Google’s internal promotion system is being blamed for pushing employees toward launching new products instead of maintaining existing ones, according to tech writer Gergely Orosz. He said work tied to keeping current systems running or helping users migrate brings little value in performance reviews, putting the spotlight on a familiar Silicon Valley pattern: Promotion-Driven Development, or PDD.
Orosz, author of The Pragmatic Engineer and a former employee at Uber and Microsoft, argued that Google engineers have stronger incentives to build replacements than to improve what already exists. The result, in his view, is a repeated cycle where similar products appear one after another, while older versions are displaced before they have fully settled. Google has launched around 20 communication products over the years, a number often cited by critics as evidence of duplicated effort across teams.
Gemini pricing and tool overlap draw criticism
Developer Nathan Clark laid out how that pattern appears across Google’s AI stack. On the model side, Gemini has been split into Pro, Ultra, Flash, and other variants. The latest Gemini 3.5 Flash is priced at 5.5 times the previous Flash version and 75% above Gemini 3.1 Pro, making the “Flash” label less associated with low-cost, fast access than before.
The product layer is also divided. Gemini App and AI Studio coexist, while Workspace enterprise accounts and Google One AI consumer accounts remain separate, leaving the same user with two disconnected Gemini experiences across work and home. At the tooling level, Google now offers Spark, Jules, and Antigravity 2.0, with substantial overlap in scope. The open-source Gemini CLI, released under Apache 2.0 and written in TypeScript, is set to be discontinued on June 18 and replaced by the closed-source Antigravity CLI. Developers have criticized that move as shifting migration costs onto the community.
After Google I/O 2026, the lineup got even larger
Google I/O 2026 ended last week with the company introducing Gemini 3.5 Flash, Gemini Omni, Spark, Antigravity 2.0, a Googlebook laptop, and an AI agent mode for Search. The reaction from developers was blunt: there are now even more products to sort through. In video, the relationship between Flow and Veo remains unclear, while Gemini Omni, internally called NanoBanana, spans both generation and editing. Image generation has been folded into Gemini, while search and research functions are split between AI Mode and NotebookLM.
Orosz has also pointed to a structural issue beyond product naming. If AWS or Azure suffers an outage, Amazon or Microsoft takes a direct financial or customer hit. If GCP runs into trouble, the effect on Google’s core search advertising business is less immediate. In that setup, he argues, internal incentives around reliability are weaker than at its cloud rivals. His criticism is not that Google lacks engineering talent, but that the company’s evaluation system keeps rewarding fragmentation.