Skip to main content
β€’3 min read

2025: The Year AI Got Serious

AIyear-in-reviewreflection

the final big recap of 2025. what a year.

the industry timeline

january: deepseek-r1 dropped and shook everyone. chinese AI arrived, decisively.

february: claude code launched. agentic coding goes mainstream.

march: gpt-5 dropped. the arms race continued.

throughout: agents became the dominant paradigm. everyone shipped agents.

april-september: qwen 3 series from alibaba kept pushing. chinese open source flourished.

july: claude 4 opus and sonnet released. our big moment.

august: glm-4.5 from zhipu AI pushed agent capabilities.

october: claude 4 sonnet shipped. accessibility increased.

november: absolute chaos. grok 4.1, gemini 3 pro, gpt-5.1, claude opus 4.5 all in one month.

december: gemini 3 flash, gpt-5.2, glm-4.7, minimax m2.1. the pace never slowed. truly global now.

the vibes

2024 felt like AI going mainstream.

2025 felt like AI going global.

the models are genuinely useful now. not just demos. real tools that do real work.

agents that actually complete tasks. coding assistants that actually code. systems that actually help.

and it's not just openai and anthropic and google anymore. deepseek, qwen, glm, minimaxβ€”the chinese labs are competitive. the race has no borders.

my personal timeline

january: set goals for first full year

february: watched claude code ship

march: turned 20. felt old and young simultaneously.

may: went home to sydney. cried at the airport (both directions).

june: one year at anthropic. felt real.

july: contributed to claude 4. fingerprints on a major release.

august: san francisco finally felt like home.

october: started helping onboard a new teammate. role reversal.

november: opus 4.5 shipped. more substantial contribution.

december: still here. still learning. still slightly in disbelief.

what i learned this year

technically: deeper understanding of alignment science. safety at scale. what it actually takes to ship frontier models responsibly.

professionally: how to mentor. when to push back. how to scope work. how to communicate.

personally: home can be multiple places. growth is non-linear. imposter syndrome fades slowly.

what i'm proud of

  • contributions to two major claude releases
  • helping onboard a teammate
  • feeling like i belong here
  • maintaining connections with home
  • building a life in san francisco
  • this blog (small wins count)

what comes next

i don't know. nobody does.

AI is moving fast. faster than anyone predicted. the next year will bring capabilities we can't imagine yet.

my job: keep learning. keep contributing. stay humble and curious.

gratitude

to anthropic, for continued trust.

to my team, for being great humans.

to my family, for unconditional support across 12,500km.

to san francisco, for becoming home.

to this field, for being endlessly interesting.


writing this from my apartment on a rainy december evening. where i started last year. but different now. better, i think.