AIToday
Large Language ModelsTomasz Tunguz (Theory Ventures)Published: Sep 3, 2026, 13:00 JST2 min read

AI Doesn't Cut Human Work, It Raises the Ceiling

AI Doesn't Cut Human Work, It Raises the Ceiling

3 Key Points

  1. What happened

    The author tested AI writing tools over three years and found AI automates mechanical tasks but does not reduce human editing effort. Across ten published essays, sentence-level edits stayed flat at around 140 per piece, even as draft revisions dropped from 47 to 3.

  2. Why it matters

    AI shifts human effort from fixing broken arguments to line-level craft. The ratio of granular edits per draft version surged from 4.4 to 43.3. Chess data shows a similar pattern: the number of grandmasters grew from 524 in 1993 to roughly 1,750 today, and the top thirty players now average nearly 2,750 rating, up from one player rated 2700+ in 1979.

  3. What to watch

    The test is whether AI’s real payoff is a higher ceiling on output, not fewer human edits—so the outcome hinges on whether users choose to reinvest saved effort into sharper craft rather than accept a pre-AI baseline. Watch whether line-level edits stay flat near 140 per piece, as they have across the dataset.

Ask the AI about this article →

Summaries like this, in your inbox every morning.

Context & Analysis

The author's journey from using prompt templates to a closed-loop taste flywheel reveals that AI's real impact is on the division of labor within a task. Early on, human effort goes into structural triage, but once AI learns house rules, it lands the thesis quickly, freeing humans to focus on rhetorical precision.

This pattern mirrors chess, where stronger engines did not diminish human skill but expanded the elite pool. The Jevons Paradox, which states that efficiency gains increase consumption, is referenced to explain why AI doesn't simply save time; it raises the bar for what's achievable.

The key insight is that mature AI tools don't aim to eliminate human edits. Instead, they transform the writer from a structural mechanic into a sharpshooter, concentrating effort on the highest-value aspects of the work.

FAQ
What happened to the number of draft revisions over time?
A data-heavy post required 47 draft revisions a week ago, while an essay on agent lifespans took only 3. However, sentence-level edits remained flat at 130–140 per post.
How did chess performance change with AI assistance?
In 1979, only one player was rated 2700 or higher. Today, the top thirty players average nearly 2,750. The number of grandmasters grew from 524 in 1993 to roughly 1,750 today.
Tomasz Tunguz (Theory Ventures)Read Original Article

Get the latest Large Language Models news every morning

For example, today's edition would include:

  • Nvidia CEO Huang Declares AGI Has ArrivedYahoo Finance AI · 1h ago
  • Saudi Arabia's HUMAIN aims to be AI 'Switzerland'Semafor Tech · 1h ago
  • Alibaba releases Qwen-Drive 1.0 driving AI modelTHE DECODER · 1h ago

AI-summarized, only the topics you pick — one digest a day via Email, Slack, or Discord.

Free · 30 seconds with Google · unsubscribe anytimeWhat is AIToday? →

Ask AI

Ask AI anything about this article. Q&As are published on this page for other readers too.

Related Articles

Next articleAI Decisions Fail Without Business Context