Feynman-style Letter: On the Agentification of Scientific Research

After reading the roughly 100-page survey Agentification of Scientific Research (2026.05), I feel the ivory tower that has been locked inside the Ivy League for centuries is being forcibly broken open by a digital mob carrying servers.

To show why AI is no longer content with being a "retriever," let's talk about "doing experiments."

1. Current State: The PhD Student "Drowning" in Vast Literature

The current research process is like an extremely painful single-player parkour game.

• Pain point: A PhD student spends three years reading literature, finding gaps, and building an experimental platform, only to tear everything down because of a mis-tuned parameter. Research is slow because the three physical stages — "proposing hypotheses," "validating hypotheses," and "writing reports" — are severely disconnected. Human stamina is mercilessly drained in those gaps.

2. Agentification: The Chief Scientist with a "Thousand-Person Army"

This survey announces a new era: research is no longer the exclusive domain of carbon-based life — it is being thoroughly "agentified."

It outlines a deeply cyberpunk picture:

• A full-pipeline physical closed loop: We now have "literature agents" dedicated to reading papers, "hypothesis agents" for writing conjectures, "experiment agents" that operate robots, and "peer-review agents" for reviewing.
• A symphony on the blackboard (Multi-Agent Collaboration): These agents are not isolated. They argue and compromise on a shared digital blackboard. A hypothesis agent proposes an idea about a new material; the experiment agent immediately runs it ten thousand times in a molecular dynamics simulator, finds a flaw, and throws the data back for revision.
• Compute is truth: These millisecond-scale, cloud-based high-frequency collisions of thought mean a biological conjecture that once took ten years to validate can be decisively proven over a weekend of idle server capacity.

3. A Feynman-Style Judgment: Science Is "the Integral of Trial-and-Error Frequency"

A so-called "genius" is simply someone who runs trial-and-error in their head a hundred times faster than ordinary people.

Agentification tells us: future Nobel Prizes will belong not only to the humans who propose directions, but also to the agent swarms that iterate on themselves at the highest throughput across boundless solution spaces.

When an AI Agent can go from fetching its first arXiv paper to commanding a robotic arm to synthesize a new drug, that is no longer a victory of artificial intelligence — it is a "genetic mutation" in the paradigm of natural science research.

Takeaway

When facing an unknown interdisciplinary problem, stop trying to brute-force it with one person's brain capacity.

Go build your own "digital research group."

If you can get five AIs speaking different professional jargon to argue day and night in your backend, every spark from their friction may light up a patch of darkness at the boundary of human knowledge.

---

*Note: this is an English translation of a Chinese forum post; the referenced survey "Agentification of Scientific Research (2026.05)" is cited as described by the original author.*