Loading
60 days from outline to defensible draft.
Thesis writing is where AI-drafted work is most tempting and most catastrophic. Tempting because the draft is long and the deadline is finite. Catastrophic because your committee will read it — and any experienced reader can tell AI prose from your prose, especially when the argument doesn't cohere the way your prior work does.
Brainback's role in a thesis is bounded and specific: Learn mode coaches your argument structure and your reading synthesis; Efficiency mode drafts non-thesis writing (cover letters, funding proposals, admin emails). The thesis body itself is yours. Every paragraph.
This playbook is a 60-day writing sprint from you-have-an-outline to you-have-a-defensible-draft. It assumes the research is mostly done and the argument shape is roughly decided. If either isn't true, back up one phase before starting.
Each row is a day (or block) with a specific focus and a specific Brainback move. Rehearsable. Adjustable.
| when | focus | brainback move | hours |
|---|---|---|---|
| Days 1-3 | Outline into chapter-level scenes | Learn mode on your outline: 'given my thesis, what's the strongest counter-argument in this chapter, and where does the chapter address it?' The tutor pushes on structure, not prose. | 8 hrs |
| Days 4-14 | Chapter 1 (intro) + literature review draft | Write daily. 500-1000 words per day. Learn mode reviews structure at end of each session; you never let it write the paragraph. | 4 hrs/day |
| Days 15-30 | Chapter 2 + 3 (methods + core argument) | Same pace. Explain-back on your core argument at day 22 — if you can't explain it in 3 minutes, revise before continuing. | — |
| Days 31-42 | Chapter 4 + 5 (results + discussion) | Same pace. Sitting with hard-to-write sections is the whole game; the tutor helps you get unstuck, doesn't write around the stuck point. | — |
| Days 43-52 | Chapter 6 (conclusion) + full re-read | Full re-read of your own draft, noting weak transitions and unsupported claims. Learn mode on the 3-5 weakest ones. | — |
| Days 53-58 | Revision pass · argument coherence | Focus on the through-line: does chapter 6's conclusion actually follow from chapter 1's setup? Learn mode is a great third-party reader for this. | — |
| Days 59-60 | Format, cite, submit | Efficiency mode for citation formatting, bibliography checks, submission-portal admin. Learn mode never touches these — they're not graded reasoning. | — |
Same seven surfaces are always on. For this specific playbook, these are the ones you’ll touch most.
Argument structure coaching. Reading synthesis. The tutor asks Socratic questions about your draft; it doesn't write your draft.
Your core argument should explain-back at a 4/5 or higher. If it doesn't, the argument itself needs work, not just the prose.
Cover letters, funding proposals, submission-portal admin, citation checks. Not the thesis body.
Every playbook has anti-patterns that feel productive but produce nothing. Skipping these is worth as much as executing the plan.
because · Detectable to experienced readers. Also cognitively bypasses the argument-refinement that writing produces. Both problems compound.
because · You'll never feel ready. Write the ugly first draft. Refine in later passes. Perfectionism is procrastination in disguise.
because · If you can't explain your own core argument in 3 minutes at day 22, chapters 4-6 will drift. Better to spend day 22 revising the outline than to discover the problem at day 50.
The general writing-in-Brainback workflow. Read the writing subject page →
Abstract: no — it's part of the graded thesis and any experienced reader will spot AI prose. Acknowledgments: yes, if you want; those aren't graded reasoning.
Grammar-level editing tools are fine. Rewriting tools (that produce alternative paragraphs) blur into ghostwriting; avoid them.
When you can explain-back the argument at 4+/5 for every chapter, AND when you can name what the strongest counter-argument is against your thesis. If you don't know the counter-argument, your committee will — and they'll use it.
Same workflow. The rule 'don't AI-draft graded prose' is universal; scale up or down based on the stakes.
→thesis writing sprint
Free tier is enough to execute most of these. Upgrade if the recall queue gets serious.