Google's year in review: areas with research breakthroughs in 2025

Original Article ↗ Hacker News Discussion ↗

DeepMind, Materials, and Non‑AI Science

  • Some wonder when DeepMind will tackle hard physics problems such as room‑temperature superconductors; others note there isn’t enough theory or data yet, but point to DeepMind’s new “automated science lab” focused on materials (superconductors, solar, semiconductors).
  • Quantum computing draws skepticism: several commenters doubt real‑world impact until it can solve non‑toy problems (e.g., factoring larger numbers), while others think it will likely work but lacks a “killer app” and good abstractions/tools.

AI-Centric Framing of Google’s Year

  • Many note the review is overwhelmingly about AI/agents; some see this as over-marketing (“year of agents” when agents mostly do coding), others argue Google has always been an AI-heavy research shop and is now just marketing it.
  • Debate over whether Google “caught up” or leads frontier models:
    • Some say Gemini 3 (especially Flash) beats GPT 5.x on quality/price/speed; others find GPT 5.1/5.2 clearly better for coding and reliability.
    • Multiple users emphasize the “jagged frontier”: models fail at tasks a child could do yet excel at complex coding and analysis.
    • One user’s failed attempt to have multiple LLMs analyze complex, real-world bank statements fuels claims that AI is still “Siri 2.0”; others respond that the right pattern is to have models write deterministic code/tools rather than directly ingesting large structured datasets.

Google’s Research vs. Ad Monopoly and Product Quality

  • Some are impressed by breadth: Nobel‑related quantum work, healthcare, weather models, TPUs, multimodal models, and non‑LLM projects (new OS, languages, hardware).
  • Others counter that, revenue-wise, Google is still fundamentally an advertising company, with ~¾ of income from ads.
  • Strong criticism of Google as a multi‑front monopoly (search, browser, mobile), extracting an “internet tax” via ads and SEO‑driven results, degrading search quality, pushing AMP, and now inserting LLM answers that further starve the open web.
  • Some call for antitrust breakups similar to Bell; others argue users’ demand for “free” services made the ad model inevitable.

Economy, Inequality, and the AI Boom

  • A side discussion disputes the claim that the economy is “tanking”:
    • One camp cites strong GDP growth and consumer spending; another points to rising living costs, unequal wage/price dynamics, high consumer and auto-loan delinquencies, and a K‑shaped recovery skewed by AI and stock gains.
    • Several stress that aggregate metrics (GDP, average debt ratios) can mask severe distributional issues; “the economy” vs. “human wellbeing” are framed as distinct.
    • There is also skepticism about the reliability of official statistics and concern around entry‑level tech job markets.

Science Coverage and AI Saturation

  • Some lament that “breakthrough” lists (Google’s, Science magazine’s) have become almost all AI, crowding out other surprising work.
  • Others argue AI breakthroughs may justifiably dominate because they could soon accelerate progress across all fields.
  • Alternatives like Quanta and independent “breakthrough” compilations are mentioned as better sources for cross‑disciplinary science.