Kwame Oteng Appiah-Nti

Programmers write code. Engineers design systems. - Engineeredweb.com: @mattfarina.

New article: a visual tour of recent LLM architecture advances, from Gemma 4 to DeepSeek V4. I focus on long-context efficiency tweaks like KV sharing, per-layer embeddings, layer-wise attention budgets, compressed attention, and mHC. Link: magazine.sebastianraschka.com/p/recent-devel…

Highlights of the new proposal

@NITAGhana I'm not sure why you're so intent on coming for the tech industry. Now you want to restrict private organizations from hiring us unless we get a government certification? What have we done to you? Seriously. You keep doing this. It simply cannot be accidental.

@nyankahmensah @TheDumbTechGuy @koboateng @Yaw_Ridwan @NITAGhana 🤣

@TheDumbTechGuy @koboateng @Yaw_Ridwan @NITAGhana Meaning I can't be a tech startup founder if I don't have certification 😂💔

A moment of stillness by the water, with Chongqing’s glittering skyline as the backdrop, blending the city’s vibrant energy with a gentle evening calm.🤍🗺️✨ #Chongqing #Cityscape #Riverside #UrbanView 📸langgejiaoniyongshoujipairenxiang

I still keep saying this. We handed over a site to Huawei and they put their product there while we worked with them in parallel for 2 years. In those two years, there was ZERO downtime on their platform because the Chinese were ALWAYS working and proactive. They never rest. To compete, you must have the same work ethic.

SNAPCHAT PAID $150,000,000 FOR LOOKSERY - STARTUP IN DEEP LEARNING COMPUTER VISION This 1-hour Stanford lecture on "DL for Computer Vision" will teach you how to build same project from scratch. Bookmark it & watch today. Stanford's full course (19 lectures) on Deep Learning for Computer Vision below ↓

Atlassian just reported $1.79B in quarterly revenue and serves 350,000+ customers then fired the engineer who built their infrastructure He shares the whole thing a breakdown of Atlassian’s playbook: > Envoy over enterprise load balancers
> sidecars for auth + logging + rate limits
> DynamoDB + SQS
> automated VM deployments

@TheDumbTechGuy @hassan_codes I agree with you Boss. I know you can make it happen.

Guys. Who knows how we can get a press conference organized? It's time to talk to this government about its plans for the tech industry.

@black_kenpachi @TheDumbTechGuy @GTgotmotionn 🤣

@TheDumbTechGuy @GTgotmotionn The name sef de worry me Aza lol

GPU shortage is worse than ever. H100s cost more today than they did 3 years ago, and you cannot get them on-demand. The big AI labs have locked up most of the supply for years. I’m worried university researchers and individual developers simply won’t be able to get GPUs.

very interesting and offbeat paper from hunyan llm team where they introduce a third reference model in the standard OPD objective, thus re-writing it like the implicit reward equation in DPO, which allows dense supervision. They then introduce the generalised OPD objective: G-OPD = lambda * log (pi_t / pi_ref) + (1 - lambda) * (KL(pi_theta/pi_ref) interesting thing here is you can have reward interpolation with 0 < lambda < 1 and with lambda > 1 you can have the student surpass the teacher and the reference model, and since you don't have a restriction on the choice of reference model, you can do multi teacher on policy distillation. They show really cool results with the same.

a new Agentic model that can run on a single consumer GPU at home: ByteDance Seed OSS 36B > very strong at coding > excellent at multi-turn tool calling & Agentic tasks > 500k context window as i have been saying, Bytedance is Tier S

@Patriot5715 @0bin_Cyber @TheAhmadOsman @Sureshsinghn @grok Interesting

@0bin_Cyber @TheAhmadOsman @Sureshsinghn @grok What you want to really do is train your own LLM. I've got Grok and ChatGpt revealing the secrets of the cosmos simply by training them on valid laws and not half-baked ones.

My house has 33 GPUs. > 21x RTX 3090s > 4x RTX 4090s > 4x RTX 5090s > 4x Tenstorrent Blackhole p150a Before AGI arrives: Acquire GPUs. Go into debt if you must. But whatever you do, secure the GPUs.

Many Ghanaians heard about the Daniel Ofori vs Ecobank case, but few understood how an ordinary share sale turned into one of Ghana’s longest commercial court battles. This visual breaks down the full story from the original CAL Bank share transaction in 2008 to the Supreme Court rulings that shaped the final outcome 18 years later. Read more here: finexinsights.com/post/daniel-of… Visual inspired by an original post by: @HamdawaySayibu #Ecobank #DanielOfori #Court #Ghana

Many people are hearing about the Daniel Ofori vs Ecobank case, but don’t fully understand how the whole issue started and how it ended. Here’s the simple breakdown from the beginning to today: In 2008, Ghanaian businessman Daniel Ofori sold about 14.3 million CAL Bank shares to another businessman, William Oppong-Bio. The transaction was handled through Databank Brokerage while Ecobank Ghana acted as the settlement/payment bank. The shares were worth about GH¢13.7 million. According to Daniel Ofori, after the shares had already been transferred, Ecobank failed to complete payment after the Bank of Ghana and SEC raised concerns about the transaction. Part of the money, about GH¢6.16 million, was then allegedly placed into an investment arrangement with Ecobank at an agreed interest rate of 30%. This is where the real issue began. Daniel Ofori later sued Ecobank, arguing that: 1. The shares had already been sold and transferred 2. Therefore he deserved payment 3. Ecobank also owed him the agreed investment interest Ecobank fought the case for years, arguing that: - the transaction became problematic after regulatory intervention - there were inconsistencies in the documents - and Daniel Ofori was still receiving dividends on the shares after supposedly selling them One major question people asked was: “If he sold the shares, why was he still receiving dividends?” Ecobank used this point to challenge the entire claim and later even alleged document alteration and fraud. But the courts repeatedly ruled in Daniel Ofori’s favour. The biggest turning point came in 2018 when the Supreme Court ruled that Daniel Ofori was entitled to the money and the agreed 30% interest. After that, another battle started: How should the interest be calculated? Simple interest? Yearly compound interest? Or monthly compound interest? That calculation changed everything. Because when 30% interest compounds over many years, the amount becomes extremely huge. Ecobank kept challenging the judgment and tried several legal attempts to overturn or reduce the amount. But in 2026, the Supreme Court finally upheld the earlier rulings and confirmed that Daniel Ofori was entitled to the compounded interest. This officially ended one of Ghana’s biggest and longest commercial court battles. Charley this case will probably be studied in Ghana’s banking and legal sectors for many years 😁 Follow for more @HamdawaySayibu

Google Cloud AI engineer just showed how they go from idea to deployed app at Google in 30-minutes using Claude. 26-minutes. free. by Google AI team. one person + Claude + Google Cloud = a full engineering org running on a laptop. worth more than any $500 vibe-coding course.