π Nathan Benaich released the State of AI Report 2026 - the ninth annual report on what happened in AI over the past year.
I've been following this report since its first release in 2018: each time it's several hundred slides with a detailed analysis of research, industry, security, geopolitics, and money in AI - without marketing and with real numbers, one of the few sources where you can really see the whole picture over 12 months, not just the news feed.
Five things I highlighted for myself in this edition:
- The frontier race is now a competition between three labs - Anthropic, OpenAI, and Google, and the leader changes depending on which benchmark you look at.
- Non-technical professions - lawyers, sales, recruiting, marketing - are now adopting AI tools faster than developers.
- AI is already building AI: by August 2026, Claude was performing 26% of R&D work on Anthropic's own models, compared to less than 1% in February.
- Inference costs are decreasing by 47% per quarter at the same benchmark levels since 2023, and the combined annual revenue of OpenAI and Anthropic reached $105 billion by the end of summer, up from $30 billion at the beginning of the year.
- Physical AI and biotech have moved from demonstrations to results: the Skild S1 robot handles 66% of unfamiliar tasks without retraining weights, and AI-developed drugs like GB-0895 from Generate:Biomedicines have reached the third phase of clinical trials.
These are five points out of one and a half hundred slides - if you're interested in the AI industry as a whole, rather than individual news, the report is worth reading in full.
https://www.stateof.ai/State-of-AI-Report-2026.pdf
https://www.youtube.com/watch?v=FBfNXq_R3oo
#ai@rvnikita_blog #stateofai@rvnikita_blog #nathan_benaich@rvnikita_blog #anthropic@rvnikita_blog #openai@rvnikita_blog
