Claim◆Article

Stagnation in long-form nonfiction writing is an alarming sign for expectations that AI will autonomously solve grand, open scientific problems soon.

Current models struggle to organize and compellingly present even well-established science. Until they overcome this limitation, scientific progress from LLMs is likely to remain concentrated on low-hanging fruit and connections between distant fields rather than revolutionary insight. ✦ AI generated

the author · Interconnects · 2026-08-12 · original ↗

Share card ↓Export .md Export .txt

Models being stagnant in long-form, non-fiction writing should be alarming to those reliant on models autonomously solving grand, open science problems in the near future. The models today struggle to organize and compellingly present some of the most established science in their area. This seems like a natural prerequisite that we should expect the models to master before they can solve broad, open-ended problems on their own. Until this is solved, the progress of LLMs for science will look closer to solving low-hanging fruit and merging distant connections across fields, rather than any sort of revolutionary insight.

Read full article ↗excerpt · fair-use quotation

✳ LLMs

Around this claim

This moment responds to

supports → The best textbooks will remain heavily crafted by humans for at least two to five years because AI currently saves only 10–20% of the effort and excels at transforming existing knowledge rather than creating its core insight.the author · Interconnects supports → Today's LLMs increase entropy rather than compressing and organizing knowledge in long-form nonfiction, so they will remain dependent on human guidance.the author · Interconnects rebuts → AI has two unprecedented qualities compared to every prior technology: autonomy (you can give it a job without specifying how to do it, and it figures it out) and a degree of unpredictability about the future that no earlier innovation, including the internet, ever had.Howard Marks · My First Million supports → LLMs are good at checking or producing individual components and making small edits, but they compound errors when asked to organize and connect those components into an entire technical chapter.the author · Interconnects supports → Models have improved rapidly in coding, mathematics, search, and research, but writing well has barely advanced because it is a difficult, relatively orthogonal skill with inadequate specialized training data.the author · Interconnects supports → AI can provide substantial value in nonfiction writing when an expert supervises it closely, but it should mainly handle repetitive, verifiable, or editorial tasks rather than the parts that convey a work's story and intellectual soul.the author · Interconnects