Many Nigerian parents can use WhatsApp, but can't fill a JAMB form, book a train ticket, or tell a scam SMS from a real one. The web is in English. They aren't. So I built Jagora. You talk to it in Yorùbá, Hausa, Igbo, Pidgin or English, and it does the task on the real website for you, out loud. Built solo in 10 days for the Meta Global AI Developer Hackathon with Muse Spark @AIatMeta @cerebral_valley Watch it work 👇

1/3 What if Africa doesn't lack intelligent people? What if we've spent so much time discussing what Africans cannot do that we've stopped asking whether the environments around them actually allow them to succeed? I think we're asking the wrong question. 🧵

Millions of Nigerians can talk, but the internet expects them to read and type in English. So I'm building Jagora for this: a voice-first browser. You speak Yorùbá, Hausa, Igbo, Pidgin or English, and it does the website for you. Built on Muse Spark. Demo next week.

I built a Lagos cooking game from scratch and it is live. 🇳🇬 Bùkà Hustle is a 3D business sim where you start with a roadside bùkà, cook jollof, serve hungry customers, buy water, wash the pot, hire workers and grow into a full restaurant. Watch out for NEPA taking light, and keep your gas filled. You can play it on your phone, tablet or laptop. It is free, with no forced ads and no energy timers. I would love your feedback. Tell me what to add next. Link in the comments.

Creativity at it's peak 🔥 We have cracked 3D animators in Nigeria Animated by @aniegreatman #3d #animation

It's coming together,I taught I would have been able to release the game today ,but unfortunately it's more complex than I taught it would be.

Unto the next

𝗗𝗘𝗩𝗘𝗟𝗢𝗣𝗘𝗥𝗦, 𝗬𝗢𝗨 𝗠𝗜𝗚𝗛𝗧 𝗕𝗘 𝗧𝗘𝗦𝗧𝗜𝗡𝗚 𝗪𝗛𝗘𝗧𝗛𝗘𝗥 𝗬𝗢𝗨𝗥 𝗔𝗜 𝗪𝗢𝗥𝗞𝗦, 𝗪𝗜𝗧𝗛𝗢𝗨𝗧 𝗔𝗖𝗧𝗨𝗔𝗟𝗟𝗬 𝗧𝗘𝗦𝗧𝗜𝗡𝗚 𝗜𝗧. Think about it. Your AI coding agent writes a feature. The tests pass. Your CI pipeline turns green. You merge the code. But what if the tests never checked whether the feature was actually correct? This isn't just a hypothetical concern. A 2026 empirical study examined 86,156 test-file patches from AI-agent-authored pull requests across 2,807 GitHub repositories. 𝗧𝗵𝗲 𝗳𝗶𝗻𝗱𝗶𝗻𝗴? 𝟴𝟬.𝟮% 𝗵𝗮𝗱 𝘄𝗲𝗮𝗸 𝗼𝗿 𝗻𝗼 𝗲𝘅𝗽𝗹𝗶𝗰𝗶𝘁 𝘃𝗲𝗿𝗶𝗳𝗶𝗰𝗮𝘁𝗶𝗼𝗻 𝘀𝗶𝗴𝗻𝗮𝗹𝘀. In other words, having a test file doesn't necessarily mean the code's behaviour was meaningfully verified. And this problem goes beyond software engineering. → An LLM can score highly on a benchmark without demonstrating the capability we think we're measuring. → An AI agent can report that a task is complete while important requirements remain unmet. → An evaluation can produce impressive numbers while overlooking the conditions under which a system fails. These are different problems, but they point to the same question: 𝗔𝗥𝗘 𝗪𝗘 𝗠𝗘𝗔𝗦𝗨𝗥𝗜𝗡𝗚 𝗔𝗖𝗧𝗨𝗔𝗟 𝗖𝗢𝗥𝗥𝗘𝗖𝗧𝗡𝗘𝗦𝗦, 𝗢𝗥 𝗝𝗨𝗦𝗧 𝗦𝗜𝗚𝗡𝗔𝗟𝗦 𝗧𝗛𝗔𝗧 𝗟𝗢𝗢𝗞 𝗟𝗜𝗞𝗘 𝗖𝗢𝗥𝗥𝗘𝗖𝗧𝗡𝗘𝗦𝗦? If you're building with AI, don't stop at asking whether the output passes your checks. Ask whether those checks could catch a convincing but incorrect result. Write tests that should fail when the implementation is wrong. Challenge your evaluation metrics. Test cases the model hasn't seen. Verify claims independently instead of treating the model's confidence as evidence. Because a green checkmark is a result of a test. 𝗜𝘁 𝗶𝘀𝗻'𝘁, 𝗯𝘆 𝗶𝘁𝘀𝗲𝗹𝗳, 𝗽𝗿𝗼𝗼𝗳 𝘁𝗵𝗮𝘁 𝘆𝗼𝘂𝗿 𝗔𝗜 𝘄𝗼𝗿𝗸𝘀. Paper: All Smoke, No Alarm: Oracle Signals in Agent-Authored Test Code (2026).

Now on @itchio ememzyvisuals.itch.io/last-b…

Play free: lastbus.vercel.app/ Headphones. Lights off. Phone in landscape.

