Researchers
Examine the methods, challenge a result or propose a comparison. The October 24 report gives the latest results with the conditions they were measured under.
Read the October 24 methods and resultsContact
Research, feedback and collaboration
I'm Jonathan Simone, solo founder of Simone Systems Research, an independent research and development company. I welcome technical feedback, research collaboration and inquiries about technical roles or focused projects.
Press: a summary you can quote at /press/.
A public synthetic Python fixture (historical commit 8374bb6ee93e) is available to run, and machine-readable results summaries are available to inspect; the local runtime remains private research software in development.
A short description or link is enough to start; please keep private files out of the first email.
Get involved
Ways to contribute to the research
Developers and reviewers
Run the public synthetic fixture, inspect the source and report a discrepancy. The review guide defines a bounded task and the information needed to make your findings useful.
Start a public-fixture reviewResearch and compute support
Help test whether prepared evidence improves a recurring task when preparation and review count. Useful support includes method review, a publishable task or compute for matched comparisons.
Examine a proposed studyTechnical and research teams
Tell me where changing documents make prior answers hard to trust or reuse. A concrete workflow can help shape the research.
Compare notes on a technical collectionProposed study
Can prepared evidence help an agent handle a correction?
Use one synthetic or authorized collection of engineering guidance and a held-out set of tasks that require an answer or edit after a source changes.
- Comparison. A capable ordinary agent with direct source access and search, versus the same agent and tools with prepared source-linked context. Use matched inputs, execution budgets and checks.
- Measure. Supported task completion, reliance on superseded guidance, and total preparation, execution, correction and review effort.
- Publish. A proposed report with the method, publishable inputs, checked outputs, cost accounting and failed cases. Agree on disclosure limits before work begins.
If preparation brings no useful advantage, that result should remain in the record and guide what to simplify. I welcome a methods reviewer, suitable publishable tasks or compute support. Scope, budget and publication requirements would be agreed before any work or spending.
Discuss this proposed studyGet involved
Compare notes on a technical collection
If you work with a document set or codebase where changing sources make saved answers unreliable, I would be glad to compare notes or test a bounded research question together. We can keep it to one collection and one check, and agree on public or private boundaries first.
The email draft asks three short questions: project or question, corpus or technical problem to compare, and any public/private constraints.
Work together
Technical roles and focused projects
I am open to hands-on engineering, applied AI research and evaluation roles, as well as bounded project collaborations with a clear question and measurable check.
- What I bring. Designing agent workflows, deterministic verification harnesses, software reconstruction from specifications and binary analysis (Baldies), human-approval gates (BigBoss), and source-linked evaluation (Consumption Engine).
- Good first conversations. A specific system you are building or evaluating, a legacy program or specification you need reconstructed and checked, or a senior individual-contributor / founding-engineer role where careful verification matters.
- What to send. Two or three sentences on the problem, the team or role, and any timeline. The about page lists my public artifacts and test records.
Other inquiries
Press and company inquiries
For press or company background, see the press summary and fact sheet or include your outlet, context and deadline. Email jon@simoneresearch.com. Public implementations are available on GitHub. I'm also on Kaggle.