The Fundamental Development Gap Map is excellent. AI has moved fast enough since v1.0 that I think it’s worth an update.
You’ve said the map “isn’t comprehensive – at all” and asked for contributions. I got excited and built one with Claude Code.
There are two places I would start.
- Break each gap into its critical path, and work out whether and how each capability accelerates each step.
- Identify how AI can and does affect each gap, and how mature that AI is.
I did this for one gap: research publishingHuman-checked
Doing and publishing research is expensive and subject to structural roadblocks, traced from a finished draft to a credited contribution.
| Step | AI reaches it | Gap Map capabilities | Elapsed |
|---|---|---|---|
| 01Production and draftingAuthor time. | works now | none | no published figureNo published figure. Author time sits outside this chain's axis and no one reports a median for it. |
| 02Submission and desk screeningEditor triage time. | works now | none | 119 daysone figure, steps 2–5Submission to acceptance, which brackets steps 2 to 5 together rather than any one of them. Median 119 days across ophthalmology journals in 2020, IQR 83-168. Carried on step 2 because that is where the span opens, and it is not a measurement of desk screening alone. |
| 03Reviewer recruitment and matchingcarries the costWillingness to serve, and matching. | 2–5 years | none | Inside the 119-day span carried on step 2. |
| 04Review judgmentcarries the costIrreducible subjectivity in ranking work that is above the bar. | 2–5 years | 1 capability | Inside the 119-day span carried on step 2. |
| 05Editorial decisionEditor labour, and accountability for the decision. | — no | none | Inside the 119-day span carried on step 2. |
| 06DisseminationArticle processing charges and platform cost. | — no | 3 capabilities | 30 daysAcceptance to first online release. Median 30 days, IQR 10-71, same 2020 ophthalmology cohort. |
| 07Credit and legitimacycarries the costHiring, tenure and funding committees decide what counts. | — no | 2 capabilities | no published figureNo published figure, and probably not measurable. The quantity is how long institutions take to count a new kind of work. |
Three takeaways
- Production and drafting is where AI has had the biggest impact, and where it will keep having one. The saving is real and large, and it has not arrived as cheaper publishing. Submissions rose 42% after ChatGPT’s release — against a 20% bump during COVID — and the work that came off authors landed downstream on volunteer editors doing desk screening. Organization Science, 2026, ~7,000 manuscripts. This has happened before: arXiv made dissemination free more than thirty years ago and neither subscription prices nor the time to publish fell, because what journals sell is not distribution.
- Review is a clear current blocker. AI can act on it, and it may stay hard anyway.
- Credit and legitimacy is the biggest challenge, and the one AI is least able to move. It is also the last step and it has no published measure.
This chain step by step, plus a second gap (telescopes) that Claude did on its own
Next steps
- Expand to the rest of the map. Cover all 103 gaps: identify the critical paths, identify how much AI is affecting each one, and propose new capabilities to close them based on that analysis.
- Verify with experts. One gap here has been read by a person. The rest is a model’s first pass and is labelled that way throughout.
What I’d like from you
I’d like to start collaborating. Does the approach resonate: expanding gaps into critical paths, and identifying where AI is having an impact and where it could?
The labels on all 103 gaps are a model’s judgment, each with a written rationale and a confidence flag, and no person has read them.AI only Every gap · the eight kinds of work · method and disagreement rates · what this does not do