melbeesgames

I believe that all of this popular knowledge applies:

a) To let the fox guard the hen house

b) To quarterback and catch the ball

c) To be judge and jury

Segregation of duties should be the basic

Creating solutions, processes, connectors, CLIs, and automations to improve my productivity. I believe that 20% of my time with AI is like improving the other 80%.

AI can make a person 10 times more productive, but it cannot replace 10 people. Different people bring knowledge, experiences, and ideas that are enhanced by AI, but are not replaced by AI.

Most layoffs justified by AI are unfounded. Cuts are for balancing the books, not for proven productivity gains with AI. I believe in and am betting on AI, but it doesn't yet replace teams. But it will.

Just as every office has a computer, every corporate document uses Word, and every data grouping/analysis goes to Excel, AI is the tool of the moment; everyone will use it, and those who don't are stuck with the typewriter.

Regarding internal apps, the big problem isn't creating/building them, it's maintaining them, scaling them, and providing support. In the coming times, we'll see many companies suffering from orphaned, vibe-coded apps that no longer work.

AI doesn't make games good or bad. People make games good and bad.

People need to understand that AI is dominating the corporate world; all areas, not just software development, take advantage of AI. Denying AI in the corporate world means losing competitiveness. You can't fight against the modernization of tools. And game development is no different; it will accelerate and improve games, but it will also cause a flood of flop games. Game design, good ideas, creativity, common sense, care/attention, make a good game. AI only accelerates, for better or for worse.

I don't think cutting is the solution; prioritizing based on value is. Create a live roadmap of value and deliver as soon as possible, collect feedback, revise the roadmap/backlog, deliver again and repeat. Pilot, scale and kill/sunset works better than ever at AI speed.

3 points: 1. Organize your repository/project: separate specs, plans, runbooks, code, config/settings, data, ... and architecture. Each with an index generated by a deterministic tool.

2. Define the standard architecture and don't let agents deviate from it without your approval. Also with deterministic control.

3. Use different models to validate the plan and then the code. If model A creates the plan/code, use model B to review it.