AI Accountability Manifesto

This just is never going to happen. AI that writes directly to assembly or even binary instructions is useless for any enterprise solution. I am a principal software engineer at a fortune 500 so I know more than he does. 1. Companies need a record of who made, reviewed, and deployed the changes. If any change causes an issue they need direct accountability for who caused it. AI generating low level code means it's not human readable and therefore breaks this chain of accountability and review. 2. Measuring productivity in lines of code is absolutely the worst metric imaginable. A vast majority of the time spent building a product is in deriving the requirements for the system, and in the approval processes mentioned above. AI doesn't solve either issue. 3. AI related layoffs are because AI is extremely expensive and will only continue to rise in costs. Increased AI budgets have not in any way correlated with more product being shipped. Therefore AI is useless for companies right now. 4. AI hallucinations are something that cannot be solved. It's part of the process and cannot be overlooked. These systems are non-deterministic and therefore are useless in most applications. 5. AI is not the first attempt to make unskilled human readable code (prompts) that produce functional systems. In all cases this has been tried, it has been abandoned because it just creates even more problems in the long term. Systems become wildly unstable, difficult to maintain, and even harder to add more new functionality to. Effectively every time a change is required they would need to prompt the AI to build a brand new system. This is not cheap. 6. AI cannot create, it can only copy what already exists.

A long anonymous forum post argues that AI-generated low-level code cannot provide accountability, does not solve the real work of software development, is expensive, hallucinates, and cannot create anything genuinely new. The joke is the confidently absolute anti-AI manifesto, posted in response to debates between /pol/ and /g/ about whether AI is useful for software engineering. EN: A long anonymous forum post argues that AI-generated low-level code cannot provide accountability, does not solve the real work of software development, is expensive, hallucinates, and cannot create anything genuinely new. The joke is the confidently absolute anti-AI manifesto, posted in response to debates between /pol/ and /g/ about whether AI is useful for software engineering.

• @gentoomemes / 21084 2026-06-11