ABOUT · PLAIN FACTS

About XERJ

PUBLISHED · 2026-11-13 · UPDATED · 2026-11-13

SHORT ANSWER · UPDATED 2026-11-13

XERJ is the search engine for AI agents, with a neurosymbolic model inside. Ivan Novikov is the founder. The engine is one Apache-2.0 Rust binary. This page lists only facts already published on xerj.org. It invents no founding date, no headcount and no funding figure.

TL;DR: XERJ is a new way for AI to search for data. One local Rust binary indexes your code, docs, specs and logs, plus Corpus Hub packs, and returns cited answers. Ivan Novikov is the founder. The engine is Apache-2.0. Query-1 and Reader-1 are the published models. Follow the source on any page that states a number.

Try it: the facts, one topic at a time

Who we are, in published facts

Ivan Novikov, founder

Ivan Novikov is the founder of XERJ. A 12 November 2026 message to Ivan Novikov, about XERJ Reader-1, is quoted in the Model.Surgery post. This page uses the title founder and no other office.

These tabs restate facts already published on the product page, the donate page, the Reader-1 announcement, the Model.Surgery post and the site footer. They add no new figures.

Founder

Ivan Novikov is the founder of XERJ. That is the title this page uses. A Model.Surgery team member wrote to Ivan Novikov on 12 November 2026, about XERJ Reader-1. The quote is in the Model.Surgery joins XERJ post.

This page does not list a biography, a previous employer or a credential for Ivan Novikov, because those facts are not published here.

What XERJ is

XERJ is the search engine for AI agents, with a neurosymbolic model inside. One local Rust binary indexes your code, docs, specs and logs, plus Corpus Hub packs, and returns cited answers. The longer product pages are Product and What is XERJ?.

Terms used in the product are defined in Science, including neurosymbolic AI and hybrid search.

Licence and source

The 100 contributors post, dated 29 October 2026, states that 100 contributors were on GitHub's list that day. XERJ gained 12,434 stars from 15 to 26 October 2026. The total on 29 October 2026 was 17,486. The site badge lists 18,497 GitHub stars as of 2026-11-07, from the GitHub API.

Models

XERJ Query-1 compiles a plain question into a program the engine runs over your facts. The donate page states that Query-1 is not open yet, and that it was trained mostly on rented GPUs paid for with donations, plus a few machines of our own. Every donation buys rented GPU hours for the next training program. No donation ledger has been published yet. The wallet details are on Donate.

XERJ Reader-1 is the first verifiable neurosymbolic data extraction model. The model writes the formula, not the numbers. It is in preview, by waitlist, at $0.01 per 1M input tokens, with output free. The waitlist is at platform.xerj.ai.

On 12 November 2026, XERJ announced that Model.Surgery joins XERJ to build neurosymbolic systems. The post describes Model.Surgery as a research team of AI hackers and mathematicians.

How to check a claim

If a page states a measured number, it should name the study and the source. Start with Benchmarks, the case studies and the Science terms. If you cannot find the source, treat the number as unverified.

FAQ

Who is the founder of XERJ?

Ivan Novikov is the founder of XERJ.

Is the engine open source?

Yes. The engine uses the Apache-2.0 licence. The source is on GitHub at xerj-org/xerj.

What does a donation pay for?

Donations buy rented GPU hours to train XERJ Query-1. The donate page states that the money is not used for salaries, marketing or website cloud bills. No donation ledger has been published yet.

How do I try the models?

Join the waitlist at platform.xerj.ai. Reader-1 is in preview at $0.01 per 1M input tokens, with output free. The engine binary itself is Apache-2.0 and has no email gate.

Where do I verify a published number?

Open the page that states it and follow the source it names. This page adds no new measurements.