← Himilo
Startups

A South African AI coding startup bets the hard part isn't writing code — it's shipping it

The Cape Town central business district. HyperDev, which raised over $1 million in pre-seed funding, operates across South Africa and Europe. (Illustrative)
The Cape Town central business district. HyperDev, which raised over $1 million in pre-seed funding, operates across South Africa and Europe. (Illustrative)Bl1zz4rd-editor via Wikimedia Commons

HyperDev, the generative-AI coding platform chaired by HyperionDev founder Riaz Moola, has raised over $1 million in pre-seed funding — wagering that the real bottleneck for a new generation of builders is turning AI-generated code into software that actually ships.

Every AI coding tool on the market was built on the same quiet assumption: that writing the code is the difficult part. Ask a chatbot for a login page, a checkout flow, a booking form, and it will oblige in seconds. The demo dazzles. Then the builder — often someone who has never opened a terminal — is left holding a wall of generated code and no idea how to turn it into something a real person can use.

HyperDev, a generative-AI software-development platform operating across Europe and South Africa, has raised more than $1 million (about R16 million) in a pre-seed round to attack precisely that gap. According to TechFinancials, the round was backed by a network of venture investors in Europe and the United Kingdom and led by Falk Albers, managing director of RIQ and a general partner at Loom Ventures. The company is chaired by Riaz Moola, the Forbes 30-Under-30 founder of the South African edtech firm HyperionDev — a link confirmed on Moola's own professional profile, where he lists himself as chairman of HyperDev.

The pitch is a contrarian one. "Every AI coding tool on the market was built on the assumption that generating code was the hard part," said Piotr Sobolewski, HyperDev's co-founder and chief technology officer, in comments to TechFinancials. "We built HyperDev because we knew the hard part was what came after, and nobody else seems to be solving that." Sobolewski is a former OpenAI engineer who, the company says, worked on technology behind ChatGPT; Moola previously worked at Google on AI methods used in tools such as Gemini.

That "what came after" is the space between a promising snippet and a deployed, working application — what the industry has taken to calling the last mile of software development. HyperDev's answer is a feature it calls Guided Mode, which walks users through actually using LLM-generated code — LLM being a large language model, the kind of system that powers a modern chatbot — to produce shippable websites, apps and fully deployable software rather than fragments that stall on a laptop. For those who hit a wall the AI can't clear, the platform offers a second escape hatch: the option to hire a human developer to refine and ship the work. It is also wired into the HyperionDev education platform, so a stuck user can be routed toward the specific skill they are missing.

The company's disclosed traction is the part that will make investors lean in — and the part a careful reader should treat as HyperDev's own figures rather than audited fact. It says it reached nearly 100,000 users in under three months, is live across 14 countries, and has seen its user-retention rate roughly double since Guided Mode launched. Retention is the number that matters in this category, and the team knows it. "AI app builders are seeing high user churn, so retention is the metric that matters most," said chief marketing officer Kenne Loubser. In a market where the standard product is a thin layer wrapped around someone else's model — and where users try one, get stuck, and drift to the next — keeping people building is the whole game.

Albers framed his cheque around exactly that combination of proprietary work and a captive audience. "While most vibe coding tools are thin wrappers around third-party LLMs, HyperDev is building proprietary technology that makes code generation actually useful," he said. "That combination of real tech and built-in distribution is rare at any stage — at pre-seed, it's exceptional." The "built-in distribution" is no small thing: HyperionDev has spent more than a decade putting tens of thousands of learners through coding bootcamps run with universities including Stellenbosch, Imperial College London and the University of Chicago. HyperDev launches with a warm pipeline of exactly the people it is built for.

What makes this worth watching from an African vantage point is less the dollar figure — a seven-figure pre-seed is modest against the $1.44 billion the continent's startups raised in the first half of 2026, by TechCabal's count — than the shape of the bet. Africa's tech story has been written overwhelmingly in fintech and, lately, in the scramble to build AI data centres. HyperDev sits somewhere more foundational and more democratic: it is infrastructure for people who want to build software but were never going to be hired as engineers. If the tools that let a shopkeeper in Durban or a student in Nairobi ship a working app are designed, in part, from South Africa — by a team that already understands how non-engineers actually learn to code — the second-order effect compounds quietly. Every builder who ships instead of stalling becomes a small node of local software supply, in a continent that has long imported most of its digital tooling.

The caveats are real. The AI coding field is ferociously crowded and moving weekly; a pre-seed buys runway, not a moat, and the retention gains will have to survive contact with a much larger user base. But the thesis is a sharp one, and it is refreshingly unglamorous. The winners in this wave may not be whoever generates the cleverest code. They may be whoever gets the most ordinary people all the way to something that works.

HyperDev's wager is that generating code is the easy part; turning it into deployable software is the last mile it aims to close. (Illustrative)
HyperDev's wager is that generating code is the easy part; turning it into deployable software is the last mile it aims to close. (Illustrative)Martin Vorel via Wikimedia Commons
WhatsAppXLinkedIn