Key Takeaways

  • Writing splits into two distinct categories: writing as reporting and writing as thinking.
  • Writing as reporting (status updates, meeting summaries, launch notes) should be delegated to language models to save time.
  • Long documents no longer prove deep thought, because anyone can generate a ten-page brief in seconds using AI.
  • Internal product momentum at OpenAI increasingly relies on interactive prototypes and mocks over traditional specification docs.
  • Tara Seshan uses The 70% Document Completion Rule to leave rough edges in proposals so collaborators can actively shape the outcome.

The 70% Document Completion Rule

Writing a document to completion before showing it to teammates creates unintended friction. As Seshan explains, “Very few great people want to interact with a perfectly polished finished idea. With a perfectly polished idea, their new ideas just bounce off of it versus something that has more crags and more rough edges that they too can polish with you together.”

To build consensus without sacrificing critical thinking, use this two-step process:

Step 1: Solo Drafting to 70%

Write the initial brief or strategic proposal to 70% completion yourself without automated generation, structuring the core logic, hypothesis, and trade-offs while leaving intentional rough edges.

Step 2: Stakeholder Stress-Testing (70% to 100%)

Shop the document around to key decision-makers and contributors, encouraging them to attack ideas, poke holes, and co-author the remaining 30% of the solution together.

When This Works (and When It Doesn't)

This rule applies when building consensus for new product areas or strategic pivots. If you drop a pristine, 20-page document onto a senior engineering lead, they will either rubber-stamp it without reading or resent being handed a fait accompli. Sharing a proposal with clear open questions invites genuine intellectual partnership.

It fails in high-stakes operational execution or compliance. If you are writing a security protocol, API contract, or legal filing, leaving 30% unfinished creates chaos. In those situations, you need crisp, fully resolved specifications before anyone touches code.

This division matches Seshan's approach to AI at OpenAI. She separates writing as reporting from writing as thinking. “Writing as reporting I happily automate or I use the models all the time to make that as simple as it can be. But writing as thinking is something I never will automate.” Status emails and release logs should be handed to an LLM. Strategy briefs and core problem definitions require raw human wrestling.

Generating a long document takes five seconds with modern tools. Because of that, length has lost its value as proof of effort. “A long doc is not a signal that you thought through something because you can easily produce a long doc that indicates that you haven't,” Seshan notes. To demonstrate rigor today, build working prototypes or draft tight, human-written theses that welcome critique.

What to Do With This

Take the strategy doc you are currently writing for next quarter. Stop before you solve every edge case or polish the final appendix.

Highlight the three biggest unsolved questions in yellow. Send the doc to your engineering lead and designer with a short note: "Here is where my head is at on this product direction. The structure is 70% done, but the trade-offs in section 3 are open. Tear this apart with me this afternoon."