@arunL4X @bishwash08 I am not familiar with how github models work. If it runs on terminal or CLI then yes you can.
English
Commodity Trader
28 posts
Commodity Trader
@Vritrax__
Nepal | Development Economist | Quantitative Trading | Futures & Options | Politics | Data-Driven
United States Katılım Mart 2026
4 Takip Edilen79 Takipçiler
@Vritrax__ @bishwash08 i can use gpt-5mini for free then ? coz i have github pro thorough student developer pack which gove gpt-5-mini and other for free
English
Tested a new NEPSE strategy. 4 AI agents vote on every stock using river levels, government contracts, news, and Twitter data. 3 out of 4 must agree or nothing gets bought.
Result: Portfolio up by 11.9%. While NEPSE up 2%. Max drawdown 7.3% vs market's 15%.
English
Check out my website: nepalosint.com
It features comprehensive macroeconomic indicators for Nepal, sourced from Nepal Rastra Bank. Along with other features too.
Nle@nlethetech
Watching Nepal’s debt rise in real time is oddly satisfying. Closest estimate we can get based on publicly available sources. nepalosint.com
English
@shit6900 It is not from google map, it is from sentinal sars-2 and 1. They usually cover entire Nepal within 5 days.
English
@Vritrax__ bro what about real time data i think google doesnot update its map every month or often,wouldn't be late??
English
Another cool project I’ve been working on is a geospatial monitoring system for Nepal that supports damage assessment, deforestation monitoring, and illegal sand mining detection. I’m still exploring its use cases and whether I can integrate it into my NEPSE trading engine.
English
Commodity Trader retweetledi
@nlethetech Thanks for the post. Himalaya AI Research Lab (HARL) is a nonprofit, community-driven initiative for building open-source, Nepali AI models and datasets. Our mission is to create Sovereign AI Ecosystem fostering digital governance, education, healthcare, and cultural preservation
English
@parya098 Damage assessment has passed benchmarking on Ukraine and Syria datasets, achieving ~96% accuracy and 93% precision. Deforestation detection reaches ~87% accuracy and 81% precision. Building a benchmark for sand mining and will focus fully on improving the project after May.
English
@Vritrax__ surely, if this works correctly then it would be a game changer!!!
English
Commodity Trader retweetledi
If you want to help build Nepal’s first sovereign AI, check out @HimalayaAILab
We’re currently building a tokenizer for Nepali to support future models. The project is open, transparent, and community-driven, with contributors from the Nepali diaspora, TU researchers, RL specialists, and industry professionals including former Palantir analyst.
Anyone can contribute. Join us and help build AI for Nepal. It’s also a great place to learn and grow, as we support each other throughout the process.
English
@MaulikShakya8 Would love to work with them. Stronger compute could further enhance the model.
English
@Vritrax__ our infrastructure ministry should be using this
English
@bibek_poudel_ The second shows damage to the Singha Durbar compound during the Gen Z protests. It could be used at detecting burned, destroyed, and damaged buildings.
English
@Vritrax__ Didn't understand all of them but the mining detection one is cool!
English
@rruummiiiiii from my website nepalosint.com . For Nepse interactions I am using my TMS for live trading.
English
@Vritrax__ where are you getting the data from? is it live fed or something else?
English
@ajaAayera I will open source the project at around may, when I make few changes. You can edit and modify as you want.
English
@Vritrax__ Sure would love to contribute if any...a Geoinformatician here!
English
@bishwash08 I also had created prediction market for my claude agents, and they predicted Nepal’s election perfectly.
English
@Vritrax__ Chalude is doing something similar
x.com/theaiportfolio…
you are heading to the right direction!
Quick question, what are token implications and operational cost? Do free tokens cut the cost?
The Claude Portfolio@theaiportfolios
The Claude Autonomous Agents have officially arrived So we're setting them up with a brand new $50,000 portfolio to see how well they do at investing in stocks Can they outperform Buffett? Here’s how the portfolio works
English
@bishwash08 I will open source the full project. But basic concept is you give the agent a trading personality, and also only allow it to use and make decisions based on data you feed it so it does not hallucinate. Specialize each agents at one field, and make them vote.
English
@Vritrax__ Interesting can you share your prompt? Do you use same agent or different models?
English
@kriitikaa_ I treat it as a pattern problem, not just tree loss = deforestation. Landslides show irregular scars on steep slopes and follow terrain, no roads or edges. Human deforestation has sharp boundaries, patchy clearing, access trails, and gradual spread over time.
English
@Vritrax__ pretty interesting! how do you differentiate between natural forest loss (e.g., landslides) and human-driven deforestation?
English
@suraj_thp Will open source, once I create a benchmark test.
English
@suraj_thp I use a satellite change detection workflow in GEE. For sand mining, we track river disturbance using NDVI drop, bare soil, morphology shifts, SAR change, and tipper detections. For deforestation, we combine Hansen loss with Sentinel-2 NDVI to flag canopy loss and hotspots.
English
@nepaliavocado @nlethetech I was a summer intern there. I worked on tracking company ownership (based on registered names), assessing tax compliance, and identifying patterns of trade-based money laundering, inspired by Palantirs’ Gotham system.
English
