FIELD NOTES ── 2026–2035

Intelligence at dawn

The next decade of AI is rising on the horizon.
This is not a prophecy, it is an observation note.

  1. 01 Agentic AI enters everyday workflows
  2. 02 Multimodal models begin to understand the world
  3. 03 Humanoid robots leave the lab
  4. 04 AI speeds up the discovery of new drugs and materials
  5. 05 Inference costs fall year after year
  6. 06 Global AI governance frameworks take shape

Prologue──Why now

chatbot is just the prologue;
The real change has just begun.

In the past we taught computers "how to do it",
For the next ten years, we just have to tell it"What do you want?"。

At the end of 2022, generative AI will allow everyone to “talk to a machine” for the first time. In just a few years, models have evolved from being able to answer conversations to being able to reason, use tools, and complete tasks that span hours for humans. The pace of change is not linear—the capabilities of each generation of models raise the starting point of the next.

This page has compiled some of the most noteworthy main lines in the next ten years:agents, multimodality, computing power, robotics, scientific research, governance, and human–machine interfaces. Some have already happened and some are on the way. You can think of it as a map looking toward the horizon.

Scenario Deduction─A possible move

Ten years, a possible road

The following is not a prediction, but a scenario deduction based on the current technology curve - the real world will definitely be more tortuous, and it may also be faster.

2026

After the first year of agency

AI walks into the operating system and browser from the chat window and begins to perform multi-step tasks for people. "Let AI run it" has become a daily expression in the workplace.

2027

Multimodal models mature

In this scenario, models combine text, images, audio, and spatial understanding, bringing translation and visual assistance to phones and glasses.

2028

robot inflection point

The outlook imagines wider use of humanoid robots in warehouses and factories, with general-purpose models making new tasks easier to learn.

2030

Science Acceleration

AI is deeply involved in the design of new drugs, batteries and materials. The research and development cycle in some fields has been shortened from ten years to several years, and clinical trials and verification have become new bottlenecks.

2032

Popularization of personal AI

The scenario considers personal agents with long-term memory, and growing public debate over privacy and data sovereignty.

2035

The Era of Shared Intelligence

The division of labor between humans and AI has been rewritten: education teaches "how to ask", medical care moves toward prevention, the threshold for creation disappears, and taste becomes a scarce commodity.

scale──A few numbers worth remembering

The sense of speed of change

10× / 2 years

's cutting-edge model training computing power is much faster than Moore's Law.

-90%+

How much the cost of reasoning with the same capabilities has dropped over several years – wisdom is getting cheaper fast.

hundreds of millions of people

An illustrative scale for weekly generative AI use, continuing to grow.

※ The above is a compilation of the magnitude of public information in the industry and is not an accurate statistics.

Two sides ─ the same coin

Light and Shadow

Light: The magnified person

  • Productivity jumpRepetitive tasks are taken over and people focus on judgment, taste and relationships.
  • professional popularizationMedical consultation, legal assistance, tutoring – services that used to be expensive have become available to everyone.
  • Science Speed UpFor the first time, "too difficult problems" such as climate, disease, and energy have new-level tools.

Shadow: Things that must be faced head-on

  • The erosion of trustDeep fakeness makes "seeing is believing" invalid, and verifying reality becomes the new infrastructure.
  • Labor TransformationSome professions are being restructured, and the speed of transformation tests every society's safety net.
  • alignment problemThe more capable a system is, the more important it is to ensure that its goals are aligned with humans—one of the most important technical questions of the decade.

Final Chapter──Returning to Humanity

After dawn, we decide how the sky looks like

The technology curve only determines “what can be done”.Is it worth doing and who should you do it for?will always be a human choice. Take part in shaping the next decade.

People walking along a riverside promenade in the morning sun
The future is not for prediction.
is ours to build.
中文