Data Maven with a Dash of Espresso βοΈ | Turning Numbers into Narratives | Senior Customer Insights Director | Tweets fueled by caffeine and curiosity
Australia
Joined August 2014
- Tweets164K
- Following34
- Followers29K
- Likes72K
fafa.π©π»βπ» retweeted
The CEO of Obsidian has his personal vault template on GitHub, so you can copy that setup instead of starting a second brain from a blank page.
Today's post also lists six other official or kepano repos around it. Those are existing public repos, not new launches.
github.com/kepano/kepano-obsβ¦
fafa.π©π»βπ» retweeted
Prompting Is Now the Slowest, Buggiest Way to Code with AI. Hereβs the Fast, Clean One
Forget clever prompts. Hereβs the ultimate hands-on guide to AI coding productivity, with insights you wonβt find anywhere else
aiadvances.org/the-top-1-of-β¦
fafa.π©π»βπ» retweeted
Stop Using Claude Code Without Ponytail (It Beats Karpathyβs Famous 4 Line CLAUDE md)
medium.com/@itaispector1/stoβ¦
fafa.π©π»βπ» retweeted
The Human Side of the Semantic Layer: Translating Data Meaning Across Teams
Data governance wins when meaning survives translation between teams, not when definitions get published, so test every definition against a real shared decision
moderndata101.substack.com/pβ¦
fafa.π©π»βπ» retweeted
Claude Opus 5.5 built this Blender scene, and the creator says it did nearly all of the modeling itself.
β it used the Tripo 3D tool for only about 5% of the work
β shaders were almost entirely procedural, with ChatGPT handling a few textures like text
β it even made the making-of video
Audio is from HunyuanVideo-Foley. The quality notes are the creator's own.
Real depth starts with plain practice.
π§ Custom training loops, GANs, vision
π οΈ 4 courses, about 90 hours of projects
β 4.8 from 1.7K+ reviews
imp.i384100.net/daYgak
fafa.π©π»βπ» retweeted
Like ζθεε is the "four examinations" of traditional Chinese medicine, literally look, listen, ask, feel the pulse. In English it's usually rendered as inspection, auscultation and olfaction, inquiry, and palpation (pulse diagnosis).
fafa.π©π»βπ» retweeted
Andrej Karpathy:
"The biggest mistake in AI right now: people are forcing agents to work instead of mastering the harness engineering for model first.
We made that mistake in 2016 at OpenAI and it cost us 5 years "
fafa.π©π»βπ» retweeted
Building an agent skill optimizer with AlphaEvolve, Harbor and Gemini 4 Argon @googlecloud cloud.google.com/blog/topicsβ¦
Find, extract, clean, use. Four steps your pipeline does by hand.
π οΈ Firecrawl compresses them into one API
π§ Scrape and crawl return clean output
π Built for LLM and agent workflows
firecrawl.link/s1IFcUo
Vibe coding is fun until production disagrees.
π§ Tests are how you trust generated code
π» Debug, review, document, then ship
π οΈ 3 courses: testing, security, databases
imp.i384100.net/5gOARb
fafa.π©π»βπ» retweeted
Another update to GLM 5.3 Flash for 2 DGX Sparks β‘οΈ
- KV sharding: when enabled, the two dgx sparks split context memory, so you get about 50% more room. This gives you around 3.1M total KV cache.
- Chats now save to disk and resume after a restart. A 101K token chat came back with 98K tokens cached, and the same answer.
- You can now set a minimum prompt cache. If it canβt be met, both nodes refuse to start. I also removed a false warning on every start.
- With thinking off, prompt overhead dropped from 15 to 8 tokens, and garbled repeated tool-call lines are fixed.
Thanks for the community PRs and issue reports that made all this happen at lightning speed!
Get it here:
github.com/MiaAI-Lab/GLM-5.3β¦
fafa.π©π»βπ» retweeted
I don't want to spend the day coordinating agents.
I want them working toward ambitious goals, in parallel, and I get pulled in only when a decision is actually mine.
This is how I'm running development that way.
fafa.π©π»βπ» retweeted
Abdullahi Dattijo put agentic RAG head to head with plain search across 2,000 arXiv papers. Here's what the extra searches gained, what they missed, and why it may not be worth it.
towardsdatascience.com/does-β¦
fafa.π©π»βπ» retweeted
OpenCode Browse is now available
it lets your agent control your real browser with your tabs and logins and extensions
opencode.ai/product/browse
fafa.π©π»βπ» retweeted
135k enrolled in ML math. From using models to understanding them.
β 4.6 from 3.2K+ reviews
π Linear algebra, calculus, Bayesian stats
π οΈ 3 courses by Luis Serrano
imp.i384100.net/qnQkoY
AI inference, fine-tuning and rendering moved off workstations this year.
π§ͺ Ollama, vLLM and ComfyUI in one click
π₯οΈ Up to 192GB VRAM on demand
β Deploy in about 30 seconds
hostinger.sjv.io/rErVMd
Analytics is curiosity first, maths second.
π Python, Pandas, SQL, Tableau
π οΈ Hypothesis testing, regression
β 4.7 from 1.8K+ reviews
imp.i384100.net/4P9GQG
