@ field note

Cursor’s Africa/EMEA opening, and why it needs an FDE on the ground

Seyi Adeleke · Senior SWE, Stitch · payments infrastructure, Lagos

Cursor is not losing this market. It has not entered it yet.

@ // the opportunity

Africa has roughly 4.7M developers, and between 2019 and 2024 that base grew 21% / year, the fastest rate of any continent. South Africa, Egypt, and Nigeria each now exceed 500,000. This is the fastest-growing developer population on earth, it is young, and its adoption of AI-native workflows is still wide open. Whoever earns trust here first will hold it for a decade.

Source: BCG, “Develop the Developers,” 2026

@ // the three gaps

Three things stand between Cursor and this region, drawn from public strategy documents and from what I see firsthand at Stitch and across the engineering teams I work alongside.

gap_01

The work here is co-build, and nobody is in the room.

One pan-African telco is collapsing a fragmented estate of apps into a single platform while re-platforming a mobile money business that spans 14 markets onto a mini-app architecture. Tier-one banks are modernising cores against regulatory deadlines. These are the same migration executed dozens of times with per-market variation, which is what long-running agents are good at and hand-rolled engineering is worst at. The reflex here is to add systems-integrator headcount, so cost scales linearly with every new market. Nobody is proposing the alternative.

what an FDE does

Pick one service, one market. Build the migration agent alongside their platform team: it reads the current API surface and the target contract, generates the adapter, backfills tests to the coverage gate, and opens a PR a human reviews. Encode what it cannot resolve as rules, so market N+1 costs less than market N. Their team owns it by the end.

gap_02

Cursor looks expensive because nobody taught the cost model.

The most common objection I hear is that Cursor is expensive. It rarely is a pricing problem. Teams route every request to a frontier model when Auto would do, close to a 10x difference on the same task. They pull whole-codebase context where a scoped @ reference saves 60–80%. They open a fresh thread for every question and lose the ~90% cache discount. Nobody taught them the cost model, so the bill reads as the product’s fault.

what an FDE does

Instrument spend per team, set a routing policy, get .cursorignore and the rules files right. Then hand engineering leadership a cost-per-outcome number they can defend in their own budget meetings.

gap_03

Trust dies in security review, for the wrong reason.

In regulated fintech, “we would be sending our source code to an American AI company” ends the conversation. But Nigeria’s NDPA and South Africa’s POPIA regulate personal data, and source code mostly is not personal data. The review stalls because nobody separates the two questions: is personal data actually in scope, and what stops it reaching the context window. US-only data residency does not settle Nigeria either, since Nigeria does not treat the US as adequate.

what an FDE does

Split the review. Scope the PII question honestly, then close it with .cursorignore, Privacy Mode enforced org-wide, model allow-lists and an evidence pack for the DPO. For Nigeria the mechanism is standard contractual clauses in the DPA, not new infrastructure. Then sit in the review and clear it.

01 / 03

@ // why me

I am a Senior SWE embedded in this ecosystem. I have driven AI-native workflow adoption inside a regulated fintech, and I can walk into these rooms because I am already in them.

Proof of work: I co-founded citydistro.africa and built its entire stack with Cursor.

@ // contact

Seyi Adeleke

Senior Software Engineer · Stitch

Lagos, Nigeria · EMEA

adelekeadetokunbo01@gmail.com

linkedin.com/in/seyiadeleke

github.com/seyi-adeleke