Abiyu Giday

Here is the latest analysis on #Africa's Electric Power generation. #rstats Read more about it here: linkedin.com/feed/update/ur…

🎯💫

#rstats 👇💫

@kyle_e_walker 🎯

Starting to see more sensible AI takes on here. A few I've noticed (that I agree with): - In many cases, a deterministic automation is superior to an LLM, and you should know when to use each - You'll often have more success with 1-3 focused agent sessions than with a swarm of agents you can't keep track of - Not everyone wants to build their own software

Interesting 👇

🔍 Distintas formas de visualizar los resultados de un análisis de clúster jerárquico con R. 📍Imagen basada en el dataset mtcars + clustering jerárquico con Ward. #dataviz #datascience #stats #analytics #rstats #programming #cluster #kmeans #ML #machinelearning #cienciadedatos

@DavidDack 🎯

I don’t think people really get how different elite marathoners are. A sub-2 marathon is 4:34 pace… for the full 26.2 miles. Most runners could train for years and still not hold that pace for a 10K, let alone a marathon. At that level, it’s not just discipline anymore. It’s talent, genetics, years of training, insane pain tolerance, perfect execution… and honestly, bodies that seem built for this in a way the rest of us just aren’t. Different game entirely.

[Tip de R] · [Paquete 📦] · duckh3: Integrá el sistema de indexación H3 de Uber con R y DuckDB para análisis geoespacial rápido y escalable. ¿Trabajás con datos geoespaciales gigantes y sentís que R se arrastra? ¿O quizás necesitás la eficiencia de H3 y DuckDB sin dejar tu ambiente de R? El paquete duckh3 es la solución que conecta el potente sistema de indexación jerárquica H3 de Uber y la velocidad de DuckDB directamente en R, sin complicarte la vida. Te permite analizar y manipular grandes datasets espaciales (y no espaciales) de forma súper eficiente. ✔️ Análisis H3 escalable Te da funciones rápidas y eficientes para manejar grandes volúmenes de datos espaciales y no espaciales usando el sistema de indexación H3. ✔️ Integración fluida Conecta el poder de la extensión H3 de DuckDB con tu ecosistema R (dplyr, sf, duckspatial), para que no tengas que cambiar de herramienta. ✔️ Rendimiento optimizado Acelerá tus consultas y manipulaciones de datos geoespaciales aprovechando la potencia analítica de DuckDB sin salir de R. Es compatible con data frames, tibbles y objetos dbplyr. 💡 Tip Si ya usás {duckspatial}, te va a resultar súper familiar, ya que {duckh3} sigue una API similar. Es ideal para cuando buscás rendimiento a la hora de trabajar con celdas H3 y grandes bases de datos directamente desde R. 🌐 github.com/Cidree/duckh3 ✍🏼 Cidree #RStats #RStatsES #Rtips #DataScience

@jack as way ahead, as alwsys, way ahead of the redt of the field 👇💫

TG got it done! Another WR🙌👏

@WorldAthletics @fitsumaregaa @KejelchaYomif @jacobkiplimo2 Sub 2 is just insane 🤯 for a marathon!

HISTORY HAS BEEN MADE 🫨 Sabastian Sawe becomes the first person ever to break the 2-hour barrier in official race conditions, storming to a historic 1:59:30‼️ @KejelchaYomif, on his marathon debut, also breaks 2 hours with a stunning 1:59:41 and @jacobkiplimo2 clocks 2:00:28, also faster than the previous world record 😤

Oh, nothing new today .. era of agentic commences and is accessible to all who could 🙂

🚨 April was absolutely insane for AI releases... • Claude Opus 4.7 (Anthropic) • GPT-5.5 (OpenAI) • DeepSeek V4 (DeepSeek) • Xiaomi Mimi V2.5 • Qwen3.6-Plus (Alibaba) • GLM-5.1 (Zhipu) • Muse Spark (Meta) • Qwen3.6-35B-A3B (Alibaba) • Grok 4.3 Beta (xAI) • Qwen3.6-Max (Alibaba) • Kimi K2.6 (Moonshot) • Qwen3.6-27B (Alibaba)

deepseek-v4-flash is now available on Ollama's cloud! Hosted in the US. Try it with Claude Code: ollama launch claude --model deepseek-v4-flash:cloud Try it with OpenClaw: ollama launch openclaw --model deepseek-v4-flash:cloud Try it with Hermes: ollama launch hermes --model deepseek-v4-flash:cloud Try it with chat: ollama run deepseek-v4-flash:cloud (DeepSeek V4 Pro is coming shortly) 🧵

All things being equal ... the operative words that resonate... "cost-effective 1M contex" ... 👇

🎉 splitGraph is now on CRAN! Leakage hides in structure: repeated subjects, shared batches, cross-study overlap. Turn metadata into a typed dependency graph → validate → derive defensible splits. #rstats #DataScience cran.r-project.org/web/packages/s…

Most AI tools feel generic because they don’t know your data. I spent time turning 3,300 scattered datasets into a structured catalog with proper tags and rich context. The difference between generic answers and actually useful recommendations is huge. Read for more on linkedIN 👉: linkedin.com/feed/update/ur… #rstats #RAG #NLP #LLM #Analyticw

@dekhane_mukul 😂🤣

This photo was taken in 1946. The guy is Ken Shimizu. He is 35 years old with two children. Shimizu never runs, sleeps late, eats whatever he wants, even drink beer instead of water. He eats dinner with many kinds of food every night. What does Shimizu do to get such a body? Shimizu doesn't have any secrets. Shimizu is the person sitting in the bottom left corner of the photo. . . . . . As for the man standing in the middle, I'm not sure who that is..... 😀😂😀😂😀

@kyle_e_walker 🎯

"Why still write R packages if AI can just do the tasks directly?" It's definitely something I've given thought to. But I've been more aggressive in developing R packages in the last 6 months than ever. Why? Because I want to know exactly what my AI agents are doing and I want to bake my expertise into the tools they use. There's an important distinction between "Claude creates something" and "Claude uses tools to create something." In agentic workflows, LLMs are doing the latter. They're not "inventing new frameworks from scratch" - that's inefficient. They are using existing frameworks, most commonly open source unless you've set up an MCP. In my consulting work in oil & gas / real estate, my projects focus on these types of tasks: - Custom-designed, high-performance mapping - Performant spatial data analysis and visualization pipelines - Carefully tuned spatial optimization workflows These are tasks I can't afford to get wrong - and getting them right commonly goes beyond what frontier LLMs can do by default. My R packages are massively valuable to me here. AI helps me write them nowadays, but each feature is carefully tested and vetted by me. In most cases, those features come directly from user feedback or client requests, so I know they're valuable. I can then feed those tools to my agents and be confident that they are implementing workflows that reflect my expertise. And as I've written earlier, I'm still a big believer in building a community around this knowledge through open source frameworks like R. AI is a massive productivity booster - but I want to make sure that productivity reflects my own expertise, not just the generic judgment of the LLM. Developing packages ensures that.

Our Quantum Bitcoin Summit last July helped push bitcoin’s quantum discussion forward. Today we're publishing Bitcoin's Quantum Readiness, a living paper on bitcoin's exposure, mitigation menu, upgrade paths, and plausible transition scenarios. 🧵👇