Eli5DeFi

Ready to feed my LLMs and agents with this data!

If you noticed, most of my articles graphic were made via Nano Banana and NotebookLM, now it's time for video, 😁

Most people have no idea how fragile the global oil supply really is, or how tightly it’s tied to the macroeconomy. That’s even more true with the recent conflict in the Middle East. I used to work in an oil & gas company, so I built a dashboard that shows you exactly why: →

Work In Progress

Everyone's arguing about which AI model is the smartest. Nobody's asking the more important question: what do you actually do with the answer? You get a brilliant response. You copy-paste it into a doc. You close the tab. Tomorrow you're back, asking the same thing, because it

Honestly, with everything going on with X lately, from random algorithm changes, crypto topic deboosts, to more AI slop than ever on my timeline. I’ve been thinking about sharing more long-form, article-style content, mainly on Substack, and still cross-posting a shorter version

94% of stablecoin holders are leaving real money on the table every day. Not because they're lazy. Because the default is broken. USDT and USDC pay 0% yield while sitting in 4-5% rate environments. That gap is $9 billion annually just evaporating. Handed to issuers who pocket

New AI models drop almost every day of the week. But let’s compare Gemini 3.1 with other same-tier models—different DNA. → Gemini 3.1 Flash-Lite 389 t/s, $0.25/M input. Built for high-volume, at-scale workloads. 86.9% GPQA Diamond. 1M context window. → Claude Haiku 4.5 101

The beauty of the crypto market slowing down amid all the bear market sentiment is that I can: - Research better with less noise - Explore other topics or verticals - Try building new things like vibe coding, AI, 3D prints, etc. - Gain more clarity on which protocols have an

By 2026, Bitcoin miners have reinvented themselves as AI infrastructure landlords in the compute gold rush. Yes, more and more $BTC mining now moving towards AI infrastructure. Here's the TL; DR — Power is the choke point. With substation wait times at 5–7 years, miners’

Where tf Machi got his money. Does he have genies that print money out of nowhere? 🤣

AI builders are bleeding money in a place spreadsheets miss: API calls. You ship an agent. It works. Then it hits 10,000 calls/day and your Claude bill explodes. So you throttle it. Now it’s slower, dumber, and less useful. You traded capability for budget. The real drain

Claude, post every scheduled draft on all accounts. Make no mistakes. 🤣

— Ollama: — Unsloth:

And yes, it can also run on a phone (depending on the specs). Check Adrien post below on how to install it on your phone.

This! I’ve always said that crypto might not have been created for humans in the first place, it’s the native economic infrastructure for machine intelligence. Crypto is made for the agentic era.

If you want to download your data from X and use it to train your LLMs, follow these steps: - Go to Settings - Select Search data - Click “Download an archive of your data” - Click “Request” Once it’s ready, you’ll receive your data via e-mail.

I diagnosed the problem years ago: stagnation masquerading as stability. The job was fine. The family was real. But my potential was rotting on the shelf, and I could feel it. I was operating at 40% capacity in a system designed to make that feel normal. This isn’t boredom;

Every token launch on Hyperliquid right now follows the same awkward script. Team raises money offchain → manually seeds a thin order book → users either catch it early by luck or buy into whatever price already exists. That's not price discovery. That's a coin flip with

We are not serious industry, 🤣