11:41 PM. Yaba. Rain. The last danfo to Ikorodu has eight strangers and one empty seat nobody will sit in. LAST BUS: a pioneering Nigerian browser-first supernatural horror experience set in Lagos. Free. No download. Headphones on. Lights off. Play tonight.

Play am here 👇 lagos-hustle.vercel.app Drop your score. I wan see who fit pass Chairman Ajala.

Agbero dey chase me for camera tax 😭 ₦1,700 to pass. Pay, haggle, or run. I ran. Free. No download. Your phone fit play am.

i will win.

Job update: Joined @plotweaverapp as a Machine learning and AI Engineer 🤭

This morning, my mum woke me up and asked me to go pick a particular leaf. She called it “Ẹ̀fọ́ Yánrìn.” I was like… “Which leaf is that?” 😭 She had to show me. Then she started telling me about the nutrients, health benefits, and different ways the leaf is traditionally used. And I was just there thinking: “Wait… THIS leaf??” 😭😂 That moment honestly got me thinking. There is SO much indigenous knowledge sitting inside our parents and grandparents that we are slowly losing. They know the names of leaves and plants in our local languages. They know which ones are vegetables, what they contain, how they're traditionally used, and the cultural knowledge surrounding them. Meanwhile, some of us don't even recognize the plants growing around us. And Ẹ̀fọ́ Yánrìn is just ONE example. There are so many other leaves and plants we haven't even heard of. Some we've probably walked past hundreds of times without knowing their names, nutritional value, cultural significance, or what our ancestors knew about them. And that's the scary part. If the people carrying this knowledge don't pass it down or document it, what happens when they're no longer here? THE KNOWLEDGE DISAPPEARS WITH THEM. And with it, we could lose knowledge about natural resources that have been part of our food, culture, environment, and traditional practices for generations. That's why me and one of my guys are starting to work on a project around this. We're starting with the DATASET before the MODEL. The long-term idea is to build an AI system where you can simply snap a picture of a leaf and get information such as: 🌿 Local/cultural names — Yoruba, Igbo, Hausa, etc. 🔬 Botanical/scientific name 🧪 Known chemical compounds & nutritional composition 🍲 Food/culinary uses 🌍 Cultural significance & traditional uses 📚 Documented scientific research & evidence ⚠️ Safety information — including what is and isn't scientifically established The goal isn't to replace traditional knowledge. And we definitely don't want an AI model making random medical claims. The goal is to DOCUMENT, ORGANIZE, PRESERVE, and make this knowledge accessible. Because think about it… We have plants growing all around us that our ancestors knew how to identify and use. But we're becoming increasingly disconnected from them. What happens 30, 50, or 100 years from now if nobody remembers these names anymore? What if our children grow up surrounded by plants that their grandparents could identify instantly, but they have absolutely no idea what they are? What if we lose knowledge about something potentially valuable simply because nobody documented it? THIS IS WHY I THINK WE NEED TO START NOW. 🌿 We're starting in Nigeria, but the bigger vision is to eventually document indigenous plant knowledge across Africa. And we're starting with the DATASET before the MODEL. We're looking for: 🔬 Botanists 🌿 Ethnobotanists 🧑🏾‍🌾 Farmers 📚 Researchers 👵🏾 Traditional knowledge holders 💻 Developers 📊 Data scientists 📸 Photographers 🌍 Anyone passionate about preserving African indigenous knowledge If you have knowledge about plants and leaves in your community, we'd love to hear from you. If you're a researcher who can help us structure and validate the data, we'd love to collaborate. If you're a developer, data scientist, botanist, photographer, or simply someone who believes this knowledge shouldn't disappear, let's work together. This isn't just about building another AI model. It's about preserving something we already have. BEFORE WE FORGET IT. Let's preserve what we still have. 🌍🌿 If you're interested in contributing or collaborating, reply to this post or DM me.

Most people think AI is about prompting. It isn't. The real competitive advantage is building systems. • Collect quality data. • Fine-tune models for your use case. • Build reliable evaluation pipelines. • Optimize inference for speed and cost. • Deploy, monitor, and iterate continuously. Models are becoming commodities. Infrastructure, data, and execution are what create lasting value. The next generation of AI companies won't win because they have the biggest model. They'll win because they solve real problems better than everyone else. #AI #MachineLearning #LLM #MLOps #ArtificialIntelligence #Tech

Just created the official LinkedIn company page for Axiveri 🚀 Axiveri is focused on building open-weight AI systems for African languages and real local use cases. We’re already shipping: • NaijaVox – multilingual ASR for Nigerian languages • Africlaude – models tuned for African developers & fintech • Open datasets for the community The goal is simple: AI that actually understands us. Follow the page for updates, model releases, and research. Link in comments 👇 Building in public continues.

Nobody sees the 40+ debug cycles, the dependency patching hell, or the training runs Ememzyvisuals bounced between clouds because one kept dying on him. Yoruba tone alone nearly made him shelve this whole release. Shipped it anyway 🔥

WazobiaVoice is out 🔥 5 Nigerian languages, one model. 13 voice personas + full voice cloning from a 5s clip. Quality: • English, Igbo, Hausa: strong • Pidgin: strong • Yoruba: still lagging, VITS fix in progress Nigeria's TTS just got real.