XERJ is open source and the XERJ Query-1 model was trained on donated money. Every donation buys rented GPU hours for the next training program. Nothing else: no salaries, no marketing, no cloud bills for the website.
TF4Nzy4KWum9gtxXQddfTop7B52nSJsr7S
USDT (TRC20) only. Sending other tokens or coins to this address will lose them. The same address appears in the XERJ Query-1 announcement and in the repository's FUNDING file; if any other page or message shows a different address, do not use it.
XERJ Query-1 is a neurosymbolic model that compiles a search prompt into a short program the engine runs over your data and shows step by step. It was trained in FP16 on GPUs rented by the hour on Clore.ai, chosen by measured throughput per dollar, plus a few machines of our own. So far that cost thousands of dollars, not millions, and it was paid for by donations.
The next training program is our estimate, not a measured number: about 100,000 H100 GPU-hours. It buys larger training worlds, robustness to fresh question wordings, sums and thresholds in the program vocabulary, fewer declined questions, and re-scoring on sealed evaluation sets. We publish every run and every score, including the ones we withdraw.
Install the engine, add the model bundle, ask it questions over your own facts. The trace is the product.
01 · REPORT File what breaksA wrong answer, a confusing doc, an install that stalls: open a GitHub issue with the request and the response.
02 · CONTRIBUTE Send a pull requestThe engine is Apache-2.0. Fixes, docs and benchmarks land through ordinary pull requests.
03 · DATA Share evaluation dataQuestion sets over facts that carry entity ids are what counting and fresh-wording evaluation need most.
To everyone who has donated so far: thank you. You paid for the GPU hours behind the first model, and the numbers it reached are published with the runs that produced them.