No. 19
2026-08-06
the AI pulse — brewed, not raw
 AI CRITIQUE
Dört adam gülümsüyor ve koyu gri bir duvarın önünde yakın duruyor, her biri farklı renkli gömlek giyiyor.

Google's top AI scientists are leaving the company to start a new startup called Discovery Loop.

Read the article    English (automatic translation)

Summary

Google's four most senior AI scientists—Jeff Dean, Sanjay Ghemawat, Oriol Vinyals, and Quoc Le—are leaving the company after approximately 27 years. They are founding Discovery Loop, a startup aimed at establishing an autonomous 'scientific method loop' (propose → apply → evaluate → repeat) capable of driving breakthroughs in drug discovery, chip design, biology, and machine learning.

Why it matters

This represents a rare and high-impact loss of talent at a critical juncture in Google's defense of its position in the AI model race. The four departing individuals held roles in units such as Google Brain, DeepMind, Gemini, and core search infrastructure. The thesis of positioning AI not merely as a research tool but as an autonomous researcher is one of the most significant short-term bets in applied AI. The first company to bring such a product to market could fundamentally alter how science and engineering progress. Google's decision to invest in and provide computational power to the very team it just lost creates a complex landscape regarding intellectual property, competition, and dependency. This situation will likely shape the sector's competitive dynamics for years to come.

Highlights

  • The company's core product is an 'automated experimentation loop': AI proposes an experiment, writes the code to run it, evaluates the results, and repeats the process thousands of times. This cycle is completed before generalizing to new domains.
  • Google is simultaneously a founding investor, the cloud and computing partner for the first year, and the employer these individuals just left. In this triple entanglement, a clear intellectual property wall has not been defined.
  • The company currently has no hires, office, or product. The CEO was selected by saying, 'everyone was looking at me.' Discovery Loop is, for now, just four friends and a pitch deck.

Critical take

The article's admission that 'the mission is not new'—noting that Dario, Sam, and Jensen have also described AI as a scientist—undermines the hype. However, the text still reads like a hagiography ('GOAT', 'supernovae in the AI firmament'). More importantly, Discovery Loop has no engineers, infrastructure, or product. The company is essentially a bet based on four individuals' past resumes rather than proven competence. The article's abrupt cutoff mid-sentence suggests that even the reporter could not find concrete news material.

Bu kritik, yerel yapay zeka modeli Qwen3.8-27B tarafından yazılmıştır. Haberi ben seçtim; kritik metnini yayımlanmadan önce tek tek denetlemiyorum.This critique was written by the local AI model Qwen3.8-27B. I chose the story; I do not review each critique before it is published.
Back to the issue