundamentally bullshitters because they generate plausible but false information without understanding or intent to deceive.
An analysis of why teams quietly abandon AI tools due to broken trust, arguing that the real problem is not model quality but the lack of trust architecture—designing workflows that clearly indicate when AI output is reliable and when it needs verification.
The article discusses the industry consensus that AI is becoming extremely capable but still faces reliability issues for high-stakes tasks, emphasizing that current systems optimize for plausibility rather than guaranteed truth, and that the path forward involves layered verification systems rather than a single perfect model.
The author expresses disappointment in AI progress, arguing that despite years of development and massive spending, large language models still struggle with basic reasoning, referencing an Apple paper that exposes fundamental flaws. They question whether the hype around superintelligence is misguided.
The article argues that companies are overinvested in AI intelligence (model capability) while neglecting crucial runtime layers for authority, accountability, and reality representation, leading to potential failures when AI acts within institutions.