I'm Javier. I spent four years doing partnerships at crypto exchanges. Now I'm moving into AI and building as I go: a for Singapore's haze and heat, a , a for LinkedIn, a for getting around, and a for Pokémon and Riftbound pulls. On the side, I grow .
Tap an underlined phrase to open the file
What this mind can do.
Every skill, and where it has been used.
| ID | Capability | Where it has been used | Result |
|---|---|---|---|
| C-01 | Partnerships | Four years at crypto exchanges, from first outreach to signed deal. | Deals closed across the US and APAC |
| C-02 | Listings | Sourced token projects and took them through to spot listing. | Including a US‑regulated exchange |
| C-03 | KOL programmes | Recruited and onboarded trading KOLs for a futures product. | |
| C-04 | AI agents | AI Job Hunter: a Claude agent that scores jobs against a CV every morning. | Top 5 jobs to Telegram every morning |
| C-05 | Automation | LinkedIn Digest: an n8n workflow that turns 2,400 stories into 5. | 2,400 → 5, daily |
| C-06 | Live data apps | Can I Run? on NEA data, and CanGo SG on LTA DataMall and OneMap. | 340 users in the first few days |
| C-07 | Audience growth | Firelit Stories and Totally Not Science, across YouTube, TikTok, Instagram and Facebook. | 20,000+ views since October |
Do people want to learn?
Firelit Stories is my experiment in AI video: 30-second shadow plays about where our habits came from, made with AI from script to screen. I started it to find out whether people would actually watch AI-made learning content. Here is what they did.
Old stage, new stage.
- Frame rate30 fps → 60 fps, so moves look smooth instead of steppy
- MotionEased movement with a purpose. Figures cross the stage instead of walking on the spot
- CaptionsThe key word in each line is picked out in gold
- PropsBigger, clearer shapes that still read on a phone
1.7KInstagram views · The grasshopper that can eat a country 1.5KInstagram views · A locust swarm bigger than a city
How an episode is made
- Step 1Find the topic
I dig through history and human behaviour for questions worth a minute of someone's time: why we fear the dark, why we stare at flames, how fire made us.
- Step 2Storyboard it
Each story is broken into scenes, one idea per scene, then drawn as a shadow play: a hook, a few beats with a date or a number, and a one-line takeaway. Every date gets checked, and where the science is still arguing, the notes below say so.
- OutPublish
The finished film goes to YouTube as a long cut and a set of Shorts. The numbers above are what happened next.
Will people learn if it's funny?
My second AI channel asks silly questions and answers them with real science. What if the Sun switched off for a minute? What if every Dennis vanished at noon? There's no narrator, just sound effects, a low-poly world and a duck who keeps pressing buttons he shouldn't. It runs on the same AI pipeline as Firelit, with the opposite tone.
3.8KTikTok views · What if the Sun turned off for a minute? 1.7KInstagram views · What if every Dennis disappeared? 62%of YouTube viewers stayed to watch · What if the Earth just disappeared?
What both channels taught me
- 01Longer can win
The 50 to 80 second what-ifs did better than the 30-second ones, as long as something new happened every few seconds.
- 02The title is the hook
Taking "Part 3" out of one Firelit title took it from 13 views to 573. A title has to make sense on its own.
- 03Keep the voice
One Firelit episode went out with no narrator, just music and captions. It did worse than the narrated ones, so the narrator came back.